Some Erdős problems are basically trivial using sophisticated techniques that were developed later.
I remember one of my professors, a coauthor of Erdős boasted to us after a quiz how proud he was that he was able to assign an Erdős problem that went unsolved for a while as just a quiz problem for his undergrads.
Not definitively. LLMs are stochastic with respect to input, temperature and the exact prompt. It's possible that the model was already capable of it but never received the exact right conditions to produce this output.
It "exhaustive brute forcing" approach does not need an LLM in the loop. Just brute force the possible outputs instead. They will contain all the most beautiful novels you can imagine!
> So this is proof of the models actually getting stronger (previous generations of LLMs were unable to solve this one).
No, it's not.
While I don't dispute that new models may perform better at certain tasks, the fact that someone was able to use them to solve a novel problem is not proof of this.
LLM output is nondeterministic. Given the same prompt, the same LLM will generate different output, especially when it involves a large number of output tokens, as in this case. One of those attempts might produce a correct output, but this is not certain, and is difficult if not impossible for a human not expert in the domain to determine this, as shown in this thread.
This is one of a number of such results achieved only in the last few months with only the last crop of models. They have undoubtedly gotten better in this domain. Saying anything else is just denial. You can run these same problems on GPT-4 or 5 all you want, you'll get nowhere. In fact people did, and you're hearing about it now because it's these crop of models that are getting meaningful results.
As others have pointed out, a key part of the prompt used here may have been "don't search the internet" as it would most likely have defaulted to starting off with existing approaches to that problem...
Context: parent originally said "you should not say 'worth mentioning', if it's worth mentioning you can just say it". That sentence has now been edited out so my comment looks weird.
Tao mentions that the conventional approach for this problem seems to be a dead-end, but it’s apparently a super ‘obvious’ first step. This seems very hopeful to me — in that we now have a new approach line to evaluate / assess for related problems.
I think they would have a very strong case that using the mouse on a product is likely to confuse consumers about the origin of the product and therefore infringe on their trademark.
I'm sure it's an oversimplification to blame the entire 1970s AI winter on Minsky, considering they couldn't have gotten much further than the proof-of-concept stage due to lack of hardware. But his voice was a loud, widely-respected one in academia, and it did have a negative effect on the field.
I suspect all Minsky did was reinforce what many people were already thinking. I experimented with neural nets in the late 80s and they seemed super interesting, but also very limited. My sense at the time was that the general thinking was, they might be useful if you could approach the number of neurons and connections in the human brain, but that seemed like a very far off, effectively impossible goal at the time.
I saw this video yesterday and considered posting it, but I wasn’t sure if it was appropriate for HN.
This channel has another video where it shows how the clean room lab is created starting from a basic backyard shed, and that was truly astounding. The positive pressure to keep the number of particles low in someone’s backyard is almost mystical to me.
Which is a pity, because lots of videos really need to be seen to be fully appreciated. Especially the ones showing stuff being made. And the ones that tend to show up here are usually worth the time.
I'm totally with the text folks on the 5 hour Fossdem sessions, though. Give me an accurate transcript I can grep or don't even bother.
I think if it's interesting to you then it's worth posting, and letting the voting system do it's thing. I only rarely post because by the time I've seen something it's usually already been posted
why even allow the cognitive overhead to worry about such a thing? it's not for you to decide anyway - let the users decide using the voting system that's built to task
Recently I saw a post about Bonsai trees on the front page. Making your own RAM is 100% more relevant to HN than quite a few posts I see on the main page.
You’re not sure if someone building a RAM clean room in a shed is appropriate for HackerNews, literally “news for nerds”? A dictionary purchase may be warranted
I think he plans to go far beyond just making RAM in that clean room. This is pure speculation, but I suspect the goal of that channel is to just make doom from scratch.
Given that the shed in this guy’s backyard is already approaching the entire national technological output of any country in the 1970s I think he may get there.
To put it that into context, some tags count as upvotes, others count as downvotes, "Troll" is a downvote. So to have your post labelled as "Troll" with a positive score, it has to have enough upvotes to compensate the penalty from the "Troll" votes, but without having another tag dominate. 5 is the maximum score.
"Score: 5, Troll" is therefore the mark of a very successful troll.
They also have "Underrated" and "Overrated" which apply points but do not act as tags. So I guess the easiest way to get +5 Troll is to have many Troll and Underrated votes, if it works the way I think it does.
There's no mention of AGI, climate change, AWS outages, Trump, crypto schadenfreude or my new MVP that you should totally sign up for even though I just vibe coded it 20 minutes ago and the DNS hasn't fully propagated yet, but the API is amazing plz like comment and subscribe.
Ok, maybe I'm being a bit cynical. Stories about bikeable cities are welcome too. And wasn't Soylent popular for a hot minute back in the day?
C certainly gives the illusion of portability. I recall a fellow who worked on DSP programming, where chars and shorts and ints and longs were all 32 bits. He said C was great because that would compile.
I suggested to him that he'd have a hard time finding any existing C code that ran correctly on it. After all, how are you going to write a byte to memory if you've only got 32 bit operations?
Anyhow, after 20 years of programming C, I took what I learned and applied it to D. The integral types are specified sizes, and 2's complement.
One might ask, what about 16 bit machines? Instead of trying to define how this would work in official D, I suggested a variant of D where the language rules were adapted to 16 bits. This is not objectively worse than what C does, and it works fine, and the advantage is there is no false pretense of portability.
reply