Orchestrating the Web: The Untapped Potential of AI

34:32There's a future for them, but they'll have to specialize away from what they used to do and why they got the big valuation in the first place because the big models are just on the march, right? So if you're a SaaS company and then I'd be a VC and I have to look at you and I have to just think like, okay, in this orchestrated web of collaborative intelligence, collaborative consciousness, where do I still see the need for you? Because I can cobble together these three little pieces and put it together with an orchestration and I do the same thing. I don't have to pay for your overpriced service.

35:09That's not a nice position to be in. So VCs are very jittery in the market because of that. Now, on the other hand, traditionally, I would have shied away from service businesses. And why? Well, people get sick, turn over employees, leave, and it's hard to scale people up. However, AI agents solve all that problem because they create repeatable, scalable, predictable revenue at scale. So some of the best VC investments will come from taking like a traditional service business, and that's maybe not so efficient, and then applying agents to it, and scaling the heck out of it, crushing their competition, and realizing the same kind returns that traditionally went to software.

36:05So I'm very bullish on service as business in like the next five years. Now, after that, right, after that stage, competition will eventually have caught up and then all your competitors will also be more efficient with AI agents. What will happen then? Well, then prices will start to come down. But this is something we see with AI now too. OpenAI has decreased their prices for their models three times in 12 months. The last price cut was just a few weeks ago, 50 % cheaper on the input tokens. And so what does that mean? It means that the cost of intelligence and the value of intelligence is becoming eroded, which means ultimately that will translate into what is the cost of operating an agent.

36:55It will decrease and so forth. So ultimately what we have is we have a lot of goods and services becoming so cheap that there's basically no reason why we couldn't offer it to everyone in our country. So for example, we can afford to put roads everywhere.

37:23we can afford to provide infrastructure for electricity and internet everywhere right so if the cost of health care became so affordable because every citizen could have an ai doctor helping them and it's just an ai model why wouldn't you give health care to all your citizens, right? What if you could do the same thing for education? What if you could do the same thing for entertainment and have AI just create endless entertainment? You can give your civilization like Panamets, Sikensis, Bretton Games, and people would have all of their basic essentials covered which basically leads to a society where um where where we could potentially eradicate poverty if we have the desire for it right and so this kind of where this all ends is we're we're in this transition period i sometimes call this like the last hurrah of capitalism right before deflationary pressures like bring the cost of everything so low that it's almost impossible for you to get ahead of other people in terms of, I own so much more than you, because everything just becomes so much more affordable.

38:43If I can spin up a few agents and with two friends or so, spin up an entire company that has the equivalent of 1 ,000 workers today in a large company, imagine what I can do. It's just eroding the differences of opportunity between us. the entire economy changes in a big way. And just to back up one of the points you just mentioned, I was watching a talk by Andres Carpathi, one of the founders of OpenAI, he recently left. But what was so interesting to me is he was talking about these scaling laws that are starting to manifest. And I would almost equate it to Moore's Law back in the day with chips, but he talked about there's two parameters, there's two things when you think about it.

39:27I believe it was the number of parameters and the size of the amount of data being trained on. And they're starting to see this pattern that as you increase those, there's no fall off in terms of improvement in the models. And it gets to the point where intelligence just becomes extremely cheap in a sense, right? Because you can keep scaling. That's why you got Sam Altman trying to raise$7 trillion. Excuse me, one of the reasons. but I thought that was really interesting in terms of how... There are diminishing returns when you scale. And what's so interesting is like studies have shown that at a certain scale, it doesn't matter whether you have transformer architecture or Mamba architecture or whatever other neural network, given enough data, the neural network will self-optimize against the data and you can almost not say which architecture is better.

40:22It's almost like it doesn't matter as long as it's able to train on enough data and encode all that sufficiently. You can get to that. But there is some diminishing return on that. Now, the interesting thing is I think it's underappreciated, but it's definitely 1 ,000 % something I'm convinced about. So in order to unlock higher levels of AI performance and intelligence, you have to increase the multimodality of the model training data. You have to. There's no other way. For example, you could have a really good impression. I'm giving a real life example. I grew up in Belgium. I watched a lot of movies from the United States.

41:14I must have seen movies with New York in it and the Empire State Building, everything a million times. And then I moved to the United States and it still took me a few years before I made it to New York for the first time. Grown, all grown man in my 30s, first time in New York City. When you go there, it feels surreal because you feel like you know that place. But it cannot replace the actual being there. there's parameters missing. You understand what I'm saying? This is what multimodality does for AI models. So you could train it on text and based on text descriptions, it can create a picture of the world that approximates the visual understanding of it, but it doesn't fully appreciate visual space.

42:05Training it on images, you can have something like Dali, but in order to create something like Sora, you need to train it on video. And to approximate the human experience, you need much more. You need actual robots embodying the physical space and navigating the physical space as humans do to have a real experience and real data that it can learn from. And not only that, humans are way more complex than that. We smell. We smell the beautiful like spring air and it impacts or impression if you're just a camera and a speaker and whatnot you can never appreciate the same impression a human would have so you need sensors you need uh you know you need to simulate the emotions and the hormones and everything to approximate the fullness of human human intellect and experience and so unless we have that all all you are is like you and I meet online, but we've never met in person.

43:06Right. And there's this Netflix show currently ranked number one. Love is blind. And you see it all the time. Like people are talking to each other, like dating without seeing each other. And they think they fall in love. And then they meet in person. And then there's like a physical response that they couldn't predict because they weren't never met in person. And for some people, it works out. And for other people it doesn't work out and it's this is the same thing for ai because you can you can put an ai behind that wall and that show and the other person might not know that it's not a human being yeah until they have to come face to face with it and then not have any physical reaction at all because it's a piece of metal so a couple things here hey that's probably the best metaphor you win for season two so far, the love is blind metaphor.

44:00But you just kind of expanded my mind a bit too, because when I always thought about modalities, I always went to text, video, audio, just kind of the basic ones. I wasn't even like thinking about the ones around the senses and all of that. And I think that just shows how far we have to go and how early we are in this journey uh in terms of the potential but you look just even going from text to video like look at what google has in terms of youtube in terms of all the video there it's insane what they're sitting youtube university anything like my kid the other day my kid brought like hey um hey dad i made your breakfast i'm like oh that's so sweet it's like nine years old super sweet kid and he's like uh yeah i made you egg souffle i'm like i don't even know you know that word like wow he He learned it on YouTube.

44:51He found it on YouTube. He's just like YouTube University. So you now have AI that can watch entire videos. It is a learn from you. Imagine an AI trained on the full knowledge, not just on the transcripts of it, but all the visuals and the text and just watch all the news in real time as it gets posted about every topic in real time, encoder, decoder, architecture in real time. That sounds superhuman, but it still has not had one time feeling its heartbeat in its chest. It's still lacking modalities, right? So we've got ways to go. And so I believe that in order to unlock those kinds of AIs in the future, we are going to have, calculators are super, super human.

45:39They can calculate way faster than me, way more accurate than me. Calculators and computers are already super human, But they're not multimodal. They're not the fullness of human. And then there's the big question at the end of the rainbow, right? What about the human soul? After you've done everything else, is there still something left that is not materialistic, that is ethereal, that we can't quite put our finger on, that gives us a chance to beat the Terminator in the movie because the human soul is something that it didn't account for? right is there still something missing that is possible too we we have to go to the end of the rainbow and just see how far we can go we can emulate humans just like you can emulate and describe a visual scene with just text but it's not enough it's not the same thing yeah and it takes back from terminator uh also la vista baby there just one last thing i want to hit on because you hit on it earlier is there going to be a need or a skill set of like coding software development or does that go away in a sense and there's a different way we interact with and build technology but where do you see that going and it may not be like the immediate future but where do you see that going in a sense so you can be a generalist or a specialist if you're going to be either of the two and you want to make a difference you better be darn good specialist or generalist for a specialist to be darn good you'll have to learn how to work with technology rather than against it that means that you know how to ask better questions than any other human being on the planet about your subject matter of expertise that you know how to challenge the AI and push it deeper and harder, just like you don't accept your first prompt back from GPT and you know how to ask the right questions because you know what a good output looks like, you'll need that as a specialist.

47:54And there's going to become a level where the AI gets so, so, so smart that only very few human beings know how to challenge it anymore. and so this is where specialists have a future in development right the next thing is and yes that's the one percent of the one percent yeah and then you have the generalists and i see this so much so many times in business i have the smartest i have a room full of phds working with me right there's smartest people in our field but ask them like to tie a shoelace and they don't know where to start or they overanalyze it. The generalist is like the MacGyver, right?

48:34The person who's so creative and pulling out of the infinite fabric of the multiverse, they pull all the strings together and imagine what a solution would look like and know how to describe it and use AI to help them create their vision into reality. How many people can't start a startup today but have great ideas because they don't have funds or time or economic opportunity, whatever. And AI is about to change all that. It's the great equalizer. I think generalist developers are really important because you represent the human race and you represent a vision and an idea. And AI is the master craftsman at your disposal in multiple fields where you can command.

49:21You know, I'll go back to army. You can command an army of skilled craftspeople to create and realize your vision. You want a new house? You know, 100 drones are ready to just stack all the bricks and do everything. Additive manufacturing just before your eyes. You envision it, and it's just constructed in front of you. Nobody's homeless. Nobody's hungry. Nobody's lacking basic resources. And the generalists and specialists are going to have a role in that. And then everybody else, it's like League of Legends, right? It's fun until you play ranked. And then you get stomped by Koreans who run around circles around you and you get really frustrated with your team.

50:07If you're just playing League of Legends casually, it's fun. You don't care about being the best or winning. You just are there to have a good time. The vast majority of developers will be able to just have a good time. But if you still want to make a difference in this AI-enabled future, you better be 1 % generalist or 1 % specialist to make a difference. Yeah, that's an interesting way of thinking about it. And your Army reference just took my mind to a weird place. And this is a weird spot to end on. But my wife and I, we watch a bunch of these weird cult documentaries on Netflix. Is there a cult of the future where a leader is, you know, employing AI agents in a sense versus humans?

50:56Like, is that like that's kind of a weird it just popped in my head. Is that something that one of those negative kind of effects and scenarios we we deal with in the future? Let's keep it positive. Right. Because it's the end of the podcast. I'm going to I don't know the name of the study and I can't name the people involved in the study. so I can't credit them. I'm sorry. I just, I'm not a vector database with perfect retrieval. I'm a human being, but I will tell you the concept of it. This was a wonderful study. It was actually done a while ago. And so, you know, at Thanksgiving dinner, the dreaded politics topic, right?

51:36Everybody has an opinion. Now I come from Belgium, tiny country, huge countries around it. We're basically the punching bag of Europe, right? So the only thing you can do is just get along with your neighbors and try for them not to invade you every two seconds they can, right? That's like the history of Belgium. So you need to learn how to get along with people. And so let's say you're in Thanksgiving dinner and people have like different political views and stuff. So now, you know, I've been in America for a while to know that can get pretty intense. But when you peel behind, when you peel a little bit behind, just like the orange and you peel it slightly behind, what do you discover underneath?

52:22Everybody wants a more just society. Everybody wants less poverty. Everybody wants more prosperity, more opportunity, more happiness. Everybody wants the same darn things. The political debate is not about what we want. it's about how to best get there see where i'm going with it yeah yeah so some really smart people did a study when it comes to like one of our top political debate topics taxation what is the ideal taxation that maximizes individual liberal liberty prosperity social mobility but also social justice and welfare for the general public different political parties are going to put out hypotheses they're going to then be voted they then have a chance to implement or complain about their hypothesis test them out in the market and then eventually the population can then vote again and then see and reward the good ideas and punish the bad ideas and so democracy moves forward right but that can take 50 years to try everybody's proposal.

53:32Okay, some really smart people figured out what if we can make a simulation of the US economy with all these parameters, and we can just ask the AI the question, find the ideal thing that maximizes everything. Now, the study came back at something astonishing. They found a way to lower the taxes and improve welfare and improve all other categories. So if we had access to such technology in our decision making and policy making, we could, in essence, it's like testing medication on a simulation of your body before you have the side effects after taking it. Why should you be the guinea pig? Run a simulation of the economy, run a simulation that includes things like diverse people groups, interests, various backgrounds, economics, social circles and everything.

54:28Take everything, equity, take everything into consideration in these models you can. And then just let the AI run the numbers. And since we all want the best thing for this country, we'll all be in agreement. And the concept of running simulation is not anything new, but it's the ability to that last point that you can actually realistically feed in all these parameters and actually have it work and spit out something on the back end is the new on log here. So, yeah, could you use that for totalitarian government purposes? Yes, you could. I mean, you know, I don't like this and I've already made too many political references.

55:11But to me, like the discussion around AI and access and everything is actually no different than the debate around the Second Amendment. The people that say, well, everybody should have access to AI and powerful AI and so forth are using the same arguments that people are saying everybody should have access to weapons. I mean, we may not like it, but it's the same darn debate, same arguments. Is it better for everyone to have access because more good people have access to it, meaning so-and-so? Or do you give access to the corporations with all the economic power in a country and then you get to corporatism?

55:49It's the same type of debate all over again, but with a new toy. And of course, this toy is way more powerful. I think AI is more dangerous than a gun. I'm sorry. right yeah so it's not ai use is not enshrined in the united states constitution but it is more dangerous than it and it's less regulated than than than even guns in this country despite the constitution so you know do you you have a handful of companies who build powerful models who are acting responsibly but then you have open source models who can do whatever they want and there's no regulation. There's no framework around it. I do think a society has a duty to its citizens, and we have a duty to each other, to have a responsible dialogue.

56:42I think that's starting to happen around, you know, what do we want for our future out of this, you know? And then the last thought is just then people like ask this often, how are you going to pay for universal, like generous income instead of basic income. How do you pay for all that? Well, currently a lot of transactions like human to human or human to agent. And with agent, I'm going to include every SaaS platform Uber Eats or every platform you're as a human pay money to a piece of software. Okay. But the real economy of the future is going to be, yeah, human to human still going to happen, just like we still barter and not always use legal tender.

57:26We still barter a lot less, but still it happens. So human to human interaction and transactions happen still. But the biggest economy of the future and the growth driver of the future GDP is going to be agent to agent interactions. And so this is when we talk about the orchestration, nobody wants to just like, everybody wants to be paid for their servers and their work, right? So the way that's going to happen is that when these AI agents are interacting, they're going to let each other know what their cost is. And there's an entire economy that's going to happen. But because these agents and AIs work so much faster than humans, there's going to be trillions of transactions for every human transaction.

58:11And so this layer of the economy, agent to agent, that's ultimately going to be the majority of the GDP of any country. And that's the layer where you do not need to tax humans anymore. So I see a future where humans are not taxed on their labor and keep all the fruits of their labor 100 percent. And where the agent to agent transactions are facilitating so much economic growth and wealth that they can also provide a universal generous income for the citizens of their sovereign nation. So this is also, but this is a little bit longer term out, but all the pieces are already in place for this. This is all converging already.

58:57It just takes a little bit of time. That's like a whole nother episode right there. The agent to agent is that, you said there's a transaction there, just philosophically, theoretically thinking through this. Is that happening on behalf of a human or how is that? does that make sense in terms of what i'm asking because you you're you got my wheels turning a bit in terms of how this evolves not necessarily like it can be but it doesn't have to be let's say that let's say let's say you play pal world and you're assigning a pal to water the water it has a watering skill and it needs to grow like berries or something right Right.

59:44So just assigning the pal to your base and say, go and make berries for me. You already know the economy of that and you know what the output is. It's berries that then your pals can eat. So just assigning the agent on your behalf to go be productive in some way, you can say start and stop or reassign. So once you assign it, you give it agency to go do its job. And just like you think human, like it happens all the time. Like we want to sign up for some service. So CFO, hey, issue me a new virtual credit card. Okay, what budget do you need? What is it for? Okay, I approve it. So you can be as a human, you can be the CFO that controls the budget for your agents.

1:00:32Or you can have an agent CFO and you give a high level budget and say, make sure the budget is allocated fairly between all the other agents and doesn't spend above that. And so agents can either self-regulate their spend or a human can act as the ultimate gatekeeper for who is allowed to spend what. But the point is, if you deploy an agent to go do economic productive activity for you, you need to enable it to go and buy, I don't know, a Canva subscription or something, right? And do what it needs to do. And so, and these are going to be very different than the current economy. The current economy is about make like a commitment of$200 a month to sign up for so and so.

1:01:14No, no, no. That's not the future economy. The future economy of agents to agents is microtransactions. Like$0.0001 for one particular impactful action that I needed to hire you for. All of that is going to happen over the blockchain. probably going to get the Federal Reserve into that because they want to control the money supply and everything. So it will be likely sovereign digital currencies that will be the official legal tender for country interoperability. But there's going to be other forms of transactions. And then just like you go to a website now and I accept Visa and MasterCard, they will be able to announce as well to other agents, I can accept these forms of payments.

1:02:00and then agents have permission to pay in these forms of payments. And then it will just be a blockchain transaction between the two agents of a minuscule amount, like tiny transactional and you're paid instantly. Like it's not like you even have to have like net 30 terms. You want to get something done. It happens like instantly and you pay like a tiny amount of money just for that action. So it's all like in supply chain. It's like just in time, GIT, like just in time manufacturing. No huge warehouses, no huge backlog of inventory. No, something comes in and it happens in that moment. You only pay for the single action that you needed help with.

1:02:40And all those transactions are going to amount to trillions and trillions and trillions of transactions all happening over the blockchain. And it will create so much wealth and prosperity. We can just tax that, tax those transactions. but like a tiny VAT, like value added tax. And this is why it would be a good idea to have like a legal tender sort of thing for this, like a digital dollar. Not as a forced way, but as a practical way, you need a blockchain type of transaction that can have value added tax attached to it. So you do not have large corporations doing tax evasion, making billions of dollars while the rest of us starve.

1:03:30And so this is a societal contract you make where you say, we want to embrace AI and we want our society to have the benefits of that wealth creation and take care of our population who is now displaced otherwise, so that humans then have time to go and live their life to the fullest. We're not made instead of cubicle nine to five. We're not made for that. Like, let us play Lego with our children. I'd much rather do that. Honestly, I'll like turn my computer off and spend time and play with my kids right now. Like, why are we doing this? And that's what I was about to ask you, well, how do humans make money in this new future where you have AI agents doing everything?

1:04:14But maybe that's the point around the wealth creation. It's just a different way of thinking about it where you don't necessarily need to be making. We solved it a long time ago. It's called parenting and children. And when you're a kid, your parents provide food, shelter, everything you need, and you're happy and you're happy. And if you want a little bit of extra money, you go babysit or do something. So you want to go on vacation, you go and do some extra economic activity, you go volunteer, go help, you get some extra money and you spend it on stuff that otherwise you can't do. But you can get a job too as a young person.

1:04:52You can go work in McDonald's when you're like 14 years old. You can go work part time. Like you can, but if you have a great life at home and you can be with your friends and hang out and socialize and play sports, would you still want to go do that? And so what we need to create is a society where you have the freedom to do what you want to do. You want to do more work and get financially ahead in the difference. Go ahead and do it. You want to instead say I have enough. Like, honestly, like Europe is a lot like that. I mean, you've got you've got people that are like. Single income households and they go to Spain and Ryanair on vacation three times a year.

1:05:34You know, a cleaning lady in the United States where her husband's not working would never be able to do that. And there in Europe, they think it's normal. Oh, no, no, no. I'm clocking out. It's five o 'clock. What do you mean? I'm not here. This is not my life. This is my work. It's five o 'clock. I'm going home. Friday in the United States would never work. You'd be like, I'm not saying fired, but performance evaluation against your colleagues would be pretty tough. We're a society obsessed with work and obsessed with keeping up with the Joneses, but different lives is possible too. And I'm in favor of capitalism and freedom and free market.

1:06:13I think people just should have the freedom and option to live a more fulfilled life. And like the last example I'll give you is that capitalism is not always rewarding what is most valuable. For example, motherhood. A receptionist makes more money than a mother. I think that's wrong. Economic value is it should be taken into effect long term value, not just short term value. A short term value of a receptionist is like I just answered a call, I make money. Long term value of a mother spending time with her child is a better society, better race, children, better society, better fabric, better happiness, more safety for everyone, more prosperity for everyone.

1:06:57And mothers, like that should be like a paid job to be a mother or it can be a man, too, who's a primary caretaker. Forget about the whole gender thing. Just think about an economy that we can create together where. where we create freedom and opportunity and equality for people without forcing them into anything that's the big flaw in like in communism you have to go punish the hard workers take from them to give to the other people and in capitalism the problem is you're taking from the productive of you're not taking care enough of the weak people in society, right? So there's the two extremes.

1:07:43But we now are having this emerging third option, which was never possible before, where the majority of our GDP can be derived from automation. And we are no longer taxing the labor of people. You want to be a great podcast host, it's very popular, you want to use everything. You're a human being, you earned it, you deserve it. Go and keep everything. But at the same time, go and take care of every mother and reward them for being a mother. And it's a job. It deserves to be rewarded in society to take care of it. It's just things that make humans truly unique is what's going to stand out and will be valued in that way.

1:08:27But this has been amazing. One of the best conversations I've had to date. So many different angles we went into. but if you want to get ahead of this, like as a business, this whole concept of AI agents, AI orchestration, tell the audience a little bit about SmithOS. I know you're still in kind of an early, you know, early access mode, but let folks know kind of where you're at and where it's going in the near future. Right. So today, if you want to build AI agents, they're either very basic or they require you to have like, a hosting environment and figure out APIs and security and encryption and code and all that.

1:09:09We made a platform that makes it very easy for everyone to build agents that are doing this thing with workflows, data, tools, and intelligence. Drag and drop, visual, no code required, although you can add it if you want to. But it's the first platform that from A to Z makes it easy for you to automate things in your business. And you can deploy them right alongside your colleagues in Microsoft Copilot, OpenAI Teams. You can deploy it to your website as a chat if you want to. But calling this a chat is an insult to its potential, but it can do it. You can deploy it as an API to power your no-code front-end mobile app or whatever.

1:09:52You can run in the background and look at new tickets coming in and automating several actions. it's um it's not yet obviously like any ai it's not yet where i think these autonomous agents will end up being they get more powerful with this web 3 the stronger the internet becomes the more powerful the ai models become the more data you have access to the stronger your agents become what we try to do is make something that is that takes the agent building itself like almost like the easiest part of the whole equation. Figuring out how you make your pancakes, where you source your ingredients, your data, and what pancake or waffle maker you're using as a tool, that's where you spend most of your time.

1:10:42Once you know how to use Smith, the rest is just drag and drop and choosing what you want to do and then pushing a button and you're deployed. We're already deploying this today in early access to 100 early adopters. And they have mission statements like, by the end of this year, we will have transformed our entire agency. Every department, every team member is going to have AI co-workers. And they have already deployed, in some cases, like 12, 13 agents already in the first two months of this year. And they're on a journey of AI transformation. I invite everyone who has like good use cases for automation and real tangible benefit that can be passed to your customers and team for growth.

1:11:30I invite everyone to just go to smithos.com and click on early access and let us show you what we've built, see if this is a good fit. And if we can start building that future together. I mean, it will happen whether Smith ends up the category leader in the space or not. It doesn't really matter for the purpose of this podcast. I believe we have an early lead. We have been working on this for a year already, way, way, way before others, because, you know, I'm secretly a time traveler from the future. It's just so many things. Right. Right. And and so, yeah, if you if you are excited about this journey and you're truly passionate about about these things and want to see where the future is headed, give us a shout out and early access and somebody from our team will will talk to you and we'll see if this is a good fit.

1:12:27Alexander, thanks for being on the podcast. We appreciate it. Thank you.

1:12:36Thanks for listening to Built Right. If you enjoy the show, give us a follow or subscribe on your favorite podcast platform. And don't forget to leave us a review. For more info on Built Right, visit us at HatchworkBuiltRight.com. The single biggest mistake we see companies make with AI is they don't properly train their teams. We see it all the time. Companies roll out AI tools and expect people to just figure it out. But using AI effectively requires a totally different mindset and skill set. And that's exactly why we built training for every level of your org, from AI training for teams and executives to training engineering teams on our generative driven development methodology.

1:13:16Or if you've already identified your AI use cases and want to just prioritize where to start, we offer an AI roadmap and ROI workshop to help you build a quick plan. It's all about going from we should use AI to actually driving real value with it. Head over to Hatchworks.com to learn more.