inimino

8 ms·

61.
by inimino 2y ago
It's the economically viable part that I think is currently hard, but agreed this is the right approach. The basic problem is that scaling up understanding over a large dataset requires scaling the application of an LLM and tokens are
62.
by inimino 2y ago
As someone who has worked on LLMs somewhat extensively, the idea that we are going to accidentally make a superintelligence by that path is literally laughable.
63.
by inimino 2y ago
> Wetterhahn Who, just to be clear, was not being cavalier, but followed all the relevant precautions known at the time. They were just inadequate.
64.
by inimino 2y ago
Simpler yet, just tell the model "Reply with 'Yes' or 'No'."
65.
by inimino 3y ago
(Offtopic, but just FYI) it's tenet (principle) not tenant (building resident).
66.
by inimino 3y ago
Bird brains, maybe. I'm not sure what your expectations of GPT5 are but I meet GPT4's limits every day.
67.
by inimino 3y ago
The idea is that you bet before the sequence is known. Nowadays we would say it is the distribution (or the process producing the random sequences) that can be truly random or not, and we recognize that saying "sequence [...] is random
68.
by inimino 3y ago
Sure, that would be true. I think people are more interested in talking about the AI we actually have today.
69.
by inimino 3y ago
You're misinterpreting it. It's not a descriptive title at all. It's a tempo indication and like all such is meant as an instruction to the player, also being a play on "as fast as possible" which is a perfectly leg
70.
by inimino 3y ago
It is absolutely doable, the question is whether you have the combination of time and skills to make meaningful progress in whatever direction you choose in whatever time you have. Probably start with a review of recent literature (i.e. not
71.
by inimino 3y ago
I used to be neighbors with the owner of this business. Good people. Never in a thousand years would have expected to see this on HN though!
72.
by inimino 3y ago
No, the point was that LLMs are not good at problem solving because they are not good at problem solving, not because they were not evolved. We don't understand how we or animals solve problems, which is why we haven't yet succeed
73.
by inimino 3y ago
Evolution operates on genes, which do not encode synaptic connections for one thing. The analogy you're making here is so stretched it's hard to begin to say what's wrong with it. Backpropagation and natural selection are abo
74.
by inimino 3y ago
> Aren't animals trained to do all of those things through evolution? Similarly how GPT is trained. No. Animals evolved and are able to do those things. Evolution is not training, and evolution has approximately zero to do with how
75.
by inimino 3y ago
Not ignored! It's precisely because software engineering discards the very parts of programming that are closest to real engineering that this term came to describe it. "Don't bother learning the fundamentals, but focus on bu
76.
by inimino 3y ago
You mean all the landscape artists and portrait and figure and still life painters throughout all human history who just casually made art of whatever they saw around them?
77.
by inimino 3y ago
The "imitation game" that was Turing's model was a parlor game with a man and a woman behind a screen. The interlocutor could ask either of them any question. The winner was whoever convinced the questioner that they were the
78.
by inimino 3y ago
Your argument assumes that p-zombies can exist, which is unproven and I consider highly specious.
79.
by inimino 3y ago
Compression and prediction are equivalent, and both measure some function of intelligence and knowledge. Current LLMs are heavy on the knowledge side, being trained on ~all the text.
80.
by inimino 3y ago
You may have missed who I was replying to in the thread. The toplevel comment quotes Teflon's website talking about "cookware" in a generic sense, like heating up a pan with nothing in it is universally bad. That's bulls
81.
by inimino 3y ago
Fortunately your stove isn't capable of burning steel. Preheating the empty pan is a basic technique in more than one culinary tradition.
82.
by inimino 3y ago
As an ancient Greek, Homer would certainly never have worn formal clothing of a civilization that did not yet even exist...
83.
by inimino 3y ago
Yes. You can also take the log of the reciprocal probability (or "luck") and then you don't need the negative sign. This value is the expected number of trials you would need for the outcome to occur once. I find this present
84.
by inimino 3y ago
Low entropy data is what's more compressible. High entropy means unpredictable, which could mean high noise (like TV static which is incompressible) or high signal as GP intended.
85.
by inimino 3y ago
If the colors are symmetrical, then it's just "my pieces" and "opponent's pieces". You could encode who has the move by how the board is ordered, and predict it from the king and queen position, then just encod
86.
by inimino 3y ago
Check out Peano arithmetic. Intuitively, natural numbers come from and are defined by counting, and that implies that "3" means inherently that something (could be anything) happened or was repeated three times. For example, if yo
87.
by inimino 3y ago
Publishers wouldn't choose papers and ink that smelled bad. There's a lot of ways to make paper and ink, so it's not that surprising that we'd eventually nail the aesthetics.
88.
by inimino 3y ago
Who benefits from this "fraud" you are imagining? How does that scenario play out in your head? "Ha! You think this costs us mere pennies of electricity, but actually our highest paid employees are personally spending hou
89.
by inimino 3y ago
Have used vim with the same vimrc for c. 20 years, probably will continue for the next 20. What I like about vim is that it just works, and any features I don't need are turned off. My configuration is minimalist, only about 20 lines.
90.
by inimino 3y ago
> For each step, we estimate a probability of success by 2043, conditional on prior steps being achieved. [emphasis added] Maybe your reading comprehension is the problem.

More ›