trott

9 ms·

61.
by trott 2y ago
> Just because you explain what you want to a human that doesn't mean they agree or will comply. Humans have innate desires that may conflict with the desires of other humans. A language model just looks for ways to continue texts.
62.
by trott 2y ago
A lot of people are worried about aligning superintelligent, self-improving AI. But I think it will be easier than aligning current AI, for the same reason that it's easier to explain what you want to a human than it is to train a dog.
63.
EU regulator rejects Alzheimer's drug lecanemab(bbc.com)
2 points by trott 2y ago|0 comments
64.
Rust can Compete with Python [video](youtube.com)
1 points by trott 2y ago|0 comments
65.
Spillover of highly pathogenic avian influenza H5N1 virus to dairy cattle(nature.com)
6 points by trott 2y ago|0 comments
66.
Nope–It's Never Aliens(scientificamerican.com)
10 points by trott 2y ago|1 comments
67.
Company wins funding to make medicine in space(bbc.com)
2 points by trott 2y ago|0 comments
68.
You are probably sitting down for too long(bbc.com)
60 points by trott 2y ago|44 comments
69.
The $100B plan with "70% risk of killing us all" w Stephen Fry [video](youtube.com)
21 points by trott 2y ago|4 comments
70.
Recursion in AI is scary. But let's talk solutions.(olegtrott.substack.com)
1 points by trott 2y ago|0 comments
71.
by trott 2y ago
The twilight zone is 20-35%: https://pubmed.ncbi.nlm.nih.gov/10195279/ (Incidentally, the author was on my thesis committee, but this isn't precisely my field of expertise.)
72.
by trott 2y ago
> matching less than 60% of the sequence of the most closely related fluorescent protein > When the researchers made around 100 of the resulting designs, several were as bright as natural GFPs, which are still vastly dimmer than lab-e
73.
by trott 2y ago
This shows how important the right title is in your news article or blog post. That post that got 9 points and 1 comment should have been called "Hatetris has been SOLVED (infinite score)" instead of "Losing the World Reco
74.
Are AlphaFold's new results a miracle?(olegtrott.substack.com)
38 points by trott 2y ago|11 comments
75.
by trott 2y ago
> Nearly all of it. Maybe you misunderstood me. I'm not talking about learning to understand spoken English. You don't need hearing or vision at all to grow up to be intelligent (and able to write English).
76.
by trott 2y ago
> Lots of assumption here. First, that we will only be training on text data, if we take into considerations all the videos and audios shared I am quite sure we would have one or two orders of magnitude more of data. 1GB of text is way m
77.
by trott 2y ago
> Uncompressed stereo is 100 kilobytes per second. How much of that is cognitively useful for learning English? On top of the textual content, audio gives you emphasis and mood. Not a lot of information in that -- a few bits per sentence
78.
by trott 2y ago
> The human brain isn't randomly initialized. It's undergone 500m years of pretraining. All of the information accumulated by evolution gets passed through DNA. For humans, that's well under 1GB. Probably a very tiny fract
79.
by trott 2y ago
The stream of data from vision does NOT explain why humans learn 1000x faster: Children who lost their sight early on, can grow up to be intelligent. They can learn English, for example. They don't need to hear 200B words, like GPT-3.
80.
by trott 2y ago
If you take Llama-3-400B, and 30x its data (hitting the data ceiling, AFAICT), 30x its size to match, and the hardware improves by, say, 3x, then you'll use up about a year's worth of energy from a typical nuclear power plant.
81.
Will We Run Out of Data? Limits of LLM Scaling Based on Human-Generated Data(epochai.org)
56 points by trott 2y ago|58 comments
82.
by trott 2y ago
François Chollet says LLMs do not learn in-context. But Geoff Hinton says LLMs' few-shot learning compares quite favorably with people! https://www.youtube.com/watch?v=QWWgr2rN45o&t=46m20s The truth is in the middl
83.
AI Agents: too early, too expensive, too unreliable(old.reddit.com)
19 points by trott 2y ago|2 comments
84.
by trott 2y ago
There's also https://en.wikipedia.org/wiki/Metamath Its smallest proof-checker is only 500 lines.
85.
by trott 2y ago
> Computer-Checked proofs are one area where ai could be really useful fairly soon. Though it may be closer to neural networks in chess engines than full llm's. Yes, if significant progress is made in sample-efficiency. The current
86.
AI's learning ineffectiveness and how to fix it(olegtrott.com)
1 points by trott 2y ago|0 comments
87.
by trott 3y ago
Very good article. I wonder why it disappeared from HN? More comments than up-votes?
88.
by trott 3y ago
> Scientists are not more rational people than $anyone_else. Well, scientists disagree with you [1] [1] https://www.scientificamerican.com/article/scientists-think-...
89.
Are You Smarter Than An LLM? (Quiz based on the most popular LLM benchmark)(d.erenrich.net)
28 points by trott 3y ago|9 comments
90.
Training LLMs using 8-bit numbers(arxiv.org)
1 points by trott 3y ago|0 comments

More ›