I don't think that's particularly true. The whole point of the Millennium Prize problems as the article states was not on absolute difficulty of the problems but on the high chance of a proof producing fruitful results leading to new concepts and theories. Any pursuit of capital T truth will of course aim for better and more clarifying abstractions and is not a subjective turn by any means.
If you mean > What is missing is an intelligible proof that human mathematicians can understand and use to advance the aims of mathematics.
Then that is not necessarily subjective either if an AI can produce an actually intelligible proof. The problem is that as mathematicians with PDE expertise have mentioned on Twitter the actual solution seems to devolve into an unreadable mess focusing on irrelevant details after a more readable first few pages in the proof. If it wasn't a Lean compiled proof and presented as a human artifact, it would be hard to assess if the deviser of the solution had any actual understanding of the solution.
If it doesn’t seem like the lllm understands the proof, it probably doesn’t.
Just because it generated something that works doesn’t mean it understands it, and the evidence from its proof is that it does not, which is not very surprising given the technology we’re talking about, which generates likely phrases based on a corpus and training.
The contention is usually that empathetic policies that favour outgroups can be exploited by that outgroup to benefit themselves. For example, a universal policy of hiring from all groups can be subverted by allowing in a group that hires only their own members.
But that belies the real contention of what exactly is the truth? What is even more fundamental that has been lost is trust in nation and institutions due to a growing values divide. We must first trust before we can accept in good faith. Perhaps in some sense we should be thankful that we now have societies robust enough that even if major institutions lose legitimacy with half the population we don't get a civil war.
It should also be noted another paper also came out today with new ancient DNA from a modern human whose bones carbon dated to around 45kya. The data suggests the individual was only 80 generations from the Neanderthal interbreeding event suggesting it happened around 45-49 kya. The event may have happened over a few generations realistically but I think 7000 years is unlikely due to the rapid expansion of humans out of Africa (or to be more precise an extremely successful expansion of a population of modern humans whose signal is difficult to filter out) all sharing this same introgression signature.
The timeline overlaps with the Peace of God movement and the Gregorian reforms which legal historian Harold Berman calls the Papal Revolution in his book Law and Revolution: The Formation of the Western Legal Tradition. Also coincides with the beginnings of renavatio that Charles Taylor mentions in A Secular Age. You could argue though that perhaps it took some time until the population fully Christianized; it is precisely in this time period we start seeing the use of familiar Biblical along with a set of standardized Germanic first names all across Europe while old Germanic naming conventions start to disappear.
> And within those pieces of fiction they will generally refer to their headcanon, but it's no more special than the fans'.
If it's the same author writing these pieces of fiction, then speaking by definition, the author's opinion is more special by being the creator of those works and therefore can create fiction that reinforces their headcanon (which is why it's called canon). So I think calling the author's opinion less special is wrong for the author's tie to the work will always be more special than the consumer by virtue of being the creator.
Falcon uses apis like eBPF when available/usable , they are not stupid if they can use something which is more secure and reliable why should they not use it.
E.g. they use it on Linux, even through they could have created a custom kernel module (idk. if they maybe also have a custom kernel module tbh.).
And pushing for something doesn't mean banning other things. E.g. they could certify "following best security practices" and not give it to vendors not using the more modern APIs, while they can't block drivers based on it with the right marketing customers of CrowdStrike wouldn't want to buy it without such cert.
I.e. while MS doesn't provide viable ways to get the functionality Falcon and similar need without kernel modules it indeed would be a bit ridiculous for them to ban such software, and as of yet the do not.
Yeah, a lot of these articles conflate use of zero as placeholder, numeral and number. But the real critical conceptual step is the last step of using zero like any other number (mostly) in arithmetic.
> It's almost as if he's being overly negative about them to make a point or something to push back against all the unbridled optimism.
I don't think it is like that but rather Chollet wants to see stronger neuroplasticity
in these models. I think there is a divide between the effectiveness of existing AI models versus their ability to be autonomous, robust and consistently learn from unanticipated problems.
My guess is Chollet wants to see something more similar to biological organisms especially mammals or birds in their level of autonomous nature. I think people underestimate the degree of novel problems birds and mammals alone face in just simply navigating their environment and it is the comparison here that LLMs, for now at least, seem lacking.
So when he says LLMs are not sentient, he's asking to consider the novel problems animals let alone humans have to face in navigating their environment. This is especially apparent in young children but declines as we age and gain experience/lose a sense of novelty.
