smokel

14 ms·

241.
by smokel 1y ago
Optimization is harder than writing out the rules of a game. For most board games, it is trivial to describe all possible next states, but it is not at all trivial to search through all of these to find the best action to take.
242.
by smokel 1y ago
That's only in heat transfer, that does not include generating electricity from it.
243.
by smokel 1y ago
Minor nitpick: it did not use preprogrammed rules for scanning through the search tree, but it does use preprogrammed rules to enforce that no illegal moves are made during play.
244.
by smokel 1y ago
Implementing a world model seems to be mostly solved by LLMs. Finding one that can be evaluated fast enough to actually solve games is extremely hard, for humans and AI alike.
245.
by smokel 1y ago
You probably know this, but things heavily depend on the type of board game you are trying to solve. In Go, for instance, it does not help much to look 50 moves ahead. The complexity is way too high for this to be feasible, and determining
246.
by smokel 1y ago
I'm not opposed to capitalism, and enjoy its benefits everyday. I don 't however agree that efficient markets are required for capitalism. I also don't think that HFT is the only way to create efficient markets. I do howe
247.
by smokel 1y ago
It would be nice if smart people would not require law to operate in an ethical way. For complicated problems, and for people who are just plain stupid, it's nice to have law and law enforcement, so that capitalist society can work pr
248.
by smokel 1y ago
They could spend their time playing chess, solving Fermat's last theorem, or invent space weapons. And why stop at humanity?
249.
by smokel 1y ago
What do you base this on? Typically it's plans like these that offer more privacy, because people are willing to pay a premium for that.
250.
by smokel 1y ago
Perhaps I should've made myself more clear. Monitors, keyboards, programming in textual representations, all seem quite unnatural. They were all the result of incremental technical progress, not the result of an ideal thought process.
251.
by smokel 1y ago
One might also take on the more cynical perspective and be disappointed that we are still stuck with these early achievements. FCOL most of us are now happy to have our AI overlords type out software on 80 column displays in plain ASCII bec
252.
by smokel 1y ago
Would you like to also be able to communicate about this number? You might have to reserve some atoms to form a being that could actually enjoy it. Considering such a perspective, the decoder for observing it should probably be embedded i
253.
by smokel 1y ago
Mind sharing what you have tried? Have you considered training a diffusion model on pixel art, and then conditioning it on a 3D model?
254.
by smokel 1y ago
This is very helpful for understanding generative AI. See for example the amazing lectures of Stefano Ermon for Stanford's CS236 Deep Generative Models [1]. All lectures are available on YouTube [2]. [1] https://deepgenera
255.
by smokel 1y ago
He uses quite a bit of tooling, including lasers . It's not like he would be able to get this far in the middle of nowhere :) In a way it is somewhat similar to people writing demos for old computers using emulators. Still great fun
256.
by smokel 1y ago
Hehe, very nice to see something outside the scope of software or PCBs with this level of useless enthusiasm. Obviously "from scratch" is a bit of a stretch here, but this is the material we come to Hacker News for. Thanks for sha
257.
by smokel 1y ago
It seems highly unlikely that they did not use indices. Scanning all documents would be prohibitively slow. I think it is more likely that the indices were really large, and it would take hundreds to thousands of machines to store the ind
258.
by smokel 1y ago
Also, as soon as the code no longer fits within the context window of an LLM, one must resort to RAG-based solutions, which often leads to a significant decline in quality.
259.
by smokel 1y ago
What you are describing also seems to align with the idea that greenfield projects are well-suited for AI, whereas brownfield projects are considerably more challenging.
260.
by smokel 1y ago
While I might agree with your standpoint, how is this different from also influencing? I've seen a lot of influencers suggest "100% assembly", "JavaScript only", "no SQL", which seem quite similar.
261.
by smokel 1y ago
If it isn't nature, then it probably is nurture. Averaged over the entire population, that is indeed mostly random.
262.
by smokel 1y ago
The comments here are priceless. In less than five years time we have gone from "That's impossible" to "Meh, it doesn't solve P=NP if prompted.". For those commenting in the latter category, it might be worthw
263.
by smokel 1y ago
As seen from the AI's perspective.
264.
by smokel 1y ago
> what does anything matter at all if the sun is going to explode It would not surprise me if this is the exact reasoning that underpins decisions made by leaders of these big companies. It's terribly hard to convince some people
265.
by smokel 1y ago
There are many places at Meta that seem to be quite interesting for researchers. You get to play with a lot of hardware, with other talented people, and you can open-source some of your work. It's all a slippery slope anyway. If you
266.
by smokel 1y ago
Note that democracy does not necessarily require one leader.
267.
by smokel 1y ago
Having worked extensively on large charcoal drawings using heavy etching paper, I’ve often imagined what it would be like if a proper robotic arm could take on the task. My own process involves compressed charcoal pencils, plenty of erasers
268.
by smokel 1y ago
My experience was exactly the opposite. Experienced developers know when the LLM goes off the rails, and are typically better at finding useful applications. Junior developers on the other hand, can let horrible solutions pass through un
269.
by smokel 1y ago
You can't get a citation for this, and I must admit that this was a bit of a hyperbole. Still, I sincerely believe that in a typical year, a typical user runs into zero or one .zip files. Of course there are exceptions, but these powe
270.
by smokel 1y ago
> WinRAR, a utility for compressing files, and has an installed base of about 500 million. Yeah, right. Edit: this figure is possibly taken from the WinRAR website [1]. It is more likely that there have been that many cumulative down

More ›