The following is from a conversation with John Cook, software and machine learning engineer who enlisted our help to gain better insights on how to be effective at providing their services.
If you’re interested, book a spot on our calendar.
Book Now https://topmate.io/whizpill
Check out Alignment Lab AI on Twitter now X.
Support Whiz Pill and subscribe to us on YouTube and Follow us on Instagram. (Same handle everywhere.)
TRANSCRIPT
00:00
Nice. Cool. So yeah, so to recap, right, you're currently looking to get people more exposed to what you're doing under the AI alignment? Yeah, we can. It doesn't necessarily need to have my company's name on it. It depends on what's the most advantageous. I mean, that's why I kind of hit you up casually.
00:27
It was more like what's, I kind of wanted to see what your perspective was initially, and then more get an idea of sort of your reach in the kind of audiences that you hit. Hmm. Okay. So for me, I, apparently I use, like I mentioned before, the hosting platform I use as in the place where the majority of my content lives on, it's called Substance. It's a, most people know it as a newsletter platform.
00:56
but it's actually so much more than that. It's a place you can upload audio, video, you can do text, of course, and you can do articles. So for me, when I first started to document the intersection of arts, culture, and technology, first of all, I needed a place for me that was easy to query, but also easy to reference, so that...
01:22
If someone was to ask me a question about a particular topic that was fairly avant-garde in terms of the space and the way that which, you know, things are moving along, I could just link them the article in question and, you know, go about my day. That was the intent there. But eventually it kind of ballooned into something that I did not anticipate, which was having the kind of reach I now have on there, which is to the tune of almost a hundred thousand and counting.
01:50
in terms of the listenership for the podcast specifically. And the reason why that is, it's because natively, Substack gives you the option to download any audio that you have on there due to the RSS feed. And you can essentially have that uploaded into your podcast account for iOS specifically. I don't know if you're familiar with that, but the way it works is that you can actually copy the link directly and then
02:20
paste it into your feed and then it's going to feed you everything from there automatically, which is pretty cool. So I did not know any of this off the bat. It's a little bit of a happy accident to be honest, but it makes sense. So much of the user base of Apple lives in the US and they're also a very predominant technology company in general.
02:45
to have that kind of reach by virtue of tapping directly into the podcast, that makes so much sense. So I kind of doubled down on that to be honest. And on top of that too, the advantage that I got compared to many of my other peers in the podcasting scene is that I don't necessarily need to rely on say, you know, Spotify to have my stuff be dispatched. And also
03:12
I know who my subscribers actually are, you know, that's another added benefit. On Spotify, if someone subscribes to me, it shows me who subscribes, but it doesn't tell me how to contact them. And that's where Substack really comes in to be able to in real time announce to people, Hey, I'm going to have a Twitter space or Hey, I'm going to have this guest on what questions do you have that kind of stuff? Because most recently I actually had the co-founder of Ethereum.
03:42
on my podcast and we talked about quite a bit. So yeah, it's... So cool. Yeah, that was the intent for me there personally. But in about a year, I would say I've basically dispatched over 60 articles, 500 hours of audio content dispatched across not just Substack, but basically every other places I could manage realistically as a solo editor.
04:10
creator, you know, the whole shebang. So, doing pretty good so far, I would wager.
04:18
Oh yeah man, that's super cool. Okay, yeah I had gotten the vibe, I think I had asked you once in a space what it was that you had done because you had, I think you had alluded to some content that I wasn't aware of and I totally forgot. So well I should check it out. One, I'm interested in that Ethereum thing, I'm trying to learn now the blockchain thing like maybe a little too late but the technologies really have to merge.
04:48
And so like on the development side, it's been a lot of bots in just a weird landscape. So I have to make sure that I understand everything as well as I think I do so I don't get ripped off. Oh, yeah. For sure. Especially in this current bull run, it's going to be very essential, man. It's a lot of grifters. It's a lot of scamming. And that's actually part of why I started this. I was that frustrated by the fact
05:17
people were so easily fooled that I just figured, you know what, there had to be a way to at the very least give people some pointers in terms of like the actual voices of the space was doing these things correctly, you know, whether it's AI, blockchain or otherwise.
05:34
Mm hmm. Yeah, I think that like it's gonna take someone's gonna have to implement something that actually serves like a utilitarian purpose enough to interact with the actual markets. And then when that happens, I think the volume of Scams is just gonna have to go down because then the You know it otherwise. I think that people in general.
06:02
won't put up with it. I think there's gonna be like a lot of infrastructure stacked around to prevent it immediately and that's gonna get a lot of people discouraged. But maybe I'm wrong because there is a lot of people on that side who are just kind of hunting around. But I suppose it's kind of tangential. So do you know much about what we do? So honestly, beyond that, we follow each other. No. Do you know that? Likewise. Yeah, like
06:31
I mean, obviously I've run into you in spaces that I do know, but in terms of, you know, based on a brief overview of, you know, your account, your bio and your website, yeah, you should.
06:44
me as a software tech who's venturing into the AI space and based on the message I've gotten from you and the urgency that you were trying to press me on it seemed to me that a lot of what you were trying to address catered towards not only finding your audience but tapping in with them in as as as close as real time as possible and glad you asked me to be honest because I'll show me
07:12
in a lot of ways marketing, that's really what my specialty are. And that's what I had to learn along the way. Cause if people don't know where to find your stuff, then it's kind of missing the point.
07:26
Yeah, I don't think I apologize if I gave you like a sense of great urgency. I'm much more of a, I'm just kind of always going very fast. And I think that's because I've just been conditioned to be that way because of the pace that everything moves in the AI space. My organization kind of got started about a year ago, although we were all already doing the beginnings of what we wound up doing together. It's just the...
07:55
when you're out in the open source and you're aware of kind of the stuff that was going to start happening as things picked up and as the market kind of adopted the models. You know, after a year now, mostly everyone else has kind of fallen off and that's how we wound up as a single solid kind of company. Where I think there might be like a just a misunderstanding a
08:26
I don't really feel that I have much need to market at the moment. I've not put too much into really gaining an audience. I think the followers I have on Twitter are, you know, it's nice to be able to like platform stuff and get the word out, especially because they're all just developers mainly. They kind of built up around just the things that we've been releasing just because of the...
08:55
just the raw utility of it. I actually got quite a bit of experience on the front of community buildings. What I used to do for a living, I would build up just servers or just communities up to, I think the biggest one I did was 140,000 users. And then I would just either do it for pay or sell them. It wasn't phenomenal money. In fact, it was quite terrible. But I think I did gain the skill of being able to do that.
09:25
identify what was going to get people invested. But that being said though, I'm certainly like, not that guy now. Like I just, I really struggled to find time to be able to do that sort of stuff. Because you really need a lot of consistency in the organization. Really care a lot about the user experience, which I do, but I'm like an unrelatable monkey now because I just stare at like a GPU and sweat in the dark all day. Um,
09:56
I guess where I'm going with all that is a.
10:00
when you're making your content and you're putting your stuff out, what sort of... You have a podcast and I guess like what is the kind of tenor of that? Like who do you usually have on? Is it like normally crypto type people? AI space types? I mean you seem to know quite a bit about the AI space when we had spoken last on Twitter space and the time before that.
10:26
Yeah, glad you asked that question. So for me, the stated mission of Wispel as a platform and concept is to discuss the intersection of arts, culture and technology, because my perspective, providing insights on that intersection is going to be very important because in my eye, those three pillars are the foundation of everything we do as humans.
10:54
and making those complex dynamics easily understood will change a lot of lives and has changed a lot of lives. Testimonies after testimonies, whether it's on the podcast itself, and after the fact, you know, people will tell me, hey, you've definitely helped me ingest this information a whole lot more easier. So, and by proxy, the audience have grown, benefited from that. So,
11:22
For me, a lot of what I started to do besides making sure I have some of those subject matters is to also ensure that I'm cloning on points that are relevant to the current state of the space slash industry. For example, one of my recent episodes I was supposed to go live at some point, I would discuss Laura, the open AI of the text to video model.
11:51
that got announced and got hyped up. Because for as hyped up as it was, you'd be very surprised that people don't know about it. Matter of fact, people still don't know about chat GPT. So that's kind of like the principle there. It's just, you know, having different, very specific points. And also each of these have a companion piece article that effectively serves as a
12:21
visual aid, if you want to call it that, for what the entire episode covers.
12:30
Okay. You know, I got to watch. Um.
12:37
So I had some labs that I was just talking to yesterday. And this is what spawned me to go hit you up. And it really wasn't even labs, it was more of a gaggle of PhDs and they were all very, very smart people. It was really intimidating to pee in the room actually.
13:04
It was free, the scientific advisory board of the active inference Institute, which is a kind of smaller sort of online open source decentralized lab thing. Um, and the topic of the discussion had drifted away from the purpose of the meeting, which was to discuss this mathematical framework. And it was because someone had mentioned that they had had some anxiety about how at Stanford.
13:34
They're not Stanford, I'm sorry, about how at the college that they worked at, that I can't remember the name of, there was a shocking lack of ability to make a curriculum, which could broadly be applied so that all of the teachers could have their courses be informed about AI without themselves having to be experts about AI. Even just the act of making the curriculum was something they were finding incredibly hard to do because they couldn't find somebody.
14:02
who could both talk like a human and understand the subject. And I think that's like very indicative, right? Like of the whole space right now. There's very, very few people who are really kind of helping bridge that gap. And it's a really hard one, but I guess in general, if you're looking for like traction on that front, certainly I think people would, people are.
14:27
right now really hungry for that. Like they very much need some kind of way of accessing the information and they're not getting it. So, uh, yeah, and I'm happy. Those are just contacts that, uh, that I'm happy to put your way because I'm going to actually have to be going and finding, um, routes of megaphoning things like hackathons and stuff like that to get people more involved. So if they can start to understand.
14:52
It's also an issue of like misinformation too right now. I think there's like a lot of narrative happening and not too many people who have enough of a scope on the situation to be presenting a narrative. You know what I mean?
15:10
Yeah. It's like a lot of strong opinions. Yeah. No, I agree. It's so for me, uh, first of all, yeah, I very much appreciate the offer. And funny enough to say that because one of my episodes that predominantly discussed animation and has applied with AI featured a college professor that, um, essentially had me live during the podcast, mind you, uh, talk to their students and
15:39
talk to them as well. And we covered basically a lot of grounds, you know, like, um, but what can AI be useful in that context? You know, what does creativity even means in that context? What are you to co-create with the AI assisted tool and, you know, other soft topics, um, that was a great, um, if you will, a study where.
16:04
If it showed me that, oh, okay, this is what I met earlier when I said that this blew me to something I did not anticipate. Cause from my vantage point, I was just documenting things that I was curious about, obviously, but also I wanted to make it simple for me to reference whenever I needed to so that if I need to reference something on the fly, I can just, you know, go back to a specific article episode or number and I can just say, cool, I got that one. I'll code that, keep it pushing, you know, and so on and so forth. But.
16:34
But now it's definitely at a place where even the nonprofits are reaching out to me, they're saying, hey, your material actually pertains to what we're doing and how do we facilitate an interaction. So that's definitely something I'm keeping an eye out for. But given that I have other ventures I'm engaged in personally, the Wispel stuff, believe it or not, it's just one of the main east side projects.
17:04
So, we've got something on our hands and that's actually why I set up a TopMate site like I showed you when you first reached out to me because I get so many requests these days I cannot possibly just attend to them all. So, you know, it's been interesting. Yeah, I definitely don't want to burden your bandwidth. But
17:30
Dude, it's serious, right? Like, this is one thing too. It's like, how are people not seeing the use cases, right? Like there's more to do than like.
17:40
Like, oh my God. And be clear, like bandwidth-wise, like currently I'm not at my capacity or not even near my capacity. I'm just saying that for the sake of transparency that yeah, like these are things that I need to be aware of because so much of what I'm doing besides leveraging AI to help out in the process, it's basically me doing it solo.
18:08
Mm-hmm.
18:11
We'll definitely have it offer. Um,
18:17
Oh, I'm so sorry. I think you're. I'm having some latency here. I'm getting there like a sound every couple of seconds. I hope I'm not talking over you right now. Let me just check and make sure it's not me. Yeah, I can hear you now. OK. Yeah, sorry about that. You had the last thing I had heard clearly was you had said that you had not. You weren't at your bandwidth like limit yet, but it's important that these are the things that you know, because you're kind of
18:47
soloing it right now as far as all of the stuff. Yep.
18:56
I feel like there was probably like 30 seconds here.
19:03
Whoa. What the hell? Okay. Let me try and jump out real quick and then jump back in. Give me like a second.
19:35
Hello, hello.
19:37
Welcome back. Yeah, like I'm trying this again. How's that sound?
19:44
I mean, it sounds better. It's certainly not what it was before, which was just like a lot of silence. It's so weird, man. Discord sometimes acts up. It's because there used to be a time where Discord was actually like one of the best audio platforms out there, but I don't know, but the camera I was saying earlier though, yeah, it's, um, I'm currently at a phase where I'm essentially expanding, if you want to call it that, where
20:10
I realized that, okay, like clearly there's an interest in this Wisp build stuff. So I just started to reach out to some people to do joint ventures with. This has gone well so far, so we'll see how that goes. But as far as addressing the specific angle of, you know, educating people who perhaps don't really know how to bridge that gap, again,
20:36
That's something I very much can help in, because what I realized along the way, it wasn't just that the information being generated by the space was not captured and was not being curated. It was also that the format which it was presented in, was not really helpful. I have a whole theory about this. Oh, trust me, we're gonna get there. But to supplement this, I'm going to give you a...
21:05
reference here, recent piece I published, ugly enough about that same professor that I just talked about regarding the AI as applied to animation. And again, like that article alone, you know, got me thousands and tons of leads, literally, because people were searching for that stuff. And the website works.
21:34
It's very much designed to attract organic SEO. So that if people are searching for AI or searching for that professor's name, they will find it. So if you check your Twitter DMs right now, you should have it in there. Okay.
21:52
Man, my maybe it is me, man. My whole computer is digging up.
22:05
Jesus. Freaking. Okay, okay. No, there it is. Subsect retrospectives. Okay. Oh, this is the one that you had sent me before, yeah? No, no, this is totally new. One, yeah, this is a different one. The other one was a repository of every single one I've hosted so far, but this is a specific one that shows effectively like a summary of the conversation, what was talked about.
22:35
who we talked about on and so forth. So that one got me thousands and thousands of leads alone. That funneled back to my sub stack. So that's why I kind of took this approach personally. It was really helpful. Because if you go and Google the teacher's name, it'll pull up the article as well. And if you Google my own name, in addition to any of the keywords, animation, AI, and so on and so forth. So it was great to see that. I was...
23:05
somehow competing with like more established quote unquote news sources. I'll be very clear here. I'm not a news organization. I'm just someone who lived near to this culture of technology long enough during my entire lifespan where I understand it really well and I can also learn it really fast, so I just shared these insights.
23:35
the way myself and kind of keep it pushing. But otherwise, yeah, man, people find that stuff, they click on it and listen to the podcast as well. And there seems to be like a pretty even split in terms of people who prefer to listen to the podcast on its own, versus people who very much enjoy the article for the article's sake. So when I realized that, I definitely made sure to...
24:04
always include an article when possible and when relevant so that even if I publish something, I'm not just going to alienate one aspect of the crowd in question. Most recently, I've been tackling the more visual media side for the younger crowd. So think Gen Z, late millennial, if you want to call it that, that predominantly use places like TikTok and Instagram.
24:32
I dispatch my stuff there and my following is not nearly as big in terms of what I have on the podcast. It doesn't really matter at the end of the day. All it's supposed to do is to serve as an entry point so that people that Google me, no matter where they Google me from, they can kind of find my stuff, have an idea of what I do at a glance and organically go from there.
25:01
I think, yeah, so, okay. So you're summarizing, so are you, is this your like a stack or infrastructure for the AI kind of recording and writing down of stuff or is this like a service? Basically, no, no. So, so what I'm doing, um, I use AI tools, um, that I did not design for any of this.
25:26
Actually, I use a bunch of software to make everything that you see here possible. Discord being one of them. So typically I would host a space where, you know, I have the guest on, we talk about what's on the agenda for the day. Um, or I have them on a discord. And once that's done, I export that and I download the audio. And from there, I take it to my audio editing software called
25:56
uh, Apple studio. And from there, I clean up the audio, up up, you know, the silence, you know, when, when necessary, um, bounce that back into a website called Adobe enhance and what it does, it's, um, takes the relatively trash, uh, audio quality from these platforms and make it more listenable, make it more, you know, podcast ready if you want to call it that, um, and, uh, you can.
26:25
transcribe, not transcribe, you can feed up to hours worth of content on it per day, but the caveat is though you have to segment it into individual hours so that it doesn't overwhelm their servers because apparently they have so much demand that they kind of have to cap it. And from there, what I'm doing is I take that audio that I get back from that and I further clean that up, then upload it to a transcription website.
26:55
from which I effectively take everything and combine it into what you see as the finished product.
27:07
Fair. That sounds like a lot of work. And a lot of skills, dude. It's like you really had to learn a lot to do that, I think. Yeah, it sounds like a lot, but also it's something that if you have a very basic framework and you kind of know what you're going for, it's not that hard to replicate. It's just for me, that's what works for me in terms of using my format to my advantage because
27:35
Some of my stuff is really long, you know, like my longest episode today is about three and a half hours long. So that's really to give you a frame of reference. But realistically, you don't need to go as long as I do. You can actually, a lot of what I've been recommending people I've consulted with over time is to take whatever you're doing on tour, dedicate about 15 to 20 minutes of your time for that specific agenda.
28:04
to cover really key bullet points that will serve as highlight content, if you want to call it that. So that when people go look for the topic in question, or the topics in question, they can actually not only find you, but also have more of a likelihood to find what it is you're looking for. And I'm going to give you a very basic example of what that looks like.
28:34
So most recently, I've had the opportunity to consult with YouTubers and one of which
28:46
Um, essentially I'm not that savvy about all that AI stuff and your account came up.
28:55
Please help out. I'll take that link again. And this is going to be like a link to the
29:06
Oh, the last thing I heard was this is going to be a link to.
29:13
to a
29:16
Well, OK, this is going to be a Twitter thread. Yeah, I'm back.
29:25
Can you hear me? Yeah. I can. Yeah, this is a link to a Twitter thread and what this is, it's just me showcasing how you can marry the stuff that you already do naturally or really anywhere else to be honest and format it in a fashion so that people can find you easier. And all I had to do really is kind of propose, hey, you wanna talk about AI? Okay, here's what you should
29:54
hammer on, here are the key bullet points, and make your content from there and publish it. And it worked, clearly, because that video is currently in the recommended section of YouTube. And five days has gained 16,000 views and counting. So I'm not saying views matter in the long term. That's not the point I'm getting at. But the way you structure your content
30:24
Being strategic about what you cover definitely helps. Because if you compare that membership to their videos, it's literally night and day. That Saurav video actually has almost three times of the other videos that they have on their channel. So things like that are useful. Oh, I'm so sorry. I didn't mean to cut you off. There was just a lag spike. You're good. What's up? So are you personally like?
30:53
Are you personally motivated by? Or I guess a better question is like which? What is the aspect of this that you personally enjoy that that motivates you to do what you do? Well, I I'm just someone who's a natural networker. If you wanna call it that I talked to a lot of people in a lot of industries and I just needed a way to make it make sense. That's that's the best way I can put it, to be honest.
31:22
Because how do I explain to people, oh, I speak to a filmmaker one day and a tech founder the other day and a finance guy the next day? For the average person, that's confusing. Why am I doing that? What's the goal? Why do I provide? And podcasting was like the best way to address those concerns for the average person to be clear. Not like people that are seasoned like yourself, right?
31:50
is just to show people, hey, I have an authority to speak on these topics because I've had insert this person on and this pertains to the intersection of arts, culture and technology because AI is really disruptive now and so is blockchain and so on and so forth. So it's not so much of a motivation as it is just a natural byproduct of what it is I already do. Fair enough. Dude, that's, okay.
32:18
So let me hit you with a whole bunch really fast. And then you can kind of just get frustrated about the wall garbage. So I've got a bunch of projects in this space of sort of culture and AI for a million reasons. It's very important on the front of we need to get real translations of these languages into the models in a way that's at the bare minimum able to.
32:48
at least translate well as quickly as we can, because it's only going to become more difficult over time because the landscape is going to become more noisy and harder to organize people around. But it's the same for cultural ideations and this sort of thing, because the properties of media are, I think, occasionally extremely influential in terms of culture. And I think that because of the rate of iteration with AI,
33:18
very much like speeding towards like a huge amount of adoption. And if we don't do that first, or if we don't quickly enough get people's, you know, just a real kind of representation of people's culture into the end of the models, like we run a chance of turning what is now a way of having that instantiated into the models forever into something that is done like in bits and pieces and never really coalesces or takes a very long time to.
33:48
Um, and so that that's one thing that, that really, uh, I work a lot on, but on the flip side of that, I've also got, well, maybe not the flip side, but I also got some projects I'm working on to kind of demonstrate into the art space and the media space, uh, how the systems can be employed such that artists are not, you know, they're not taken out of the picture because I think they, uh, like AI, I mean, just like yourself, like to do
34:15
any kind of self-expression with a tool that's designed for it is, you know, it's like, it's very enabling. And it really, it really doesn't matter so much, I think that like, you can make something that looks pretty good as a base level, it's just prompting when the value of art has not for like a long time been how good it is. It's always been as far as I'm aware, like the, the person who's making it. And so if you're
34:45
interested in that sort of thing. I have some artists that I'm talking to from all over the world, and all of them have these crazy projects, and it's more or less going to be a, let's get some stuff out, let's build a thing. The thing is gonna do some stuff that'll probably, like morally be at least enough of a green light to get some researchers and devs using it, so that'll kind of reinforce itself. And overall, just...
35:13
make a system that benefits artists first for making art and do it early enough that there's that it just becomes like the bare minimum that anyone puts up with and maybe solve a bunch of problems going down the road or at least make some cool and pretty looking things. And then the other thing that I wanted to hit you with was it sounds to me like what you have is a very powerful stack of tools and techniques and skills and like personal knowledge and all the stuff that allows for you to
35:43
optimize on people's ability to get engagement. And I don't mean that in the way of like an SEO thing, but it sounds like you just really have a good workflow down for producing high quality content, which people are clearly engaging with. So if you wanted, if you're not, it is why I asked what you were personally motivated by. Cause like, I know that AI itself to research and develop on that front, it's something that takes a vast amount of time and focus and it's.
36:11
really, really painful for like a lot of people to try to do. And not that I would say that you don't understand the systems or anything like that, but if you wanted, I can totally build stuff pretty fast. And I have quite a large team of researchers and developers and we all do open source things, but I think in general, everyone's just kind of mostly trying to help the space move as quick as it can so that people can be competitive in the AI space and not swallowed by Microsoft. So we could totally.
36:39
automate your stack if you wanted, or help you kind of build out into more of a service in and of itself. And then that way you can do both things and maybe fund yourself a little bit better. I don't know. But I certainly wanted to just put that on the table. That definitely sounds like something interesting and that piques my curiosity. And again, to be clear.
37:06
None of this, right? Like the Whispill stuff, the people I've had on, the articles, you know. None of this was intentional in the sense that I did not set out to become like the go-to person, if you will, of like everything tech. Because I don't just talk about tech. I talk about so much more than that. But for the purposes of, you know, the tech currently, a lot of what I do very much is useful because what's essentially happened...
37:34
I've gone viral at first because I covered a guy that built his own custom pixel art model for AI and what happened for that was he, he's a, he's an artist like a natural, you know, professional artist turned engineer who, I guess saw a need for him to scale his business in terms of, you know, freelancing.
37:59
And I took note of that. And I also took note of how good the model in question was. And I was like, the fuck, why is no one talking about this? Um, all these so-called AI experts in this space, you know, are hosting space and all about that. So I kind of wanted to fill that gap. And I did then immediately, shit started to explode and I was like, oh, cool. Okay. This is what I was sitting on.
38:29
So I just kept it pushing. Um, and, um, the idea of automating what I do would definitely be helpful in the sense that, um, if it's a direct to market endeavor where all I need to worry about is reading the thing and it transcribes it, makes it into a video, um, makes it into, you know, an audio piece ready to.
38:56
you know, dispatch, you know, that's where that stuff really comes in. But to my knowledge, there are services that already do that. And that funny enough, I kind of believe it or not, I kind of enjoy the process of creating and sifting through that information myself. And I think a huge part of what I do actually comes down to me having like a hands-on approach to select.
39:23
you know, like what's actually worth discussing on any given day. Um, so that once it's ready to go, I can actually dedicate my full attention to that. Um, but otherwise, man, um, great ideas overall being fun around here. And, um, you definitely have me. Um, also I sent you, I'm going to double check real quick just in case, but I sent you a video of the pixel archive.
39:50
Yeah, yeah, that one. There you go. So this is the example I'm talking about. And mind you, this is episode three. We have a fourth one locked in along the way. But yeah, effectively this guy, like he's been in the space for basically forever, you know? Like even before Chad's GPT blew up, he's been involved in a hugging phase and everything in between. And a lot of these ImageGen guys, your old hats. Yeah, bro. Like they really are.
40:20
So yeah, I just found him, talked to him, I gave them a feature several times and along the way it occurred to me that yo, this is what my stuff is for in a lot of ways. It's to not so much cover the tech angle per se, but to actually provide people with more of a nuanced insight in terms of not only who built it all, but why they're building it because this is a literal case study of someone who is an artist, right?
40:50
So that challenges the narrative that only lazy people use AI, build their own custom tools and made that stuff basically worldwide. And yeah, it went crazy viral when I first dropped the first episode for Context. That podcast was downloaded well over 28,000 times within the first 24 hours. So really goes to show, man, there's like I said, there's a hunger for that information.
41:20
like properly distilling it so that people who care about it can find it. And also make it easy to share too. And that's where I think people that host spaces exclusively are fucking up big time because, um, I don't need to be on Twitter spaces to only force having to have an audience, you know, my advantage is that. Yeah, no, I'm telling you, man. Like I noticed from experience, my content I've literally got reached out to by
41:48
channels that have millions of subscribers asking me, hey, you basically have an entire outline for a YouTube video. Can we use that? And I would pick them up and talk about my personal ease and my conditions and we'll go from there. So I know other ways, my stuff has transcended the podcasting medium. It's effectively-
42:17
This is what I was identifying when I was saying that like you should build some stack underneath it because it's very much it looks to me more like it's a it's a media system. Yeah. More than a podcast. I mean, it's very impressive. I wanted to say, I didn't mean to give the impression when I said automate this. I certainly didn't mean to automate away work. In fact, the only reason that I think that that would work with you having a stack that you're deploying is the fact that you know what you're doing.
42:48
I think because in the space right now, it's almost always if there's an AI thing involved, it's a naive implementation of some sort of basic tools by somebody who is aware of a way to exploit the kind of labor savings, but not aware of how to do it well, like do the task well. And so I always said that like there's not Google would
43:15
pretty much right that there's no moat except, I mean, there's one moat, right? And the one moat is that it's just like your own personal skills and talent. Very much, I think that that kind of thing would look like you largely curating systems just like you do now, or I guess that's probably pretty sterile way of putting it, but like, you know, just doing what you enjoy and being able to profit by it more, not to try to...
43:45
Over the convention thing or anything, but. Then there was another thing you had said right before that, but I I thought was that I really wanted to respond to you, but man, I lost it.
44:10
Welcome back.
44:16
Either.
00:00
44:49
I am.
44:52
Man, this is tragic. So.
45:03
Can you hear me now?
45:06
Yeah, like something happened and I got like hard crash on here and I, but I should be back. Let me know how's my sound. It's beautiful and amazing. All right, perfect. Yeah, like I think earlier you were trying to touch on the pixel art guy perhaps and what he's done all worked on. Just to- Yeah, I mean that for sure. Like the, I mean, I think like I can't really speak at length about the high level of
45:35
Enablement that you get as an artist if you've been already to art but it is the same thing as far as what I was referring to with the stack that I said we could help you build and I Mean largely, I think you know, you know what? Okay, so the thing tonight said I wanted to comment on but I couldn't remember what it was was You had mentioned that you Really didn't intend to operate Like in the tech space you just kind of wound up here
46:05
And I think that that's a function of the fact that with tech stuff, it's like pure utility. So like the more like the transaction is truly like, I made something that's very useful. And on the flip side of that, it's people being excited for more stuff that's very useful. And so I can see that being a huge reason that you might've fallen into the space that you're in just because of the use case of people have no idea.
46:34
you know, and they really want to know. And the information is just not very accessible right now. It's a big opportunity. Yeah, it is. It is. And, um, you know, all of this to say for me, I'm at a place right now where I'm pretty happy with, you know, how things are going along and I'm definitely exploring options, but, um, I'll be very transparent here. Um, it's only, it's going to be at the leisure of what I'm
47:01
already doing not only on the front of this podcast and what it represents, but also on the front of other stuff I'm doing that I don't disclose publicly, you know? Should be. Yeah. So yeah, that's basically the gist of it. So I suppose lastly, we can touch upon like the more front facing side of things in terms of, you know, what I've done to put my stuff out there regarding the social media specifically,
47:30
interesting strategy at first. I didn't really optimize for growth, right? That was not really my priority. What I needed was a way, similarly to how my articles and my podcasts are easy to query across the board, I needed a way for that to happen on my social media front. So the way I've done that effectively is by capitalizing on Twitter and its ability to tag people and also
47:59
to search organically, using the search feature on someone's profile. So I started to effectively, as I went along, I would put out video excerpts, basically, that would not so much summarize the gist of the space, but kind of give you an idea of, like, okay, this is what we talked about. This is why we talked about it. And essentially, like the actual thread itself, it's like a...
48:27
I think it's like a dozen or so tweets long, but it should give you an idea of basically what it is so that as you scroll through, you can actually tell that, oh, you've covered those topics and those topics. And from there, when I need to, I can just kind of go back and pin it on a space if I feel like it, because I know how to find it pretty easily on my account. And I can also retweet that on the fly. And I take a very similar approach basically everywhere else so that...
48:56
whether it's YouTube, Instagram, TikTok, you name it, I can actually do that and keep it pushing. Matter of fact, on my medium, what I started to do, funny enough, was leverage the potential of embedded videos that the platform offers you as a feature. And I would effectively convert people that find my articles organically and funnel them back to my TikTok account.
49:25
At least when you're on mobile, what happens is when you stumble across the article, it gives you the option to be able to tap on the actual video itself. And if you have the app installed, it's going to take you to the TikTok app automatically. And it's the same thing for YouTube, by the way, and same thing for Spotify too, as a matter of fact. So once I realized that, I just kind of went all in. Granted, people are going to be people.
49:53
what they have on their device is gonna be a reflection of their habits. So if this is someone who's not really YouTube savvy or YouTube native, then doesn't matter how much you optimize that, it's not really gonna work because if they don't have the app installed, you know, it won't matter. But if they do though, then that's someone you can literally capture off the bat as they land on your page. And vice versa, when they land on your YouTube video, you can actually cross-link.
50:21
the original article source so that if they are curious to look into it some more, they can. But for me, I definitely took this reverse approach of I'll do the content first, publish it, curate it on the back end, make it easy to search for so that when it's ready to dispatch and go case, I can just basically blast that out. That was like the intent here.
50:48
Um, and this is a very applicable process that literally anyone can do. All it takes is deliberate, you know, um, strategy so that you're not just throwing stuff at the wall, as I like to say, and for me, man, I'm just someone who I'm very, I try to be efficient, um, even though I enjoy what I'm doing, I also know it's like colossally time consuming if I let it get out, get ahead of me. So.
51:16
That's where the intent for me came there, to be as clear as possible in terms of how do I get from A to Z in terms of results. So that works. For those videos specifically though, I use this app called CapCut. And it's a fairly powerful app and it's native to iOS and there's a lot you can do with it. And it's also AI enabled, oddly enough, actually.
51:43
To my knowledge, CapCut is one of the first, if not the first, um, AI powered tool, if you want to call it that, way before AI was mainstream. Cause a lot of the TikTok people was using it to do all sorts of things. Transcription being one of them, um, subtitling, so on and so forth. Um, using AI to do motion tracking and so on and so forth, all these stuff, you know. This is a fleet of tools that were available natively for free that I
52:11
I started to use organically for my own needs, you know. So I built up that catalog over time. And lastly, I sent you another last example of what that looks like when I take the content I already make and publish and I kind of combine the power of search I was talking about. If you click on that specific link I sent you that sends you over to my Instagram.
52:39
that post should say, oh, what the fuck is Whispel? And it's gonna proceed to explain to you. And it should lead you to like a Bing UI interface. And that's what I mean, right? Like a lot of the traffic I get literally comes from people that are using AI tools to search for them. Because when these topics kind of like come about and the way they're structured, it effectively allows people to land organically on my stuff.
53:09
because especially because of the fact that so much of it is inscribed already and properly numbered and sequenced so that if someone is looking for that specific pixel art episode that I talked about, they're going to be able to find it easier that way. And you can actually query Bing as well currently and ask it, you know, what's the summary of the articles that I've published in the last, you know, few months or so and what have you.
53:37
So once I realized that that's when I really went all in on this in the sense of cool. I know how people search for things. I know what yields the results that they're searching for. It's a matter of, you know, dispatching that. I mean, keep it moving, you know? Okay. Yeah. I mean, it's a lot of optimization. So.
54:02
I feel like I should take a class. Um, OK, I think.
54:10
I suppose just to add on to the AI enabled tools thing, there is really something special that when you find something really good, because there's so much now and most of it's terrible and more work than you're saving yourself. And I totally get it. I mean, if I had to go back to before, like the before times back before like,
54:39
GPT brought us fire from Olympus and, you know, everything started changing really fast and I learned how to build this stuff. I don't know what I would do, man. Like I like, I live on all of these optimizations. I've been building. I'm really excited. Uh, I think one of my primary motivators is just like making the stuff accessible enough so that everyone can do this. Because having like having real control over all of the information that comes in and goes out of your.
55:06
of your computer is ridiculous. Like that was really not a thing that I thought I was going to, like I didn't even consider it as like a tangential benefit of doing this machine learning, AI, LLM, whatever thing. But the level of efficiency, man, it's insane. That's probably why I sound like I'm in a rush all the time. I'm just doing so many dang things at once now, because I can automate away like so much effort. I'm actually doing quite a lot on this front for agent frameworks, because
55:36
To me, I think it's absolutely absurd that we're an entire year in and like there's still just more work to set up than they save you. Which I mean, it's just AI and it's literally software wrapped in software and you can talk to it. It's really should be like very straightforward but there's just like nobody who likes AI and UI at the same time. That's the unfortunate truth. The one case I can think of that somewhat comes close is this
56:06
relatively new startup called NVIDIA AI. Oh yeah. About that. Okay, so you know about it. But yeah- They made one of my pitch videos. Funny enough, man. I first learned about them from one of my co-spirits in the space. And I was like, okay, like an interesting proposition there. You know, let's see how it goes. So I paid attention. I didn't use the app yet at the time. But-
56:34
Most recently, they just deployed the ability to clone your voice. And yeah, once I was like, cool, like they're definitely up to something here. They're ready to scale. So, I don't know, boy, good on them, man. Cause again, like they're doing stuff that was pretty straightforward for content creators specifically that other apps have done before, such as CapCut, right? CapCut has been doing AI subtitles for quite a while and so many other AI things. So.
57:03
Having another AI entity in the video space to come in and compete made perfect sense. And what NVIDIA serves as a purpose in my pipeline personally is to generate, not generate, fetch b-roll so that I can contextualize what I'm talking about, which further compounds the effectiveness of what I've published. Because all I need to do at that point is to go back to my transcript, just take maybe
57:33
five minutes of it in total, summarize that transcript, and then feed it back into the NVIDIA prompting tool. And at that point, it's gonna yield me the video I'm looking for that gives me the gist of the imagery that is suitable for what I'm looking at. And yeah, man, worked beautifully for me so far and doesn't really seem to...
57:58
stop anytime soon, so I'm going to keep using it. And that's how you do UI, personally. You don't have to make it fancy or make it crazy complex. All you need is a clear way to get the person to the result that they're looking for. And if they know how to already do the stuff within the framework of their own pipeline, you're really good to go. So that'll be accessibility, man. So to be honest, I have yet to talk about in video specifically. But.
58:27
I very much would talk about that with people at some point, if you're open to it at some other time. Yeah, absolutely, man. Anytime. If you want, I can try to go hit them up. I think I have a contact over there. Yeah, that's dope, man. That's dope. That would be dope. Because again, having the actual guys over there be spotlighted, first of all, would be dope. But also, showing the creators of this space right now who...
58:53
predominantly again rely on spaces which makes no sense. You know, like not only you limiting your visibility in terms of what you're putting out, but you're literally wasting your time. I'm not joking. You know, there was actually a whole, like the whole reason I'm even on Twitter at all is like directly related to this very thing. So I've never been, like I don't have any social media really I have like a LinkedIn and a Twitter and I have those two things for the company.
59:21
But I don't have like a Facebook or anything like that. I've never been a social media guy before. And I joined Twitter at about the same time, like a bunch of other engineers joined Twitter because we're, I think largely there's many parties, including my own and then all the labs that we work with and everyone I know in the open source who's on the research side, who's like very concerned about how do we make sure that no matter what else happens, this AI thing goes about as right as it can go for everybody, you know? And...
59:52
we were noticing, I think independently, largely, but you know, I know a couple other people that I kind of showed and who showed me stuff that there's like these negative cycles that are occurring with a narrative that are kind of like adding a bunch of noise to the conversation and abstracting things away to like kind of erroneous foregone conclusions and you know, in general, just kind of taking it in the most American route we can where we kind of take the whole conversation, just drop it to its lowest possible level that doesn't really represent anybody's views.
01:00:22
And then everyone argues until like nobody's best interests are like looked out for. And so we joined the spaces because like in the spaces, there's a lot of people who were just like talking about AI, like they knew what they were talking about and like spreading around all this misinformation. And because of the nature of it, everyone kind of drifts between the spaces. So we were just acting like a filter for a while there and just as soon as people would come in and start talking about AI, we'd be like, okay, let's talk about AI.
01:00:47
And we would get really into it and talk about the technical stuff and kind of highlight what is actually going on under the hood and all the other nice things. And this seemed to have calmed down. So it looks like it did kind of actually work, which is nice.
01:01:00
And I just sent you another thread I've compiled together, kind of outlining the specific steps I engaged in and using in video within my pipeline when I was first experimenting with it. And again, man, it's relatively easy to do. And literally, you can take that whatever you're doing on spaces already. It doesn't have to be a multi-hour endeavor. It could literally be 20 or so minutes. And from there, you're just good to go. You just have your content.
01:01:29
You transcribe it, fire it up and feed it back into that NVIDIA AI tool and you know, what up, you're done. And you move on to the next thing. But people, they mostly care about the cloud that exists on Twitter, which I cannot get. But if I was doing this for cloud, I would not have been this far along by now, I can tell you. I've literally consulted with the actual former manager of Toolpod to give you context.
01:01:58
They have a nonprofit where they host events and they talk about emerging technologies and how they can be addressed to serve artists, whether it's blockchain, AI, the whole shebang, otherwise known as Web3. And yeah, this was someone who did not have the time of the day to figure out what was what and what I was doing effectively served that purpose for them. So
01:02:25
And a lot of it was enabled because of these AI tools. So this is where I come in in this competition of like, hey, I get it that AI is disruptive. I get it that AI is controversial. I get it that it's a place full of grifters. And honestly, tech in general is just a place full of grifters really. It's always a new shiny hype thing. But the stuff that's not just shiny hype, you know, like it's actually worth looking at in earnest and it's also worth dissecting.
01:02:54
properly so that you know what you're doing and you know how to best leverage it for yourself. That's what I think people were really missing out on and that's why I made it a point to build in public in a lot of ways, right? Like the majority of my podcasts were hosted on Twitter spaces and people have seen in real time how I do it. My biggest room has been, I think about 2,000 people give or take and that one...
01:03:21
funny enough, touched about AI. So no surprises there, right? Like AI is a popular topic, so it's gonna attract a lot of people. But point being that the NVIDIA AI tool and many other tools I currently use in my stack, they're really useful. They're helpful. They help get it done. And that's why I made it possible for other people to either ask me about it if they are that curious. So at the very least,
01:03:51
um, reference the past stuff I've published so they can see, oh, okay, this is how I can also do something like that for myself or insert whatever topic I care about.
01:04:05
Yeah, I think, I think it's cool. Like, I think it's a really, it's really, I mean, it's really an issue, right? Like this whole, this whole deal of the noise level to signal level. I think with AI in particular, just because of the speed that everything's moving, but it is interesting that the pressure is so high, kind of advancing people forward, that you can kind of, I mean, because like I'm in the same way where I just kind of fell into what I'm doing, which is, I know it's like
01:04:34
to fall into, but it really is just enabling people to kind of pursue whatever they're interested in. And if it's even tangentially related, you wind up getting a company out of it. I think that like, it just is like a related note to is people are, I think largely super intimidated by the space. And they don't, they don't attempt to like really get into it, you know, they're interested, because they're afraid of the
01:05:05
sort of lack of academic credentials or just otherwise like, I'm not really good at math or I don't do it that much. To which I say that's dumb. It's actually incredibly excessive. I mean, it's not incredibly excessive, that's wrong. It's actually super terrible, but you don't have to code really at least, to at least do like fine tuning and start to really understand what's going on. You'll want to eventually, but.
01:05:32
We, I know that like with our stack, we've got everything boiled down to just like a couple of Docker commands. And that winds up being all you need to do most of the time. That's all open sourced. But I think even like with hugging face and transformers library and datasets library and all these layers of abstraction that are built up, like it seems like it won't be too long before you're at a place and we're all at a place in general where we can just say like, I got this problem. How do I do it? And you hit a button and it just does it for you.
01:06:02
And like that's fascinating, man. Cause like that's not that far away. Yeah, essentially. So for me, from a media production perspective, what's going to be interesting to observe over time is how bigger folks use this and how effective it becomes in terms of not only media production, but also how receptive people are to it. Um, that that's a lot of what I talk about on my own podcast is that, Hey,
01:06:30
If this is the future we're headed towards, you know, like how ready people are, both on the end of the creator, but also on the end of the consumer. And I've seen some case studies here and there in terms of people, either in parts or totality using AI and publishing said content on YouTube. And these guys have earned, you know, thousands, if not millions of years doing it. And it's not the usual lazy generate me this, you know, and then, you know, press enter kind of thing.
01:06:58
It's actually fairly involved and fairly elaborate. So it's going to be interesting to see, you know, like whether or not people are going to be more open-minded over time. But even if they are it, Because like they're all going to be using it. Exactly. Yeah. Even if they are it, you know, in the grids scheme of things, you know, it's really not going to matter, you know, like it's something that's going to be a matter of, are you using AI audio, AI image, AI video or AI whatever else? I think that like even
01:07:27
I know that like a lot of the projects that I'm working on, I don't know if the only one like are tying all these things together. I think that like, dude, I think that nobody is expecting how weird this actually is going to get. And I am fully aware that I'm saying that like the day after OpenAI announces they just bought this company for, they're partners now with this company that's making like humanoid robots and all this other like.
01:07:54
we are still at the very, very beginning, like there's so much further to go. And it's really hard to articulate how confident I am on that in a way that's not walls of technical jargon and arm waving. But it's cool. I think that I think that like really, you know what it is? I think this is software is what we're doing is we're just now we're doing software. And we thought we were doing software before, but we were doing before was we were just eating bricks. Now we get software.
01:08:25
And we're going to replace everything. And it's going to be entirely different. The use cases are literally, I mean, there's no boundaries as to what you can do with it. It's not more work to make a model that does something entirely different from what all other models do than it is to make a model that just does what we're doing with them now. It's really just about free time and creativity and being clever. I know that was like a totally tangential rant, but I've been like, you know, I don't know. This is why I do what I do.
01:08:55
It's very much on topic. So to sort of like take a little bit of a pause and step back in terms of what we've covered so far. So we've covered, you know, our respective mission statements, why we do what we do, the objective of fully in the relatively near future, who we're both respectively connected to, as pertaining to the tech space and how that might be relevant, but also
01:09:25
We've talked about some of the ways we've both used AI so far. And to be honest, the last few things I think it would be worth covering specifically would need to address the social front in regards of, because you mentioned that social media is not something that you are not really that active in. And personally, neither was I, or at least not to that extent. I was pretty happy with what I had going on. I had an Instagram, a private one.
01:09:55
I had a YouTube account, again, a private one. I was not using it as a way to put my stuff or myself out there, but what occurred to me is that, cool, it is opening like a whole suite of solutions, slash problems for people who are new creators, people who kind of suffer from this, you know, AI UI syndrome you talked about earlier, right?
01:10:25
where people kind of know how to get something done, but they don't know how to put it out there. That's another thing I think that's definitely worth taking a look into as we do more, hopefully more of these chats and collaborations and partnerships and so on and so forth. Because for me, man, I really feel that the real potential of this space beyond the media being published about it,
01:10:51
boils down to enabling the people that are involved at the ground level to be able to basically be their own mouthpiece so that they don't have to rely on something like a podcast to advertise themselves. They don't have to rely on a YouTube video done by somebody else to propose the validity of what they do. They can just have like, not a ready-made stack per se, but just basically...
01:11:20
a place where they can just kind of go and self-select, here's what they need, here's what they're looking for. And then from there, they can just kind of like hone in on that specifically and dispatch content and that's it, right? It's actually something I worked on before privately with people. And I've done that in a relatively small form factor for them, mostly as a consultant role. That was really my play here. I kind of already understood media.
01:11:49
pretty well as is, you know, I'm actually a professional photographer and videographer. So yeah, like media definitely is my bread and butter. But point being, these guys were looking for a way to not automate per se, but kind of pin down, cool, if I'm someone who's not a creator whatsoever, I have zero knowledge going in, how do I do this? And helped out in that way. And I feel like
01:12:16
that's not just replicable, but it's actually something that if done well, can be as effective as TikTok is. Because I feel like in a lot of ways, that's what TikTok, that's what made TikTok successful. It's not so much that TikTok is a social media platform that caught on with the kids. I think at the core of it, what really made TikTok successful is the fact that for the first time and probably ever, like the barrier of entry for
01:12:46
content creation was so low that someone could literally get a phone, record themselves or what they're doing and post it and there you go. Like that's it. That's your mouthpiece. But in this current age though, with the algorithms and the cross platform aspects of it all, that needs to be a common core. It's going to be a little bit more involved in that. But the points still stand though.
01:13:12
A landing page that literally gets you to the steps and motions of, hey, are you a developer? Cool. Here's the format you should engage in. Are you a musician? Here's the format you should engage in. Are you a writer? You know, again, so on and so forth. So that's something that as a platform concept slash idea, you know, like it's something that's going to be really helpful for people, I think. But you know.
01:13:40
Tom will tell whether there's gonna be a need for that, because I generally don't know what the future of the space is gonna look like. For all I know, this whole thing might crash tomorrow and like all of this is basically pointless talk, but nonetheless, because of the fact I have so much experience under my belt already, especially talking to the people that are literally involved in these industries, you know, they're telling me that, hey, like I wish I could be involved more, but-
01:14:09
I don't know how to make content. That's usually the number one feedback I get. It's either I don't know how to make content or I don't know. This sounds like a business. That's a use case, isn't it? Yeah, it is. It is. Or, or they will say that I, I don't want to make content. That's like another one too. And I'm like, it's only a matter of time before someone solves that. Cause from my perspective, what I'm doing.
01:14:35
It's kinda that, right? I'm one of those people. I don't like quote unquote to make content, but to further break down what I mean by that, it's not that I don't enjoy making content, it's that the idea of getting into gear and do something that's scripted and then put it out, that's what's annoying. But if I'm doing what I already do naturally, which is talking, right? That's like my medium of choice. I can just use that effectively to then supplement what I'm doing.
01:15:04
spread my message out there. So I feel like in a lot of ways, a lot of people can do very similar things that cater to their style. Whether you're a live streamer, a video person, or like myself, an audio podcast writer kind of person, it doesn't matter how you do it, it's only a matter of how you capture it and convert the information in the process so that using AI or not in the pipeline is going to enable you to effectively
01:15:32
build your brand and put yourself out there. Mm-hmm. I think, um, yeah, it's, that's a
01:15:42
I guess, so what do you think the solve it? So if you, okay, so say you go out and you're running an AI company and you don't wanna make a podcast, you don't wanna make, you literally do not wanna make content maybe, because I know that like a lot of people in this space, they don't wanna talk to anybody, you know? They're very uncomfortable in social situations and they certainly don't wanna get up any kind of megaphone or platform, they just wanna work and do their thing. And so I guess like, what is the solve then for that? Like how do we enable that person to take the like,
01:16:12
you know, sort of general kind of interest in discussing the specifics of the topic without melting everyone's brains and boring them to death. I think that's really the main thing is like this layer of communication is very much, it's just the terms are different because like fundamentally right now it's still mainly research at least as far as I'm aware and
01:16:38
implicit in that is this just innate respect for people who are curious about things regardless of what level they perceive themselves to be at. And so I think everyone's generally happy to do that part. It's very much just the other part of like, I suffer constantly every day dealing with my command line and thinking about numbers in weird ways and visualizing, you know, odd logical systems. And now I'm an unrelatable monkey. And you know, it's probably way worse for
01:17:08
somebody who's like actually good at this, you know? And it does seem implicit. That's actually, this is one of the topics that was brought up when I was talking to these other people from these other labs is like, they were all like kind of addressing it's like, yeah, actually, you know, I also like had no idea what anyone was talking about. And now that I know what everyone's talking about, I can't explain it to anybody because like they don't know what I'm talking about. It's weird because yeah, it's just hard. Cause it's like not really good linguistically to get the terms across, you know?
01:17:37
So it's not only a content problem from the standpoint of getting people to understand at a glance what you're doing, but it's also a translation problem where people, for lack of a better term, don't really speak your language. And that's honestly part of what occurred to me as I kept on doing the podcast and article specifically. It was surprisingly effective.
01:18:06
rudimentary to be able to tell people, Hey, this is what I'm doing. Um, it still worked though, because a lot of what I doing, what I was doing the entire time besides, you know, documenting everything and making it public was ensuring that when the time came for me to specifically reference something, I could just do so in a few clicks or a few swipes and you have the thing there and it kind of works. I think that approach. I just like transcribed everything.
01:18:36
Like should I do like what Rewind does and like literally just take everything that appears on my computer and just put it in a database? Because it doesn't actually sound like a bad idea. The way you just said that makes that sound really, really convenient. Well, personally, I wouldn't say literally transcribe everything, maybe limit it to the thing you actually need. And that actually applies at scale with every other piece. That would be useful. Yeah, no. So what I'm getting at, though, is
01:19:05
If you are someone who fits that bill of you don't care about a platform, you don't care about content in the sense that you're not trying to be an influencer, you're not trying to be a content creator, this still applies to you because you could have a landing page where its whole job or software for that matter, its whole job is to record you in your process of you doing your thing. And once you're done...
01:19:35
it packages that for you and you can then move on to the other thing. Whether it's, you know, screen recording, very similar to how a BS does it. Um, or be it, um, something more along the lines of writing where you quite literally type and write what it is you want to ideate about, and then it kind of gives you a framework of like how to best address that, um, in a timely fashion. You know? So
01:20:05
that we live in the future right now because like you're blowing my mind right now. Like I hadn't even, I mean now it seems so obvious like yeah just make a website and I just whenever I'm having these kind of conversations I just let it go to the website. I just put it there and let a model like like just summarize it, drop it out, let them have access and keep a confident. Yeah actually as a matter of fact funny.
01:20:29
But funny enough, one of the platforms I used early on sort of does this already. The problem that they, at least in my opinion, they suffer from is that they're not exactly clear on like who it is they're serving or what it is they're serving. And that website in question is called Zellis. And what it does, right, it's supposed to keep track of the spaces that you host or co-hosting or speaking and for a fee.
01:20:58
or you can transcribe everything and summarize it. And once you're done summarizing, it will effectively.
01:21:08
create content for you in the format of an audiogram so that as new people speak on the topic at hand, you can literally see the stuff switch out in real time. So I think it's pretty cool that you can do that. You know what I'm saying? As a matter of fact, I'm- And that's like a local stack, so you don't really need too much hardware for that. It's like a whisper in a 7-B, just fine tune it and like diorize. But that's-
01:21:38
The real challenge though, to be very clear, is going to be to getting people interested to use it to begin with. And that's where I think someone like me kind of comes in in my insights in terms of like what role, you know, the intersection of culture and technology has played specifically to maybe like showing some of what I'm looking at. I think the case study in question should be TikTok in terms of looking at their model and how they've done it so that anyone who aims to
01:22:07
be operate at scale can do something similar. What made TikTok popular in your humble opinion? I know you're not really social media savvy, but trust me, I'm going. Well, I think certainly probably AI has something to do with it. I mean, the quote unquote algorithm sure was involved somewhere.
01:22:27
I think it was because it was quick and accessible really, right? Like that was the thing about TikTok. It was like, it wasn't YouTube. You didn't have to record a long form video and send it to desk or go get your, I think at the time it was more like standup comedy truths for the thing, but it wasn't, didn't have to be anything structured. It could just be a video that you recorded on your phone and then you just upload it. And it's like five seconds long, 10, 20 seconds long. And it's really easy, I think, when you have a lot of that to curate a feed where it's just funny people or something and you're just.
01:22:55
through like 30 second, like, you know, hamster brand clips of like quick jokes. I mean, I did anyways. Uh, I think that that as least as far as that's what people stayed for as long as they did, maybe I'm as far as why it came out like, yeah, it's just accessibility, right? Like it was just too easy to like jump right in. Exactly. So, so to, to, to further elaborate on my point, what effectively happened was that, um,
01:23:25
TikTok, not only the way to disseminate content, but also the way they create on the fly so that all they need to do quite literally is just, you know, get in there, all the tools are already built in for you to caption it, to edit it, and you're good to go. And that whole thing sparked an entire subgenre of apps where their entire purpose is to basically
01:23:53
edit short form video and tap cup to my knowledge is one of the most successful ones. Canva is a close second technically but Canva has a more broad application that goes beyond short form content but it still applies and it's no surprise that they're now a billion dollar company to be honest at least from my vantage point it makes perfect sense that they turned out how they turned out.
01:24:20
But to go back to my point at hand or in terms of really honing in the fact that Hey, um, TikTok was successful, not just because they had the store available. They're not just successful because they were easy to use. Actually, they were so successful because they made it dumb easy to share what she was doing so that if you see someone make a video and you've thought it was funny, you can just download it.
01:24:47
and it automatically stamps it for you. And you can just worry about posting that elsewhere and rinse and repeat. And on top of that, when you share stuff with a link specifically, not just downloading, it actually encourages the algorithm to keep track of who you shared that link to so that each subsequent person you shared this stuff with are essentially hijacked for lack of a better term.
01:25:15
It's weird how it works in action, but yeah, man, a very clever on their point. Yeah, you know. Genius. I was so confused. Oh, no, no, I think I just assumed you'd stop, but there was just lag. You had said, you know, we were just talking about TikTok's general kind of cleverness.
01:25:46
I don't know, but the, sorry, the more I think about it, um, this thing of setting up a pipeline to automatically kind of record these things and go put it on a website and just like make it accessible immediately without thinking about it. Would you use a tool in a, like if you, if this context were different within this conversation, would you use a tool wherein it had this feature as a sort of ancillary benefit to kind of instead building a database?
01:26:15
of your structured context, like while you're doing what you would consider to be like, this is my hobby, this is my work, this is my thing that I do every day or whatever. And just kind of categorizing all of that back and forth in the screenshots of your screen while you're doing it and whatever else. And then like summarizing that and taking notes and kind of just putting that into like an obsidian type mind map maybe or something that a model could navigate really easily with so you could talk about it.
01:26:45
And then instead of posting instantly, you just discuss with the model what should go up for the day. I mean, that was a lot to say, but in practice, I feel like that would save a lot of time for a lot of people. Because I'm considering now building that, because I was like, I will build the pipeline. And then in my head, I kept wanting to add more things. That sounds robust.
01:27:13
I get the gist of what you're hinting at. So for me, going back to my point about Zealous from earlier, if you check out that post I just sent you on Twitter in your DMs, you will see that this is kind of a prototype of what that could look like. I'm not saying it should look this way, but it's definitely in the ballpark. What Zealous effectively did here is on the back end, it takes the very content from your spaces.
01:27:43
and structures it in a way so that on the front end, you can just kind of put that out as this audiogram that people can have a general glimpse of like, oh, this is the content, this is what you do it for, cool, yada, yada, moving on, right? But what you described to me though, it's a little bit more involved in that. It's a bit of... Yeah, I'm thinking some of this stuff all the time. Like just kind of assist, like generally something to curate the data that comes in. I keep saying...
01:28:12
I think of everything in terms of pure data. What I mean to say is something that's just helping you out all day, that you can just have, not like an assistant type thing that's going to be requiring the model to know stuff, more like a semantic interface to all the details throughout the day, that you would typically just move past. It's effectively something intended to parse.
01:28:38
the good bits of what it is that you personally care about as the user so that you can only focus on those and then once you're ready to publish, you publish, right? That's literally what it is. But again, zealous, I don't know if they're trying to do that, but that's what zealous in a lot of ways kind of does already, but it requires a little bit of, what's the word here, deliberate intent from the user for them to get there on their own. It doesn't naturally encourage you to go there.
01:29:08
what you're describing does, right? Yeah, like that should be passive. You shouldn't even have to think about it. Yeah, effectively, yeah. Very much akin to how second nature swiping and pinching is on a touch screen. People instinctively will try and do stuff like that as well, because I'm telling you, as a user, forget me being a power user, because I'm technically more of a super user at this point. But as a regular user, the instinct a lot of the time
01:29:37
kind of ask the thing that you're working with, hey, can you put this here? Can you save this for later? Can you bookmark this and so on and so forth? These are very instinctive, or can you copy and paste? That's another one. These are very instinctive and sensitive features that people think about and do all the time, but no one has really figured out how to do yet because LLMs are so new. So to go back to my point about CapCut,
01:30:05
That's why I think what this company has done that's really so successful and useful. And NVIDIA also is in the ballpark, but whatever company or whatever stack sort of figures out how to somehow, um, undense all of that into a simple interface where all you need to worry about as the end user that doesn't care about being a content creator, that doesn't care about.
01:30:33
being an influencer, all you care about is curation of content, then yeah, whoever figures this out, they will literally, quite literally have. VICTOR COST
01:31:02
This is honestly, I've consulted with... Talk about it after this too. Like I think if you think that's really gonna make money and you have a voice and I got a megaphone too, I could just make that like... Yeah, man. So... It's like what I wouldn't be able to do is, probably what you would be able to do actually, because like the thing that I wouldn't be able to do is find all the annoying things and tear them out because everything I do is so annoying all the time, right? Because it's all super technical. Yeah. So I'm getting kind of excited. It sounds like...
01:31:32
At the very least, it'll be fun. Yeah, no, no. I mean, it sounds like a good project. So for me, personally, again, I've consulted with tons of people about those same topics before. So trust me when I say that none of what we just talked about is new. Someone, someone is tackling a similar thing. Well, there's Rewind, right? Right, right. But it's going to be a matter of how are you catering this to serve a specific demographic and a specific purpose? That's really what's going to be important here.
01:32:01
Forget the staff, forget the tool. If what you got going on is something that at scale functions to be in service of a particular group and does it well, then congratulations, you know, like you basically earned the money here. So, but for me, I really think that as you do your endeavors moving forward, it will be great to get into the habit of basically
01:32:31
talk to people besides spaces and get to know a little bit more about their process. It's going to give you a whole lot more insight in terms of what it is that needs to be tackled and get done because the podcast in a lot of ways functions as like a patient zero of what I'm talking about. Yeah, actually, never mind, actually never mind, it's already covered. Yeah, no, it's just like for me, I feel like it's a little bit of a
01:33:00
It showcases in real time, literally. If you go back all the way to episode one, and you kind of go in chronological order, you can see that as I go along, I iterate in the way I deliver information, the way I even structure the article, the way I dispatch the cross-reference links and stuff like that. All of that has evolved as a direct byproduct of me talking to the very people that were both
01:33:29
guest, but also both readers of the end result. So for me, it was more of a case of, okay, now that I have this information at my disposal, you know, how do I, how do I compound that for the benefit of the people that it's aimed to serve? So I feel like similarly, that this idea of, you know, capturing information, that's really where it comes in, having, having the idea, the ability to
01:33:57
in real time, as close to real time as possible. This is so frustrating with Spaceman. It's all freaking data. Like, that's all it is. The whole everything, literally the substrate that we're communicating on that. It's freaking information. Yep. It's so infinitely frustrating. It's been a year. And people have been doing this the entire year. And nobody's got it yet. Like, the only thing, like, the stuff that I've been building is hard. And I grant the, not.
01:34:25
search for the effort of a lot of people. But it's not that a lot of people still probably only 10 or 20 people at scale. And we've really done a lot. It's crazy that, for example, we have the ability to optimize the training stuff. So it's like 30, 40, 50 times cheaper. And NUS is making the models have 100 times more context length.
01:34:53
all these insane optimizations that are just totally changing the entire state of technology. And dude, how do we not have a note taker? Like how is that not standard? It's, it's, it's, so this is exactly my point, right? Like, um, if it's not standard, then don't think about building the standard, don't make it standard. Think about making it as closely accessible as possible to where someone would logically go to use it.
01:35:23
so that if they're, say for example, then yeah, like a build that as a proxy to what TikTok would be. But if they're more of a writer, then iterate towards more of what the writer would use. But again, it's gonna be like a case by case study and it's gonna require experiment. But when all this hadn't done though, from a perspective of just information sharing, yeah, it's something that I think there's a lot of value in. And I know this because I'm someone who literally works with data all the time.
01:35:53
You know, like when I break it down, my podcast is raw data that I'm structuring and formatting in real time. Yeah. Text and the platform or rather, I guess, application that's really useful for me to do that it's called the eScript and what it does effectively is, you know, convert the audio into text. Um, and then I can edit the text and set text when I edit it also edits the audio so that if I want to delete.
01:36:23
Keyword or key terms or even a whole paragraph is gonna delete the actual audio that this paragraph is attached to dude That's huge. That is so big I've been doing everything like manually Yeah, literally like literally and and and Granted to be honest what annoys me about that fucking discrete tool is that for whatever reason? a they don't disclose or properly anyway
01:36:52
When you sign up for their service, there's like tiers involved. You kind of have to sign up for it. Yeah, it's. You want to.
01:37:17
Yeah, I was annoyed with the fucking noise. I needed to shut that down for a second. But yeah, so when all is said and done, dscript, this is just a tool that effectively allows you to take audio, convert it into text, but set text has real impact on the audio file, the source audio file. So that now, as I do my work, typically, I can basically
01:37:46
take all the arms and the arms and the arms and the legs and all these annoying filler words. And on top of that, if I feel that someone is getting a little bit too long-winded for a particular segment, I can just basically edit that out and keep it pushing. What's annoying about dscript though is that it seems to me, because I've used it as a power user, once you work in the format of several hour long files in conversations, you know, it falls apart very quickly.
01:38:15
It doesn't seem to be able to handle that kind of thing. The more shorter stuff, the better. I know why. Yeah, so I bet what's happening is on the, oh, I'm sorry. Yeah, go on. It was lag there again. I was going to say, on the back end, what's happening is they're probably running Whisper or some TTS or STT model. And to get them to be able to align the words with the second level timestamps so that everything's perfectly in line.
01:38:45
It's a matter of you can do. There's like three things you can do. You can spend a ton of money on a ton of compute and blow it out and do it that way, which lets you get some context in there. Or you can blow the model up and just not work like it just won't work. Or you can do a thing called monotonic alignment where you take a very specific model and you.
01:39:13
Well, I have to I'm trying to like, break this habit I have getting like really deep into the technical side of it when it's not appropriate, but like. I'm betting because like with monotonic alignment, the quality of the model should it like it stays consistent forever. Like you can do it for 1000 hours. Yeah. So I'm betting they're blowing it up because if it's you're only doing like a couple hours, like, dude, you burn through like on Google collab whisper will burn through like an hour of audio and like a minute and a half.
01:39:41
If you're doing it well, like if you're doing it like that. So it could be that this is the case. Or it could be they actually have an entire model doing everything, which would be insane. But maybe they do, because like you never know. Someone who's very familiar with audio and audio engineering and the whole pipeline of how that stuff works, I can definitely tell that on the back end, what they're doing is that they're processing the audio in whatever format you feed it in. But.
01:40:09
What you get back is definitely something that highly relies upon the text output. But for it to process the text as you not only edit the file, but also as you trim it down for your needs, it's actually basically restructuring that same audio file in real time as well. So it takes a lot of compute to do to begin with.
01:40:37
I know this because like regular. What did you say? What's the resolution? I'm sorry, what's the like? The cause the part that's going to take the most is going to be actually literally editing the file, right? Like it's going to be the actual computer to do that. So I'm wondering if there's not a I just know a bunch of tricks for optimizing on weird data. And so like I might know something that might save me some time. Yeah, so for me, I really think that.
01:41:03
It's definitely not a, it can be a resolution issue in terms of like what format of audio you use. So for example, if you use like say WAV compared to FLAC or MP3 or M4A, you know, that stuff definitely matters. But the ultimate point being that if you're working with that format, ideally for, at least for what this script currently offers, keep it at most to 20 minutes.
01:41:31
20 minutes is low key pushing it. Like, because I can, again, I can tell on my end, like it's just taking time to process all that and yielding me back the results. So I found that the sweet spot is like 10, 15 minutes, you know, like if you, and that's actually perfect for people that make YouTube videos, because now all they have to do is take the original audio file, feed it in there, and just, you know, have the text ready to go, and then...
01:42:00
edit whatever they want to edit out and you know boom bada boom you on your way and it's actually even more perfect for short form video. Now as applied to short form that's when stuff gets really interesting but it's gonna come down to again like what's the intent here, who's the demographic, you know what's the purpose, all that stuff. That sounds basic when you say it out loud but you still have to think about because if you don't what you're gonna effectively do is effect...
01:42:28
to give a suite of tools to people that they don't either care to use or don't know how to use. And in this day and age, you know, being specific about what you're offering is going to be paramount. So that's my two cents about it. But honestly, beyond that, man, we're basically at the top of like the two hour marks of, you know, this whole chat. Personally, I've had fun, you know, talking to you.
01:42:57
I very much look forward to talking to you some more at different times about other things. Very much happy to do that. And I'll be fairly transparent here. Even though I kind of have, you know, things going on and, you know, I have stuff in the chamber, I would definitely would want to spare the time and have like a proper podcast episode with you. Because I think like, a lot of what you're doing actually would be very suitable for the kind of audience I have. Because they're very much...
01:43:26
They're not techies per se, but they're very much the kind of people that care about what's currently going on and why they might want to be into it. Or rather, why they might want to educate themselves to use certain tools. You know, like you sound like the kind of person that knows how to really break it down like that. And I could potentially facilitate, you know, ask guests I've had on the topic before, and I'll do like a super episode effectively of like
01:43:55
all the AI powerhouses. So, you know, just for thought. But quick recap, though, we've covered a lot of material today, surprisingly, man. We've covered, you know, like, you know, social media distribution platforms, NVIDIA being one of them, CapCut being one of them,
01:44:24
If you were to go that route, how to best assist the user? So that, are you trying to be a curator for their content? Or are you more so trying to be a record keeper of sorts? That they can kind of go back to and consult later on for their own needs. Or do you want it to be more of a distribution platform so that the creator can then dispatch them? Oh man. There, you know? Like, so- Yeah, I think I want them to be
01:44:53
Enabled like I just want to control the information right like that's all I'm trying to do. I think that's things like make it easy I'm so sorry. I was like totally muted when you were talking to me, and I think he just it was probably Like I was ignoring you, but no yeah absolutely. Let's have another conversation. Let's let's chat in DMs and stuff. I want to I'm happy to do a podcast. I would love to I would love to do that I'll totally tell you more stuff about us and the kind of things that we put out probably if you have like people that are kind of
01:45:23
Like that space of like not quite super, like they're not, they're not PhDs. They're not even really in college for this kind of thing, but they're techie because they're in the AI space and they're running the open models. Like those guys, like that's our domain right there. Okay. Sounds, sounds good, man. Um, lastly, if you don't mind checking in your email at the moment, there should be a link in there that, um, gets you to a portal while you leave the testimony, could you take a...
01:45:52
15 to 20 seconds to do that real quick while we're still on the chat. Yeah dude for sure. I don't see... I got one from DoorDash. What's the address it comes from? So if you search topmate, it should yield you the actual email it sent you when you first booked. Got it.
01:46:21
So they should take it with them on each.
01:46:28
video calling. Probably it's not that one.
01:46:35
Thank you.
01:46:40
This man, I'll check this back here.
01:46:49
College base, advice and answers from the Top 8 team, not really what I'm here for.
01:46:55
Expert missed session. Yes. No. Yeah, I don't have the link yet. Probably it's going to wait until after 12 would be my guess. So another three minutes. Yeah, so it can actually get you there. Okay. Well, nonetheless, yeah, be sure to do that because a lot of what I do. I'll make sure. Yeah, thank you, man. So yeah, because a lot of it actually comes down to literally show people, hey, like, I'm not just some guy like ranting about these topics. I know what I'm talking about.
01:47:25
So no actual people who was doing real shit in the space like yourself. So that's where this stuff comes in. But otherwise, man, this was a pretty tough combo overall. And I'm hoping because again, when he first hit me up, he generally felt like the FBI was at your door and like about to get down. I'm sorry. I'm just like always working. So I'm always like in a rush. Was it that intense? Yeah, really.
01:47:52
Oh, from my vintage point, yeah, like I was telling me because it felt so casual. I was, I was sent a barrage of paragraphs and I'm like, Whoa, okay. Like sounds like this guy, a lot of urgency. What is going on here? Um, yeah, I typed really quickly. So I like, I don't even think about it because I'm like all day I'm coding and like, you know, I don't like to take my hands off my keyboard to even navigate my computer, so like it's very little effort for me. And I think.
01:48:21
you know, it comes off as loud, my bad, Nathan. No, it's all good. I'm not annoyed by it. It was just me kind of like in hindsight realizing that reading tone via text is like so difficult and you never really know like what the headspace someone actually is when they talk about these things. But yeah, like that's, at this point in time, man, considering me a resource, like I said, I can't...
01:48:48
do my best to assist in the meantime with the whole scheduling my own time and my own also video sessions I'm supposed to do, but beyond that, there's a lot. Yeah, I like to do asynchronous. So like, I just like to lead like when I find DMU, like don't think I'm expecting you to respond like with any kind of time. It's usually it's because I'm info dumping like in between meetings. Yeah. Because I'm actually going to go into one here in about 10 minutes. So that's like my whole life right now.
01:49:18
And to really wrap this and top this off and go ahead and basically state your name, who you are, what you do, credentials, if that even matters and kind of like give a brief summary of like how you felt this conversation has helped you. And also what your key takeaway was and will be done for the day.
01:49:46
Um, yeah, I guess I never really did give my name out. My apologies. My name's Austin. Um, it's John Cook on my website, which is alignment lab.ai. I'm a developer of cool AI stuff and tools. And, um, I think my key takeaway here was that everything I had reached out for was 1000% already done, which is great because I don't think that I even was able to clearly articulate what I was.
01:50:14
wanting to talk about, I just knew that you kind of had the stuff and clearly you did. So I most like appreciate you for being so organized and kind of on top of everything. Been a pleasure. You're very welcome man. I'm on that note signing off and catch you around later man. See ya. Be safe. You too man.
[Whiz Pill Podcast 55] Consultation Call With Alignment AI