Symposium Part 3: “Roundtable on Implications of LLMs and Active Inference Agentic Architectures” ( Program )

SPEAKER_03: Okay, well, why don't we start? I'd like to thank everybody for being here. This is great. I think we've got a wonderful group and I'm really looking forward to the conversation. Let me put this in a little bit of context in what's happened in the last two sessions. What we really wanted to do... was be able to understand the transition between large language models that everyone talks about, but I think there's certain issues that we're seeing is that they're not getting the same level of improvement with the successive generations. So there's certain issues around explainability, transparency. What is the business case for them right now? Um, how do they make money and then also the fact that they're enormously consumptive of energy and energy resources. Um, and so part of that recognition, I think, is in the broader communities where people moving over to. agentic architectures. And part of that mentality is how to automate basically current workflow processes. So you see certain roles and tasks being automated, and then you may have specializations and the supervisor roles. But there is the discussion that we've had in the last two sessions, in which Andrew did an excellent job presenting, is you're moving into a new kind of model based upon really science-based principles of what it is to be intelligent, what is it to be alive, what is it to be autonomous, each thing with its own agents. So we have a new sense of agency that's coming in, fully autonomous agents, which is very different than what we have right now. and and i think that everyone would recognize that the the ability to automate the cognitive processes and we're seeing this it's not just a 10x improvement it's it can be a thousand x improvement and then when you put these different things together you're getting a whole different kind of organization there but we're in a new era that we've never been in before and i think our our reflex is to say well we have to have people in the loop But when we think about building new kinds of institutions, the idea of an institution and a rule of law is that they're above the individual. You actually build something that people have roles that they're supposed to play within that context that mirror the intentions of the institution. But now we're creating a whole class of agents, I would argue, and this is part of a whole discussion here, that actually can embody and by design a certain institutional perspective and scale that way. And that has real implications because what's the business case for that? What's the social contract? When people are saying, what happens to employees? And I was just mentioned earlier that I was listening to a piece by Peter Diamandis when he talks about his moonshots and his abundant society and saying that when you really have exponential change, you do it outside the organization. You may have a special group that is dedicated to that, reports to the CEO or whatever, but they're not part of the normal organization. And I think this is the case, what we have here, but it's even more fundamental than that. It's not like you're developing a new product, you're developing a new kind of enterprise. And I think where this, so there are some challenges at the legacy level. And I think, but on the concept of how to deal with, a new way of addressing climate change and biofinance, it's sort of a green field of how to create a new kind of economy. So it actually creates a lot of opportunity. So we're going to have certain optimism around that. But it is a big change. So having sort of laid that groundwork, what I'd like to do is begin with Simon, if you could begin, because you've been doing some work on AI risk. You've been writing just recently on these. On these issues from the enterprise perspective, and you can take it in direction that you're feel comfortable with, but I think it'd be good to to inform our group around that. And then Andrew, you can follow up on that because it…