kingstnap

22 ms·

361.
by kingstnap 1y ago
Only 4 revisions to requirements sounds like a really concrete app idea.
362.
by kingstnap 1y ago
Will they be cheaper to buy? Yes. You are effectively just buying compute with AI. From a simple correlational extrapolation compute has only gotten more cheaper over time. Massively so actually. From a more reasoned causal extrapolation ha
363.
by kingstnap 1y ago
Facts are public domain and can't be copyrighted (you can only argue your creative expression of them, ex in a textbook, is yours). But I suppose that wouldn't apply to disinformation. So your made up bullshit is yours to keep!
364.
by kingstnap 1y ago
Anyone able to hook your boy up with the older stuff after this gets built? I wanna test neural architecture search on ML models for algorithmic tasks.
365.
by kingstnap 1y ago
Some water usage has highly questionable counting methodologies. Like using if a datacenter is using hydroelectric power you count the evaporation from the dam reservoir as "used water". I'm not an expert but imo correct acco
366.
by kingstnap 1y ago
Make it long enough that it's not worth any human's time reading it. Like full-on balls to the walls slop. Pass it between different LLMs to add additional conclusions and subheadings. In fact, do it in parallel where one chatbot
367.
by kingstnap 1y ago
It's a ridiculous amount claimed for sure. If its 2 kW per it's around 5 million, and 1 to 2 kW is definitely the right ballpark at a system level. The NVL72 is 72 chips is 120 kW total for the rack. If you throw in ~25 kW for coo
368.
by kingstnap 1y ago
Yep. True webscale is astronomical levels of inefficient and wasteful usage of bandwidth and CPU.
369.
by kingstnap 1y ago
Are you really suggesting that people who are intelligent are purely that way because of their environment and experience? Any amount of observing children will show that equal instruction will not net equal outcome.
370.
by kingstnap 1y ago
I wonder how commodification of water is supposed to work in this case. Would we rent out water rights? A somewhat tangential point: I think it would be very cool if petro states started using oil money to set up massive solar powered desal
371.
by kingstnap 1y ago
> Only testes and ovaries are clearly distinguishable Is it just me who reads these kinds of things and just cringes. Always some nonsense along the lines of "well its not binary so therefore it's a spectrum". You have obv
372.
by kingstnap 1y ago
Somehow laptop makers all write complete garbage firmware. I'm sure dell does the same terrible handling of DGPU power and badly written ISRs that pointless raise system latency. I had shoddy crashes for months that would cause my dell
373.
by kingstnap 1y ago
Exactly this. Selling stocks <=> Buying USD. Are they buying bonds? Investing in foreign countries? Real estate? Gold? Crypto? Holding cash is surely a losing bet I doubt we will see any deflation even if a shit ton of people lose th
374.
by kingstnap 1y ago
No amount of LinkedEn speech can fix the poor part of it. In 2025, it's abundantly clear that the mask is off. Only the whales matter in video games. Only the top donors matter in donation funding. Modern laptops with GPUs are all $2k+
375.
by kingstnap 1y ago
Revocation seems really nasty to deal with. The whole chain of trust model is that your browser vouches for an authority that vouches for a website that everything is legit. You can't just ducktape on an idea like that cert for "w
376.
by kingstnap 1y ago
Well we have AI GPUs now so you could do it. Each MI325x has 256 GB of HBM. So you would need ~32 of em if it was 2 bytes per scalar.
377.
by kingstnap 1y ago
Runs of spaces of many different lengths are encoded as a single token. Its not actually inefficient. In fact everything from ' ' to ' ' 79 all have a single token assigned to them on the OpenAI GPT4 tokenizer. Sometimes
378.
by kingstnap 1y ago
There is this deeply wrong part of this paper that no one has mentioned: The model head doesn't hallucinate. The sampler does. If you ask an LLM when x was born and it doesn't know. And you take a look at the actual model outputs
379.
by kingstnap 1y ago
Extremely poor takes. I distinctly remeber back in 2017 or so I was using tensorflow to train a CNN. At the time Python really hadn't gotten its shit together with regards to typing. Googles Tensorflow documentation was heavily filled
380.
by kingstnap 1y ago
This seems like the kind of regulation that causes more harm than it saves in anyone's house actually burning down. I did a cursory search and didn't find any EU reports of a house fire caused by a L1 charger. Sure you might burn
381.
by kingstnap 1y ago
Cheat on the test by bringing the answers memorized and ready to use in your brain!
382.
by kingstnap 1y ago
It makes more sense when you approach it from linear algebra. Like you can make any vector in R^3 `<x,y,z>` by adding together a linear combination of ` <1,0,0> `, ` <0,1,0> `, ` <0,0,1> `, turns out you can also do
383.
by kingstnap 1y ago
If you don't want to answer clarifying questions, then what use is the answer??? Put another way, if you don't care about details that change the answer, it directly implies you don't actually care about the answer. Related s
384.
by kingstnap 1y ago
Not OP, but I think one way things shake out (at least for news) is this: Step 1: News journalists offer an LLM friendly private API to OpenAI and others. Maybe they create some special templated markdown articles. Step 2: OpenAI offers a
385.
by kingstnap 1y ago
Human curated high quality information on google is getting rare, its as simple as that. There is a renaissance of extremely high quality YouTube videos from creators with very few subscribers. Particularly in Math and Science content. But
386.
by kingstnap 1y ago
Teething pains are surely expected. But the egos of some of these people have to be inflated worse than ai wrapper valuations.
387.
by kingstnap 1y ago
DNA damage inevitably accumulates. The big reason children are younger than their parents DNA wise is because the parents' DNA undergo random recombination to create something that is the mixture of the two. This doesn't help over
388.
by kingstnap 1y ago
How can stuff randomly enacted by executive order ever be "permanent". Especially given how legally dubious it is to claim that everything is about fentanyl and therefore this is all somehow security and not economic.
389.
by kingstnap 1y ago
It's all but inevitable that compute gets cheaper. AI winter -> Datacenters have too much compute to sell -> Compute gets cheaper AI summer -> Hardware companies are incentivized to make better chips-> Compute gets cheaper
390.
by kingstnap 1y ago
LLMs have introduced a lot of reading slop into the mix. But they definitely help when you don't feel like typing and can just describe the problem at hand. I'm hoping better sleep helps with the activation energy / anxiety.

More ›