The reason LLMs fail today is because there’s no meaning inherent to the tokens they produce other than the one captured by cooccurrence within text. Efforts like these are necessary because so much of “general intelligence” is convention defined by embodied human experience, for example arrows implying directionality and even directionality itself.
I’d like a source for that. College graduates are no longer at an employment advantage compared to their uneducated peers. The average age of a new hire increased by 2 years over the past 4 years.
Young people in the west have definitely seen declining salaries, if only by virtue of the fact that they’re not being offered at all.
Money and power don’t usually make you smarter, in fact they usually make you dumber. You can have every anti social belief and the intentions of the antichrist, but if you’re smart and run your system well everyone will still benefit.
AI actually has some optimizations unique to the field. You can in fact optimize a model to make it work; not a lot of other disciplines put as much emphasis on this as AI
Yes, in "runtime optimization" the model is just a computation graph so we can use a lot of well known tricks from compilation like dead code elimination and co..
For inference assorted categories may include vectorization, register allocation, scheduling, lock elision, better algos, changing complexity, better data structures, profile guided specialization, layout/alignment changes, compression, quantization/mixed precision, fused kernels (goes beyond inlining), low rank adapters, sparsity, speculative decoding, parallel/multi token decoding, better sampling, prefill/decode separation, analog computation (why not) etc etc.
There is more to it, mentioned 4 categories are not the only ones, they are not even broad categories.
If somebody likes broad categories here is good one: "1s and 0s" and you can compute anything you want, there you go – single category for everything. Is it meaningful? Not really.
Well yes, that's what he wrote, but that's like saying: stealing can be done for variety of reasons, including by someone who intends the theft to be discovered? Killing can be done for variety of reasons, including by someone who intends the killing to be discovered?
I read it as "producing racist videos can sometimes be used in good faith"?
They're saying one example of a reason someone could fake a video is so it would get found out and discredit the position it showed. I read it as them saying that producing the fake video of a cop being racist could have been done to discredit the idea of cops being racist.
There is significant differences between how the information world and the physical world operate.
Creating all kinds of meta-levels of falsity is a real thing, with multiple lines of objective (if nefarious) motivation, in the information arena.
But even physical crimes can have meta information purposes. Putin for instance is fond of instigating crimes in a way that his fingerprints will inevitably be found, because that is an effective form of intimidation and power projection.
I think they’re just saying we should interpret this video in a way that’s consistent with known historical facts. On one hand, it’s not depicting events that are strictly untrue, so we shouldn’t discredit it. On the other hand, since the video itself is literally fake, when we discredit it we shouldn’t accidentally also discredit the events it’s depicting.
So make fake videos of events that never actually happened, because real events surely did that weren’t recorded? Or weren’t viral enough? Or something?
The problem is deeper than economics. It’s the festering wound of reconstruction turning putrid. It doesn’t have to be the end of the US, but it certainly can be.
Also, I’m not sure the US economy was even great for most of the periods you mentioned. The question of if the US survives to have the same economic standing that it did in the 1800s is not that compelling
reply