Just because transformer-based architectures might be a dead end (in terms of how far they can take us toward achieving artificial sentience), and the outcome may not be mathematically provable, as this author seems to want it to be, does not mean that the technology isn't useful.
Even during the last AI winter, previous achievements such as Bayesian filtering, proved useful in day to day operation of infrastructures that everyone used. Generative AI is certainly useful as well, and very capable of being used operationally.
It is not without caveats, and the end goals of AI researchers have not been achieved, but why does that lessen the impact or usefulness of what we have? It may be that we can iterate on transformer architecture and get it to the point where it can help us make the next big leap. Or maybe not. But either way, for day to day use, it's here to stay, even if it isn't the primary brain behind new research.
Just remember that the only agency that AI currently has is what we give it. Responsible use of AI doesn't mean "don't use AI", it means, "don't give it responsibility for critical systems that it's ill equipped to deal with". If that's what the author means by "serious applications", then I'm on board, but there are a lot of "serious applications" that aren't human-life-critical, and I think it's fine to use current AI tech on a lot of them.
Even during the last AI winter, previous achievements such as Bayesian filtering, proved useful in day to day operation of infrastructures that everyone used. Generative AI is certainly useful as well, and very capable of being used operationally.
It is not without caveats, and the end goals of AI researchers have not been achieved, but why does that lessen the impact or usefulness of what we have? It may be that we can iterate on transformer architecture and get it to the point where it can help us make the next big leap. Or maybe not. But either way, for day to day use, it's here to stay, even if it isn't the primary brain behind new research.
Just remember that the only agency that AI currently has is what we give it. Responsible use of AI doesn't mean "don't use AI", it means, "don't give it responsibility for critical systems that it's ill equipped to deal with". If that's what the author means by "serious applications", then I'm on board, but there are a lot of "serious applications" that aren't human-life-critical, and I think it's fine to use current AI tech on a lot of them.