I appreciate the attempt at making sense of conversational interfaces, but I don't think natural language as a "data transfer mechanism" is a productive way of doing it.
Natural language isn't best described as data transfer. It's primarily a mechanism for collaboration and negotiation. A speech act isn't transferring data, it's an action with intent. Viewed as such the key metrics are not speed and loss, but successful coordination.
This is a case where a computer science stance isn't fruitful, and it's best to look through a linguistics lens.
I quoted the bit I wanted to react to, namely the idea that businesses have to give customers what they want. I read that as strongly influenced by an economically driven - rather than morally driven - world view.
Leaving money on the table does not necessarily mean a failing business, except for some extreme definition of 'failing'.
I meant to argue against your first assertion, not your second; I'm not concerned with whether it's a bad financial decision or not.
> Newspapers are a business and have to give their customers what they want.
I really want to challenge this idea. Businesses can have missions quite distinct from what the majority of their prospective customers would want.
If I had practically unlimited money I wouldn't ever think of funding a news organisation and then only have it produce content that customers wanted. I would have a purpose for it, stemming from my own ethics.
I think it quite naive to consider Bezos has not done the same and that this decision is simply in line with his personal political interests.
Neoliberalism is a really poor substitute for personal morality and accountability.
Donald Schön’s work on this topic is really enlightening. It’s not just that there’s knowledge in the heads of people that can’t be linguistically expressed well, but also that expression of it requires interaction with a specific situation.
This is also represents a massive gap for AI systems to become actual in-the-world problems solvers.
Finding a proof is a search in a large space, akin to searching from abstractions to concretions. LLMs don’t do anything like this, and so you’re looking at the planning problem again. It’s not clear to me how framing this particular problem as a language problem is helpful in any case.
I remain sceptical about any form of combination of reinforcement learning and LLMs.
Acting successfully in the world when faced with complex issues requires learning useful ad-hoc concepts from the specific situation you find yourself in. It's plausible an AI can learn template tactics from large datasets, but I don't think that's enough.
There are various fields, like creativity research and design thinking, where it's understood that non-trivial problems need interaction with the environment to frame a problem in a way that allows an approach to a solution. This is because of the uniqueness and novelty in the situation itself.
It might be my lack of imagination, but I don't see how a deep learning on a large data set will get there.