Viewing profile — versteegen

> For me it’s the massive amount of resources it takes to produce and run one It's amazing that LLM pretraining is both extremely data inefficient at learning concepts and cognitiv…

You misread. "Pain and suffering" not "death". Of all the pain and suffering in the world, a vast amount of it really is our own fault. Famines and wars shouldn't happen. And if yo…

To save anyone else the trouble: discussion there is not really worth looking at (largely a flame war), except: the author of this disproof seems to be a crank, and the disproof's …

...but we're talking about compaction, and opencode's compaction is (or was) terrible. I've seen so many horrible problems that I keep it disabled (with an envvar flag, because eve…

Wow, remarkable. Clearly Aum is a different league from the lone-wolf "Fort Detrick guy", treat the risks separately. But I'll take these questions as rhetorical. (See my reply to …

I didn't argue "must be regulated". I'm arguing AI is powerful (at achieving things, and hence has dual-use dangers). Many people won't even admit that, which is the part that real…

Thank you for taking this seriously enough to write this, and anyone else likewise. But this is attacking a strawman, amateur bioterrorists. AI is a force multiplier in the hands o…

The old observation that people tend to define AI as whatever computers can't do yet is as true as ever. It's getting a bit absurd, moving from demanding "general intelligence" to …

Isn't it ~$3000 per week? Extrapolating from the current limit on Pro plans.

Disagree. Actually, in API cost equivalent, the $20/mo ChatGPT Plus plan gives you ~$100 of usage, while $20/mo Claude Pro gives you >$250 of usage (I measure at ~$300 in my last w…

> It looks like it wrote a python script to generate test cases in our file format for testing. Just... you know, as a side quest. On the one hand, agents have done this sort of th…

Algebra is useful because graphs are algebraic objects, and a lot of CS is about graphs, in particular search/planning. But no, I never saw rings mentioned except for generating fu…

Having studied CS and maths to post-grad, I think you exaggerate. Although a CS course might use these tools, they didn't in my experience go into explaining or defining them. The …

That was a tendency of 5.4 and earlier, OpenAI specifically worked to avoid it in 5.5 and I find it happens rarely know. It really felt like 5.4 had been intentionally trained to s…

"Thinking" seems to be a political term now, people have completely different definitions of it, based on how they wish the world to be organised, and find defining it differently …

Being correct doesn't give you licence to use insults.

Not half, only on the order of 10% of the cost of inference is electricity.

Pretty sure there isn't. HN has lots of hellbanned bots lately (I have showdead on [1]) but I never notice ones pushing a political slant. But I suppose nation states conducting a …

I always run opencode with envvar `OPENCODE_DISABLE_AUTOCOMPACT=1` because I discover a new horrible bug in its autocompact every time I don't... including the config .json option …

BTW the quotas for Go have very recently changed, now only $15 for some models instead of $60. Which is not actually a difference for DS4 Pro, because they lowered the token pricin…

To summarise the full results table further down the page (which doesn't render on the page for me!): Kimi K3 beats each model (out of 35 benchmarks, excluding missing): vs Fable 5…

It's a custom 3 bit quant of Llama 3.1 8B and other shortcuts. The quant is not good. Their newer arch switches to standard 4 bit quants, should be far better!

I think MiMo 2.5 seems to be better than DS4 Flash, at the same price. DS4F can write pretty advanced code but it way overthinks the simple stuff, its CoT is full of errors (immedi…

I've always been amazed that any time (very rarely) that I look at the tex source for a paper it's full of commented-out things not cleaned up. Nobody thinks to look?