#160 Exploring the Uncharted Territory of AI with Garik Tate

0:00 From the heart of Dubai where tomorrow is being built today to the world, welcome to the CTO show with limit here. We redefined technology and reimagine possibilities with mehmed delve into the riveting Realms of AI cybersecurity and digital technology experience the thrilling highs and lows of startups. Immerse yourself in the spirit of Entrepreneurship and witnessed the future of business. Innovation, being written in real time. Now, without further Ado, let's tune in and explore the future. Hello, and welcome to a new episode of the CTO show with mammoths.

0:48 My name is Muhammad. And I mean each episode, as you know, I cover different topics from emerging Technologies, like, artificial intelligence, AI, cybersecurity, digital transformation, Quantum Computing. And also at the same time, recently, I'm doing more interviews with subject matter, experts startup, Founders and entrepreneurs. And he today, I have Gary who's joining me garak? I would let you tell the audience. What are you joining from and what you do? Yeah, thanks for having me met and so happy to be here and I'm calling in from the Philippines. So, I'm originally from the, the USA. My basis is here in the Philippines and I, my AI futurist and CEO of a hollowed-out team at Valhalla.

1:34 We specialize in working with Purpose, Driven tech companies, helping them build out their development teams and their products while leveraging the power of AI. You could say, our expertise lies at the intersection of a AI IQ and EQ. Wow, that's fantastic. Now, in the first place, how did you first become interested in Ai and what led you to become an AI futurist? So, the I think it all started when I was quite young as, as most things do, there was a I was thinking about this guy. I get this question asked normally and in the very beginning my answer was kind of flat.

2:17 Like, I don't know. It just was always obvious. Like it was, it was the thing. I don't know what sparked it, but I think, I think it started. And, you know, same way that a lot of us, kind of, nerds get into things that we start with sci-fi started with the books, we read. And, and There's booking a recursion and it the more I dug into this idea of building intelligence from scratch. The more I started obsessing over it on pretty much every level you can imagine because It strikes me as a very Grand project, that is really at the very tip of the iceberg of everything that it means to be human for the last many thousands of years and were, it's pushing the limits of our Hardware. It's pushing the limits of our philosophy. It's pushing the limits of our ethics. It's pushing the limits of our industry and it's all coming together in a way that just makes right now. A very exciting time to be alive. And so I have been obsessing This for as long as I can remember, it looks like a long.

3:22 I would say, help session. I can see this also get it. Now from what you are seeing, what is the current state of AI and because you mentioned, you work with a lot of your customers to see how they can implement this. So, where do you think the future holds in terms of, you know, future of business, and the impact on society? Yeah, there's a fantastic quote of the, the future is here. It's just not evenly distributed. So so a lot of incredible Innovations have have arrived.

4:04 And I think we're entrepreneurs gonna be making a lot of money, is the same way they've always made money, which is going to be on The Cutting Edge and they're going to be playing Arbitrage with the future. They're going to be figuring out where the puck is going in there. I'll be moving there there first. And and I'm advising companies into thinking about how to predict what's going to be happening with AI and where it's going. There's quite a few models I give them but but one that I want to share here to get started with is that the the closest analogy of where we are at right now to history. That's always a good foundation. It started with is at the turn of the 20th century as Electricity was being distributed throughout civilization. So think of Thomas Edison and light bulb think of the the power grid throughout Europe throughout the this throughout the west where the transformation meant that any technology we had before, we could add power to it through the form of electricity.

5:13 So a hammer became a jackhammer a, you know, No a a carriage became a car originally. The first cars were electric a screwdriver, you know, became a power tool. So on so forth, and what we're doing right now is something similar where. At this point we are adding intelligence to just about anything that you could imagine and because the intelligence at this point has an advanced, the point where it can emulate human communication. Literally, let the answer is it, like everything is going to be changed. So what I suggest people to be thinking about, is to be taking their subject.

5:56 Matter expertise, and this is, you know, same with most Innovations, right? Is it. Take your subject matter. Expertise takes what you find unique, what makes you unique? And to figure out how a i is going to be entering into it. How is what you're going to be doing going to be augmented by it and If that is a little bit too broad than the the way that you break it down, is you think through some sort of value added chain, I'm going to assume that you either an entrepreneur that has a business or going into a business and you break down that that value chain of what that business would look like, or what that process is.

6:36 And then think through every step of that value chain is both adding value. But also subtracting value because costs accrue friction is added. Miscommunications could be added but Also theoretical value is added. So, you know, in e-commerce to there's a point where they see the advertisement and they click on it, then they put in the order button or so on and so forth. At every one of those stages, you can add in different forms of AI, to increase the efficiency and decrease the, the cost. And I guess I'll leave it there for now. I think I mean there's also some interesting comments about what are the recent Innovations in Ai and what's coming down the pipeline? But I'll, I'll leave it there for now. Yeah, we come back there but just, you know, like being the devil's advocate here, right?

7:26 So someone might say, because you mentioned a very important, I would say phrase there and you said it's like the moment is here and, you know, you did. Similarity with previous things. Now someone might say getting but you know the AI Is Not A New Concept. AI has been with us now for sometimes why now all the hype is happening around AI. Is it because You know, the generative AI what we are seeing with tools, like, Chad GPT from open-air. I bought from Google meet journey, and all these tools that why the hype happened now and not before, knowing that, and I covered this in one of my episode of the history of a i AI you know, the concept goes back to very old times and then Computing I mean in modern age has its start in the 50s.

8:25 You know, in all this movement that happened Din the universities across the u.s., but why now is the moment of AI? So in in my opinion, that there is a certain level of arbitrariness to it because the technology that underpins a lot of the recent, you know, gizmos and things that are being distributed masses, to be bar, chatter BT, Etc, were based on an innovation in 2017 from a paper called. Attention is all you need. Which was kind of a precursor paper to the Transformer. And so, the Transformers, it's a fairly technical concept that was an innovation on to our existing, you know, understanding of how to use neural Nets, and how to use machine learning and the fundamental, well, actually, I'd be very curious momentum.

9:20 If you're familiar with Transformers what you, what, you perceive your perspective would be on what they added to the game, but From what? Has been impactful for me about Transformers, is it? They unlocked a much greater degree of parallel processing, amongst other things which basically meant that we could be. We could do horizontal scaling with our machine models in a way that we couldn't do before. So all of a sudden it became a lot more tenable for us to consume 10% of the internet in order to create a model, like gbt 3.5, which was the chat, CBT and, and it's still billions of tokens or however many it was.

10:07 So, the Transformers had a lot of innovations. That was one of them. Another was that it Transformers allows you to chunk a piece of input data and to add you consume that input data to tell its context. So so what I mean by that is the original paper attention, all you need was was focusing on how to translate languages. So the example I gave was from from French to English and various phrases in French, you know, will have Different sentence structure so it's not one-to-one.

10:47 You know, translation you'll have Yoda speak and the this the grammar structure will be will be thrown off. So the the Transformers were able to more efficiently discern that by being able to tell how the words existed and came together into sentences. So, long story short, is that it could, it could consume larger pieces of data and approach something much closer like human. Ish understanding. I wouldn't say that they are anything like human consciousness but they could, they could better understand larger volumes of data and parse the patterns between them. So those were the two things. So why did it not launch off in 2017 late 2017?

11:30 When, when that paper came out, you know, I think it took about For four and a half years before enough Engineers really understood the potential and really started to build out the infrastructure to make chatty Beauty. And then the quote-unquote hype has been that Chad gbt was released and now anybody had access to it for free. They took a gander with having a billion dollars of funding from Microsoft. Now most of that was in server server tokens but putting that aside they had a big investment enough To go Mass public and you know, the future is here. It's just not evenly distributed, people didn't know this technology. Really technology was out there and so it took that long before it hit mainstream and here we are. Yeah, I remember and this is now I'm not the devil advocate anymore.

12:22 I am who I am. Usually, I think I started to see like discussions online and then, I can't remember exactly how Came up with because at first there was the API of what people knows now, as Chad GPT. So yeah, open AI, they release the closed, beta or closed platform. And at that time you know like they put some use cases there. And I was thinking like okay this is really interesting, I'm not a coder but by any means but of course I understand I have the knowledge. Okay so they did then API that you can give some input and then It gives you some output that's cook interesting. And then you know in December or November I believe end of November when they release chat GPT right? Oh man.

13:17 Like like yeah. This is something that we never saw before and now to your point Gary and this way now I'm not the devil advocate and I repeated multiple times y. It is the moment because you know, the anyone that is familiar with computer science. knows that, you know, natural language processing and knows that, you know, having ability of a machine to understand the human like it's like the Paramount I would say theorem Theory whatever you want to want problem that everyone wanted to solve and you know this breakthrough because You know, I'm like you interested in AI from long time and the key for doing other things with a, I start by letting the machine understand actually, what you want. So, if I can let the machine, understand what I'm thinking about, I can later go and tell the machine to go and do it, because it can understand me now. Take taking this from this point, I want to understand from you and for the audience, of course, how this will affect the future of jobs careers and our lives from your point of view. Yeah, I think that there's as always is with large amounts of change.

14:44 There's a lot of fear, there's a lot of uncertainty and it's becomes quite hard to to, you know, for individuals who haven't been thinking about this stuff to predict what's going to happen next and I I think that the, the two things that I talk with people about who are worried about AI taking their jobs, disrupting their livelihood, is that first of all, AI is not going to replace. Anybody it's instead, it's going to amplify some people to be so hyper productive that they're going to replace people. So, in other words, your jobs not only taken by an AI. It's not be taken by someone who's using a, I better than you.

15:31 Yeah. And so in some Industries there are they are more constrained by Supply than demand what I mean by that is if if you talk with people in the tech world, you know, we kind of need ten times more. More coders than we have right now. Like like this, the constraint has been on the supply side. We want more code. We I would also argue that we want more art in the world. That there are a lot of things that we have taken for granted the supply and demand curve are where they're at, but give it a few years and we're probably going to really find that the demand has increased exponentially but of course, there are other roles one that comes to mind is I believe McDonald's has recently started testing I for there for taking orders through at drive-throughs and things like that and so it's like is that going to increase exponentially as well? There there are some historical examples such as with bank, tellers back in the 1980s.

16:37 When the ATM was introduced. We thought bank, tellers were going to go out of business. They they were gone drop down by like 1/5 because 4/5 of their time was spent just punching papers and, and doing things at ATMs can do, but when ATMs were popularized, in fact, bank, tellers jobs and not go down there. She went up because then the bank's wanted to create more Banks, they want to create more outlets. And so the bank teller job was not was not affect the way we expected. So I think there's going to be a lot of surprises in this journey, but By taking advantage of this technology and not seeing as a race with Silicon Valley, but instead, just erase with your industry, as you already were doing, you already were competing with people in your field.

17:26 Now it's just there's, there's a new, a new tool in the tool belt. That's a much healthier way to look at it. The other piece of advice, I give people is that when you are in a state of fear, we know this with with Neuroscience that the blood Pools away from your neocortex and your 32 percent, less capable of making creative decisions and and higher-order executive decisions. And so I really encourage people to You know, it's real advice to look on the bright side and to get excited for this technology is going. If you are if you're scared, if you're playing on defense, if you're going on the passive and try and like whether this out, then you're not going to be at the level that we really need you to be.

18:16 And so those are generally how I think about most people's tactics for, for weathering things. Yeah. On this point. What? I would say, like, feeding complaining, I don't know, like, all this work that I can think about Shawn - would not change anything. Like even we saw some of the big names even in Silicon Valley. They signed a paper for stopping, you know, the research and on AI, some people ask me, what do you think? I said, okay, look if openly I or Google, they don't go and make it, someone else will come. Make it because, you know, it's matter of as we call it in the tech world.

19:03 And, you know, in the Consulting world, you need a proof of concept, right? So you need to proof of concept. And once you have the proof of concept, succeeded. So we know we can do this the Genies out of the bottle, so to speak. Exactly. So if Mammoth will not do it, get it tomorrow. We'll go and do it. If Gary doesn't do it, someone else will come and do it because we know that it can also collecting Yeah it's gonna take some time with them. Probably I'm not sure. But with all the processing speed you know like building a large language model. Yeah. I think it's time it needs all the city's resources but you know like I believe in six months time we can see another open a i for example, right?

19:43 But the can yeah the same thing. Have you followed the the leak with with a llama from from from the the Facebook a I have you followed this story? Oh yeah. Yes. Yeah. So once that so Book had their own large language model. They called it llama and it was it was leaked out. But something really interesting has which I did not predict occurred which is that once it leaked out to the masses, the open source Community has taken it and run with it and they have found Justin certain efficiencies of how to create new large language models with with using the Llama framework that are orders of magnitude more efficient.

20:25 So while it cost a Million dollars in processing to create chat CBT. We are, we are now creating new large language models that are as good as G BT 3 not as good as GP for yet, but are as good as G PT 3, which it was is pretty much good enough on orders of magnitude less. I think Google came out this paper chinchilla which, which had some of the major breakthroughs that helped the open source community and then And there's a few other breakthroughs, they had as well. But at this point what from what I'm seeing is that people with a beefy enough home set up and a few few days of downtime can be creating these large language models, which I do not think I thought the constraint was going to be a half a billion dollars in a server Farm.

21:18 But this point true that the genes outside the bottle and if anyone's interested in that, I would suggest the Google paper called I think it was called We have no moat and it was a leaked Google Document. If you Google we have no motor Google and neither does open a.i. you'll see a link Google document that goes in a lot more detail on that. Yeah. So surprisingly and, you know, this is one and information to give to the audience if you are watching or listening to us. So it's my, it's my top hit episode. Actually on the show, it get the highest number of listening and I didn't put on YouTube because was a solo recorded without camera, but you know everyone. Started to also ping me about it when I covered Auto GP T and this is exactly, you know, like they used actually the Llama open source, right? So so to do a few things and it's interesting. Now you mentioned getting something, and I'll ask you this question, not from technical perspective to just also shed some light on the, you know, like the future of work that we talked about.

22:21 Now, you always separated between GPT 3.5 and GPT for, right? Yeah. The reason is and I did also a like a dataview kind of between birth and GPT for and always I repeat myself that to me GPT for has the ability and even if you have access to GPT for today, you know, like they are not sure if they still show it or not open. They are they show you something like a score for the model? So for example, GPT 3.5 is fast GPT for like Less fast. I mean slow down a little bit and then you have reasoning, right? So the reasoning part. Now from an AI futurist perspective. Like, when we say reasoning, we mean someone that has the ability to come up with something that doesn't exist with a.

23:18 I be able to do that as a great question. You know it says it's for this one could guess what? So so I guess I guess bottom line I would say Yes, because we just see how fast were advancing. And how many breakthroughs were making? How far is that in the future to come up with true creativity? It's hard to say. But I will, I will talk a little bit about what's the gap between where we are. And that point so I think we're going to cross it. We're probably on across it this Century but I want to talk a little bit about what that Gap is.

23:59 So if you go to chat CBT and you ask it, What's two plus two? It will give you the answer for. But if you ask it a very complex math or not complex but a very long math problem like what's, you know, five thousand trillion 400, you know, so and so forth times and equally large ridiculous number. So number. So it's a math problem. That is not very popular online. It will give you a wrong number. And, and if you punch into a calculator, you will verify that the wrong number. But the first few digits will be correct. The last few digits will be correct of the new number. Yeah. So what what it's doing there is, it's recognizing patterns, it's not actually calculating the numbers in.

24:44 This is kind of a weird, a weird thing to get your head around that when you're asking these questions, it's not reasoning through the problem. It's not calculating it even though it's a computer, you expect a computer to calculate it for, for Christ's sake, but it's just recognizing patterns in the internet. So if you ask What's two plus two? It will have millions of examples of people saying two plus two equals four so it knows the answer is 4. But if you ask it a question, it hasn't seen before even a math question that you could reason through with enough time it will give a wrong answer but it will get parts of it correct. Because it has already recognized a pattern that to get the first few digits.

25:25 And the last few digits of a very complex multiplication, it only has to look at the first few and last few digits of of the two What do you call them component? Component numbers. So, it is the pattern recognizing has gotten us further than we thought it. But it only has the illusion that actually has a model of reality. So the way that I think about this is that they're us humans. When we are reading through things, we reason things both bottom up where we have empirical historical data of what we've seen and and we have examples, but we also think through things top down you can Imagine something you've never seen before. Yeah. And right now, these Technologies, they and I would encourage people to play around with Chad, you beauty is do, is you punch in some of these numbers? It's quite Illuminating.

26:18 What this technology is, is it, it has a illusion of top-down thinking and it's, um, it's done that only through bottom up. And so, I think that as we are developing Better and better Technologies for these a is to have internal simulated versions of reality. So we're creating like these virtual Minds like physics simulations and other things that are a little bit more obeying, the laws of physics. So to speak, I think we're going to get closer to an AI That's able to do something a little more top-down and that's going to be quite exciting. I think that's going to be what the next few Innovations are. Transformers, they've taken us further than we ever thought they would but they are Going to peek out, and we need a few more Innovations.

27:06 A few new truly creative new Innovations and no one's talking about yet. We need a few more of those before really going to crack that problem, but the question got it is, don't you think that always, we need someone to punch this as we call it prompt that who gonna give the prompt today? I is it like an other AI that is giving that prompt how it will work. So you're asking in like a hundred years from now. What's going to be the process of maximizing a eyes utility? Yeah, that let me give you a more like straightforward example.

27:49 So assuming that a, I will have the power to come up with completely new business model or a completely new. Software. Let's make it more simple. Okay, so someone need to tell whatever, you know, the model they are using that. I need a software that I believe it should do. One, two, three, four. And then the AI will go and write the code. They I will do the UI they I will do the ux but who is giving this ignite to start you know the process. Yeah, you know. So if we're talking about a I having greater and greater impact into the real world, so it was like the idea is you ask, Chad, your beauty make me a million-dollar company. So it's based like you know find a way to put a million dollars in my bank account right now and make it happen. Yeah.

28:42 And you if everybody has access to that then then a lot of things are gonna start to break down at least in terms of like what is the definition of some of the things like if if everybody has infinite money than money is less valuable, right? And we're still in a physical universe. So you know what we're aiming for is a post-scarcity civilization where where you know, people are all given enough. And and, you know, I think that Star Trek Vision still is at the heart of a lot of nerds like you. And I trying to make the world a better place, but, you know, as an AI futurist. And I'm looking at this stuff, practically we Need to keep in mind that there are these Technologies are existing in a physical universe.

29:31 And so if we are Asking it questions like give me comparative power over other humans like make a trillion dollar company. Hmm. Bit If everybody has it then it won't work because it's comparative value and Technologies democratized. So then the real game we have to play the infinite game where everybody has enough or be one person has the only access to Super Ai and then they ask and make a trillion-dollar thing and they're a eyes more powerful. So, it works. Yeah, so it's one of those two scenarios and I think that It's a really beautiful thing where we are right now we and I hope it plays out that the philosophy.

30:19 That's underpinning. A lot of these companies is it we want to democratize this company, we don't want a few Super Players to do it. And so that was the idea that opened my eyes. I was founded with now, is it still living up to its ideals? You know, I think there's a debate to be had about that, but principles are still very much in the culture of the open source market of Allah. People work at open. A I, you know, we're all just human beings at the end of the day. And so I think, I think it's really important, we keep that Vision and that inspiration and that ideal at the Forefront of this movement. So what that guy said?

31:01 Yes and I can see got it. Like you are on the positive side of the thinking, right? So you're optimistic Yeah, I would, I would say that generally speaking. The, the realization that that's allowed me to be optimistic, is to look at the writings of earlier centuries and to see what blind spots that people had when they were. So, so confident, and to see how much better we are off right now, and to see how many of those pessimists were wrong has So as I think, luckily happen enough times that there's a pattern, we can recognize and that pattern can be extrapolated without just Faith. Like like, you know, I think that there's questions we need to have about gray goo and what, if it isn't super, replicator Ai, and, and Cetera.

31:54 Anything that there's there's a rational logical reason to be optimistic even about these more, esoteric problems. But fundamentally I think there's a lot to be optimistic about on many dimensions. I'm on your side and just, you know, too. I would say complement, what you mentioned garak. I believe I will because you, you touch on that and you said, maybe either we all be in an abundance. And I believe this is this is the case. I believe all of us will be in abundance, life will not be the same. We are living it now, of course it will take maybe hundreds of years, I'm not sure I'm hoping that it's less whereas because the reason I'm saying this because if anyone You know, and we saw life examples, maybe it was like a joke.

32:42 We saw this guy who went viral on Twitter and he did something he called it, the hustle GPT. Right. But guess what? I have people that I know that we follow each other's. Even we were in cohort together. They did it. I mean okay it was a small scale but they did something and get them 500 bucks and you know they said okay let's try to get 5,000 bucks. Of course if now you go and type okay I need this site has some that Make me 1 million dollar, it will not work. And actually chat GPT will answer you. I think they tweaked something and to tell if there you are, man, I'm just, I language model and I cannot give you any Financial advice. But I mean, this showed us again, I call it proof of concept, right?

33:26 So there's something called proof of concept that. Yes, a, I can generate ideas. Let me give you one more example. And actually, I was planning to do even a Hey, they're more like maybe walk through about it two days back and you know like you and I we share this geeky thing. So I looked at you know by office here, I have a Raspberry Pi and they have a Arduino as a city sitting here and you know, all of a sudden I said, hey, let me go to chat GPT and ask, you know, give me some cool idea that no one may be have done it. Of course, up to the knowledge that they had up to 20 21, and guess what? It gave me Really cool ideas like maybe no one will be interested in, but when I start to read, oh look at this like, you know, it's suggesting me, you know, the ideas are just me, the components that I need to have giving me step by step, you know, guide and actually it's funny enough. It's told me I'd and hey, like this is just, you know, a fun project, do it yourself.

34:30 Kind of thing, even subject, you know, I said, okay, what can I do? He said, yeah, like, you can put it into a website. Fight for hobbyists who would like so you can put this idea that I'm giving you. Okay. So this is the augmentation guys that, you know, I'm enjoying now. So if you are always thinking of this kind of take my job, you know, like this will end the world. No nothing like this. Nothing like this will happen, actually, you just need to have fun and actually I was not very fan of you know, like reels and Tick-Tock and all these things. But now when you go that side you see like there are a bunch of people young people who are Sugar are getting educated on AI and they are using it somehow either to create content or to do something, which is very, very positive.

35:14 Now, we wanted to say something and we kept it aside for the use cases Garrick like the most, I mean close ones that we can see in the near future like where we going to see the changes fast. And when I say fast we're not talking about yes. Maybe in the coming six to 12 months. And then down the road how you can break down that to us. Yeah. So one thing that's going to be coming out very soon is I think we're going to have a lot of companies that are going to be producing better interfaces with with where this technology. So if you start if you start using chatter BT not just for fun but actually for your work, you'll often find that the You know, the hard part was figuring out how to describe the problem, who would have thought at the questions was harder than the answer.

36:15 I asked you to make a big report, but telling it what to make a report on, who's it for, what's the relevant, you know, factors? That's the hard part. So I think we're going to be creating better and better interfaces to reduce the effort of those parts. So there's big communities right now that are trading and selling. Prompts the, they call themselves prompts engineers and prompt marketplaces. I think that that is just the very, very early days. And probably will look nothing like a Marketplace in the coming years, but I think a lot of people going to make a pretty penny, creating better interfaces.

36:55 I'm thinking of like, Jasper, Ai and writer Ai, and those tools, it's really just as using chat, CBT on the back end, but there they created an interface. Makes it easier to extract. Good copy, that's number one. Number two, is that the the a eyes that are being produced by like barred with Google and tragic Beauty with opening a, I are very general in their knowledge and a go to Tech. Go to Startup is going to be to create a specialized AI that performs very well in a subfield. So think of illegal AI right now that the Is AI hallucinates and you can't have that in the legal profession or a medical profession.

37:41 So by creating finely-tuned a is that make less mistakes and are less generic? I think a lot of value is going to be added to the market and a lot of very successful startups going to be born. I'll tell you one pet project, I'm excited for it, I'm excited for AI. That scrapes, every piece of text from a variety of let's call them gurus like mentors like Steve Jobs, Etc. And then Crates on AI Persona so that you can have a conversation of asking Steve Jobs. Hey, what would you do? And it takes every transcript, every conversation from that, perfect guy, every interview and you can have Steve Jobs or anyone as your business Mentor. That's an example of an AI that becomes fine-tuned and narrow down with with the right data inputted. So those are our two, there's a few others but those are two fundamental lenses.

38:32 I like to put on and advise people to put on when thinking Of new companies and new Innovations, still to be created. But guys are we going to see? Like only like, for example, historically, it seems like we are tied, 2-2, big names all the time. So it was Apple Microsoft. I don't know. Like we have all these two big names. So here, now we have now Microsoft in an open-air. I backed by yeah verses verse. This is Google, are we going to see like this consolidation? And the reason I'm asking you because nowadays, like even you startups are coming up and there is the famous meme on the internet.

39:19 When they you know, when they remove the the cover it's like running on top of open a iapi. Yeah. Or maybe Google still didn't release an API for board but I'm sure they're gonna do that very soon. Yeah. And the reason why I'm asking you like don't you think That actually these two companies will have huge impact and control over any startup that will come in the future like and how we can tackle that actually that's a great point. And you know, I think that there is a tendency for a lot of companies become a lot of Industries become duopolies, you know, PlayStation vs Xbox.

40:03 So on, so forth and It's the AI Wars are really shaping up for Google and Microsoft to be re going head-to-head. Apple is because they like to be very insular are not using the word AI in any of their press releases. They're using words, machine learning Transformers Etc. But they're avoiding that word because they don't want to Hitch their brand to to that. So, I think there's gonna be a lot of players that are pretty big that are maybe not the center of attention in this, but More importantly, I want to go back to what we're seeing in the middle of today's conversation, which is the open source, Community with the Innovation, from llama, being being leaked, and the Innovations with chinchilla and low, low input training, which is meaning that this technology becoming a lot more democratized through through developers, building building custom apps that are not tied at all to either of these big players.

41:06 And so I think that it's very natural for a lot of Industries have duopolies and I think that we will have something like that, but enough options that we won't find ourselves in some sort of nightmare scenario where a shadow Monopoly. Actually dominates too much. I would have said that even three months ago until until the realizations from that white paper. Once again recommend you know we have no mote from from Google. We're definitely recommend checking that out for more. Evidence for that. Yeah. And you're right. Like things are changing so fast because what maybe I've said, even one month bag is it doesn't apply anymore.

41:50 You mentioned something about coding and I'm interested like when we can expect a proper full fledge code generated by because now you know the feedback so far that it's Not the perfect code, it still need, you know, people to go and check it back. Of course, always, this is why we need, for example, quality engineers. And we need like Q&A and all the stuff. But like, really like, when you think day, I will be able to generate proper code. Is it any? Any, any near future? Yeah, that's, that's a fantastic question.

42:32 It's not going to be a single moment, you know. I'm not sure if, if people think of it this way. But I remember when Siri first came out and everybody said was terrible, and they weren't interests. And then five years later. Siri, just works. And then people are like, well, of course it works. It's been around for five years and there was never a moment of like and you it's quote, unquote arrived. And right now, the copilot tool that Microsoft has for its a GitHub product already there. Noted that I think more than 50% of all code committed in this year has been written with copilot, which does not mean 50% of developers are using.

43:14 It means a 10%. Developers are paying $8 a month are writing five times more code than anyone else. Um, so so really what we're looking at is still in the beginning stages of like we said before, people being Amplified, not not a, I taking your job but people being Amplified I'd becoming more productive. And I think that that's kind of what I think. That's really what's going to look like for for a decent amount of time. And we're probably pretty far away from the idea of you just typing in. Hey, making a million-dollar, you know startup it's just not going to happen that way at least while we're constrained by physics. So, I think it's a long, a long journey with a lot of little micro milestones and probably was not gonna As an AI futurist.

44:02 I'm not hitching. My my planning around that anytime soon. Okay, just to put context and maybe if there's anyone who doesn't understand like how coding works and to your point Garrick, so when when programming languages were created, so they were just, you know, putting on top of some Language just for the sake of a programmer to write a piece of software in a human understandable format. And then you know, with time the code, you know grew. I mean it's start to have more libraries that people can reuse. Why?

44:46 I'm mentioning this, the reason I'm mentioning this because this is remind us of how large language models. If you are today you are not a geeky guy, you are not a coder. So the way large language models works is If they have a huge amount of data sitting somewhere, and then when you go and put the question, like, even if you are just saying, hey, give me a. I don't know. Write me an essay about Construction business. So it goes first, you know, what information, same as you used to do in the past, I used to go to books and then go to articles and then try to rename rewrite it in your own way.

45:29 So that's what AI does now, connecting back to to code. So, actually, we have a library of code that a, I can understand what it does. So if I go and today say, to chat GPT, I want a function that written in Python. I need to specify, of course the language, I Python code that goes. And for example, check a website if it's up or down, check how much is the page load speed and put that into a specific format output to a text file? For example, just because we are thinking non-ui now. So, what? What, what, what opening our HR G PT will do to go and find, okay, which Functions within the programming language, with python can do that and then to write it for you. Maybe it's not the ultimate Way what Garrick mentioned is that it requires that because they need to train, the more data you have and the more you use it and the more feedback you gave it becomes better because it's what they call a supervised model. So basically you need to give to keep giving feedback to it. So it can you know, interact better with you and just for the sake of the audience here because I've been using Using chat, you Petey's maybe one week after it went out. The thing is that because they can collect data and we know that they can collect data about all our interactions. And I can tell you even with the same problem that I used to have three months back or four months

46:59 back, if I use the same ones, now I got much better answer because they know better about me and what are my needs. So, yeah, like it's amazing. I want to share one. I want to share one one thought, just on that. Point of it learning. We're going to people talk about prompt engineering, becoming a new job description. I think we're going to have a new job description that future as well, which is, which is it? Large language model optimization, which is the same as SEO and ASO with Jess is going to be. What if you you ask, hey, what's a pizza place word? Give me a list of ten pizza places. And you prompt it so often to include Mom and Pop shop in Atlanta that all of a sudden Mom and Pop Shop, Atlanta shows up in more people's queries already. I have some friends who are working on this to influence, other people's answers to questions, like how they train, they are training the model.

48:05 Now of course that's going to come an arms race. You know, SEO never SEO is still a work in progress, It's a constant dynamic system, it doesn't, it doesn't really hit equilibrium. So Beak. So like it's just gonna be another, another Vector of of games to play, but that is going to become another. I believe industry on top of SEO and on top of prompt engineering and and all the future holds, yeah, on the prompt engine before I forget because you talked about, you know, people doing prompts and trading prompts. So actually this space guys, if you thinking to keep this business model, I don't advise you at all. Because and I mentioned this in a previous episode couple of days back. This is being sorted out by itself. So you mentioned Garrick Jasper.

48:54 I and just very I, they have a function. So if you missed that episode, I'm repeating it now and I'm not affiliated with them. I just use them as well. So, just where they get a really cool function where when you type The Prompt when you type, what you want to do is it like a blog post or whatever. Have a button now enhance it enhance. So when you click on enhance I'll give you a very simple example. If I say, I need the truth, let's make it a bit more. I need, they link it in post about sharing a new product with the market. And when I write this phrase and I hit the enhance button, what Jasper I would do, it will go change my prompt and to write something like this.

49:43 You are a marketing specialist in social media and I want you to write, you know, like it make it much longer the way that you should interact actually with large language models. And when I saw this what? Like, yeah, we don't think we know we don't need to enhance the prompts anymore. I mean, there is a an AI that can do this. So as we are coming with me, yeah, I think, I think there's a lot of, a lot to be said there, I know we're wrapping up but I think there's a prompt. Engineering is definitely a lot of Transformations. Fundamentally we're still constrained by physics and then different skill sets. Put some people in better situations and others so I think that I don't know if I would say it's wasted effort but certainly is not going to be anything like what it is right now.

50:30 I completely agree with that. 100% one thing I want to ask you guys, let me maybe people now who are listening to us. They're saying, hey, these two guys. What are they? What they are talking about that. We are interested in the eye. We want to get involved this in this day. I think what a second from a career perspective. Maybe they want to start a company or maybe it's just a hobby. So, what advice or resources? Would you recommend for these individuals? Yeah, so the thing that I tell everybody who is interested going to AI is to check out playground dot, open a i.com, which is where they show off the the Hidden variables about how their large language model actually accepts inputs.

51:19 So if you're just doing chat gbt you're missing, big chunks of story. So I would say it's just most people who want to get, maybe not fully technical, maybe they don't want to code, but they do want to understand the AI on a, on a more intimate level than they would just by playing around with it to check out playground dot, open Ai. And if you're interested in starting a company, my company of a hollowed-out team is right now, looking for partners. To start new Ventures. Our mission is to be creating a, i based SAS companies to build up and then exit from in two to three years. So we're looking for partners and businesses, so anyone who wants to can reach out to us, but if you're just starting as a as a hobbyist, I definitely recommend reading more of the open, a.i. documentation playing their playground great. It's good.

52:07 That you mentioned this because this something I asked, so I gotta put, you know, the website. Of the company in the episode description and of course I will put also the profile access to Garrick so you can get in touch with him Garrick. Like this is my famous last question nowadays. What is the question that you wished me to ask you and how you would answer it? Oh that's a fantastic question minute. I felt like you were a fantastic interviewer. Let me just let me just go back to you and Cam went this was a lot of fun and I I decided to go a lot more technical than I normally do with the CTO show and you were fantastic to talk with an exchange with so great. Great job.

52:54 Thank you, my pleasure. The question of what I wish you had asked I think we covered most every I don't know where the question would have been. I love talking with a future of a, I think there's a lot to be learned from how it's from looking at darwinian Evolution from looking at how biology has has been solving similar problems over billions of years and how we're now solving similar problems with AI with similar results. So maybe, maybe if we do a follow-up episode, we'll talk. More about that.

53:35 But, no, this, this has been fantastic. I met, and I think we pretty much covered everything we wanted to okay, great. That I'm really happy Garrick. Like, I'm really thrilled to have you today on the show because I think you shed light on a lot of aspects that people have thought, okay how the future looks like and from healing from someone like you, it would be really enlightening for the audience. Thank you very much for being on the show today. Day. And guys, like, as usual as I and every episode, like, if you have any questions to Garrick, or to me, you can reach out to us, for me. You can reach out to me by email, LinkedIn, or Twitter, where I'm the most active.

54:17 If it's the first time you are watching this on YouTube, you can subscribe to the channel. We are always having, you know, Superior content, like, great interviews with CEOs, subject matter experts and you are hearing about all the new technology from AI. To emerging Technologies and all the rest. And if you are listening on your favorite protesting platform, also don't forget to leave us a review and also subscribe to the podcast. And if you are interested to be like Garrick today to be a guest, I would love to discuss this with you. Like this is how actually we are growing the show more and more and I will be discussing that with you, one to one. Of course we can choose the topic the form at the time.

54:58 The date can arrange for all of that. And don't forget also to Leave us a review with appreciate this and until we meet in next episode. Thank you very much, take care. Bye bye. Fantastic. Thank you. Meant that subscribe button. Share the show with your tech-savvy friends and fellow entrepreneurs and leave us a review on your favorite podcast app. Your support means the world to us.