AI Models – Techstrong AI Podcast EP32

Hello, and welcome to the Techstrong AI Podcast. I'm Amanda Ani, and I'm excited today to have Jacob Lorgan. He is the CEO of neurologic.

How are you doing today? Very good, very good. Thank you, Amanda.

Thank you for being on our show. And our topic of the day is AI models, and you said recently that AI model collapse, uh, you believe it mirrors how our historical narratives have been influenced by religious and political forces. So can you share that a little bit with our audience?

What do you mean by that? Yeah. Well, you know, matter, when I think about AI model collapse, uh, often compared to historical shifts shaped by religion and politics, uh, because if you look back at, uh, pivotals of moments in history like reformation and the enlightenment, uh, this one just shift in power.

There were paradigm shift in how we understood the world. So similarly to what's happening with AI right now and the collapse of old models, is that we had you say, we have absorbed all the knowledge from say, all time that has been written, and try to kind of collect all that to you could say, trains the train, the, the general, uh, ai, uh, general AI models. Uh, and then at the same time, we want them to kind of be adjusted for how we see the world today.

Uh, not being biased, not being too radical, uh, and not being too opinionated. But at the same time, that might be, you could say the, uh, you could say that might be what most people think, uh, or mo most people's opinion or view of the world, although it might not be correct. So as we kind of filter it through kind of you can say what we believe is useful, and we are producing that much more content.

Uh, just like with internet, with all the media outlets, uh, those pro suddenly made available A lot more content were written by people who are not journalistically trained or did not, uh, you can say, uh, look, source, look up, source, et cetera. The same way when they get filtered through and so much more content is produced, and the AI starts to absorb that content again, it becomes, you could say, a reinforcement. So it's, you say it's not because I think it's so much, it is so dangerous.

We just need to be aware that at some point it's just not gonna get any better. And it's gonna, you say, say, wa get watered out. When we see kind of a, a huge advantage is, is where we, where we want to you say, uh, empower people to do things they couldn't do before.

And we can start making use of you say, when we do steer, uh, the, uh, the, the, the outputs. And you could say, uh, and, uh, introduce a regime where it's, we, we where we need to sit up, put up more, you could say control and testing, like in legal or like in software that that, that we have introduced with Henrik ai. Then we, when we are empowering at talent, and you could say, make me build, uh, software or make you as good as a paralegal, we actually do wanna, wanna, you could say, add those, you could say somewhat, uh, guardrails or synthetic data processing.

So, so that the outcome is, is predictable. So we can ask for an outcome and we get, uh, uh, uh, and we get an outcome that we can test and say, well, this was what I expected. This is what a con contract or a rental lease should look like.

Or I want an application for food delivery. Yes, it's got all the elements that is required for food, uh, food delivery app. And then there's obviously some creativity around, you could say how the design looks in, in a, in, in an application, or you could say the language in, in a contract.

But, but in, in, in essence, there is a, a, a right or wrong. So there's so much what we can do with, with synthetic data, data now, what we have, you could say, train the, the general, uh, the general models. So With synthetic data, when it, when it comes to training, what are some tips or advice you have for business leaders that are trying to ensure that their AI is trained well and and trying to avoid the bias?

Yeah, I mean, one of the main challenges, uh, develop de developing purpose-driven AI is aligning the AI with the real world needs. Uh, AI can't just be generating code or performing isolated tasks. It has to be about solving the actual problems that people face.

Uh, within ai, we address this by integrating, uh, you say AI agents that are trained, uh, across multiple disciplines, what we call, uh, disciplinary me metrics, which is the framework that we spend about two years building. Uh, this framework ensures that the AI understands the context of what it's building, whether that's regulatory, compliance, user experience, scalability. There's not just about, in our sake building software, it's about building something that's reliable and, and, and, and, and, and us usable.

So you say it requires there a high level of domain expertise in both you could say, uh, training, uh, and organizing and, and testing the testing the outputs. We hear a lot about centers of excellence and different focus groups when it comes to training ai. What do you see as the future of training ai and do you think groups like this are helping?

Oh, absolutely. I mean, for you say, uh, for, uh, as legal, uh, ai, uh, I don't think that could be trained by somebody who, like, by a group of people that had no, uh, experience within, uh, you could say legal work. I believe that while you are putting the application into use, it would be su it's super beneficial to have, you could say, legal experts on the sideline to either offer, you could say final improvement if it's needed by the, the, by the user.

Or you could say, uh, or you could say, uh, and, and constantly testing, testing the data, making corrections, putting in building scenarios, uh, that kind of pushes, pushes the limits. And I think that will, you could say, continue to happen over time. Uh, uh, let's finish.

You could say some, some talk about like, we, in, in the future that will never, that there won't be, uh, the need of, uh, software developers. Uh, I see quite differently. I think we need, you could say more software developers, but you could say similar to, to you could say journalists, uh, today and, and, and like yourself, there'll be like, there'll be so much outlet of, of opinion, or in our case software, uh, that the, the scale will kind of multiply it by you could say mil millions.

Uh, and you could say software developers will then be the expert that is, you say, required when you need something. That is, you could say, when we need to innovate, uh, in, in fields that, uh, that is, you could say that is pushing, pushing the boundaries of, of software or IT compliance is needed, checks and balances, et cetera. So you still need, you could say these, uh, curators, uh, in, in, in, in the world of something, say, created by, by, by ai.

So, uh, so I just see, I see new and more exciting roles, uh, for people who are domain experts today. And I definitely believe that when it comes to, to, to building, you could say these specialized, uh, AI and AI platforms, then they should be built by, uh, people with the, with the right level experience to achieve, uh, the, the outcome that we are, that, that save the, the optimal, uh, outcome. So it's, it's, it's that synergy.

Well, this technology, we've seen it advance so rapidly when it comes to AI and future use cases with ai. What do you see? And are we anywhere close to AI hitting the age of singularity and being just as intelligent a as humans?

Uh, no. Uh, not at all. But it, it will become our assistant going forward.

I also don't believe that, we'll, uh, I'm, I'm very optimistic around ai. I don't believe we'll ever see kind of terminators, uh, running around trying to, uh, take us out. Uh, it's nice.

It's a fluent process. No, it's a fluid process like everything else is, uh, has been, I mean, a, a fun example. I mean, I think I was five when I saw a movie where, uh, the, uh, main character was speaking to an ai, uh, having a conversation that was obviously my biggest dream by then.

So I was super excited when suddenly you could have like a real voice conversation with an AI that felt somewhat natural. Uh, but that lasted for about five minutes and that was it, uh, because the double C wears off quite quickly. Uh, so, and I think where, uh, just like say internet, I mean, if you don't have it, you can't live without it.

But in your day-to-day life, you don't really think about that. It's there. I mean, I, I wouldn't be able to find myself, uh, fi figure out driving around in London, New York, or even San Francisco for that sake, without navigation that requires internet.

But then again, it's, it's, it's not kind of, you say overwhelming my life. And I think AI will be the same thing. It'll be a natural assistant that makes me capable of do, do, to do things I couldn't do before.

And who knows, if we look into the future and if, uh, Elon is, uh, uh, is, is is successful in, in kind of extending our brain capabilities with synthetic, uh, brain, brain cells, and we can hook that up to ai, will that then be kind of a, a fun emotion at some point where, uh, we don't really know whether we are, we're using our synthetic or our biological, uh, brain to, uh, to solve tasks. But it even then, it would probably be something that is, that feels, uh, quite fluent. Uh, I don't think it will, you say, take over our lives, but obviously as we get older, it will seem weirder to us.

Uh, but for new generation, it, it generations it will be absolutely natural. I mean, the anyone born today will grow up with, uh, with AI and, and not think that is never not existed. Uh, and it would be a natural tool.

So Yeah, I believe that is gonna be the case. So if there was one key takeaway that you could leave our audience with today, what would that be? Uh, that AI has the power to d democratize innovation.

Uh, and with, uh, Rik ai, we are allowing the barriers to entry so that anyone can build software, create solutions, and solve real world problems without being limited by technical skills or resources. Uh, so this is about empowering people to turn their ideas into reality at scale with the power of ai. Wonderful.

Well, I really appreciate you coming on the show and sharing your insights with us today. Thank you very much. Thank you for having me.

And thank you to our audience. Stay tuned. There's a lot more coming up.