I was thinking about this for a while as well and I think there are several considerations... for one, LLMs tend to be fairly single-modal in their way of tackling specific problems. This aspect is likely to get amplified with multiple LLMs working in this organizational manner, leading to potentially very naive decision making or overly high optimism. I think this would remain true at a fundamental level, regardless of roles that each LLM is given. A key aspect of a human way of thinking is that humans have a kind of "breaking point", where, despite our interest in pursuing a specific direction, if we gain "enough" quantity of convincing information that goes against our present understanding of an issue, we will eventually reverse course. This reversal usually represents a kind of team-wide "strong acceptance" that the prior approach and all logic associated with it has been invalid. We can completely flip. Can LLMs do such a thing? Since the models are statistical in nature, I would question their ability to "aggressively adapt" to situations that they haven't foreseen; being more likely to pivot back to a class of solutions that got them in trouble to begin with, which could lead to cyclic failures. As the LLM architect, it would be a difficult call to make, whether or not to allow an LLM to act "drastically", since there could be endless unexpected consequences, but the fact remains that "drastic" adaptation is completely necessary, at times. A significant dilemma. Another consideration is data collection. Decisions are only as good as the data that they're based on. Humans are quite "innovative" when it comes to gathering data to support some specific direction, and we often combine this data with much more abstract observations about the world. We can also utilize complex psychological strategies, including for example assumptions about the evolution of adoption rates of certain classes of services or products by customers over time, which then drive competitive and "future-looking" development efforts. LLMs would probably be much more reactive and comparatively conservative in those types of roles. So, the concept is interesting and I'm sure it will be utilized in the future, but there would be plenty of concerns as well. Anyways, thank you for your videos as always! -Nick
@NicolasEmbleton
4 ай бұрын
A CEO LLM would probably help most companies in the world make make better decisions overall. Replacing the heads circles would also remove a large host of conflicts that basically destroy the companies from within. IMO this is a very hopeful future. Also very likely to happen relatively soon on the edge.
Пікірлер: 2