airgapstopgap

14 ms·

31.
by airgapstopgap 3y ago
Not really. They have a way of squaring this circle, by changing their inference code. Speculative sampling [1] would still make their first claim a lie – sure, there'd still be the original GPT-4 model, plus a smaller draft worker.
32.
by airgapstopgap 3y ago
This is not responsive to my arguments. Google can be arbitrarily far behind OpenAI or Anthropic, OP's idea that they feel threatened by LLaMA when they (well, Deepmind) have reached LLaMA level 18-10 months ago is still wrong.
33.
by airgapstopgap 3y ago
> But there's no little numeric model embedded inside the language model that would tell it how many dollars, or be able to compare dollars against dollars This logic applies to any function an LLM may perform, therefore it cannot p
34.
by airgapstopgap 3y ago
Do you realize that LLaMA-1 is just a very slightly smaller, comparably performing replication of Chinchilla [1], which DeepMind had completed a year prior to LLaMA's release? And has RLHF-ed into a suitable chatbot "Sparrow"
35.
by airgapstopgap 3y ago
I believe it's more like they used Anthropic human preference data [1] or similar, and accordingly Anthropic/progressive American notion of honest-helpful-harmless behavior. Thus I've seen models misgeneralize towards prudish
36.
by airgapstopgap 3y ago
Note that this is apparently a 7B version of a 104B model trained with the intention of competing with OpenAI offerings on the Chinese market [1]. There is a number of those projects: Baichuan, ChatGLM2, InternLM and some more iirc, and the
37.
by airgapstopgap 3y ago
I'd argue that this tedious posture of caring about the environment is vastly more responsible for the climate disaster; it is very popular with the anti-nuclear crowd, after all. > the kind of moral arrogance from humankind we need
38.
by airgapstopgap 3y ago
This is quite an amusing way to describe it. Clearly motivated, too. > debris may destroy the environment around it "Destroy the environment" invokes some rather grandiose associations for what is essentially pulverizing a stee
39.
by airgapstopgap 3y ago
I cite Western sources solely because I presume most people here don't read Russian. I had to look it up. My personal hypothesis about the dam is that it was negligence and perhaps a misguided attempt to "partially demolish it&quo
40.
by airgapstopgap 3y ago
Are you perhaps confusing Venezuela with El Salvador?
41.
by airgapstopgap 3y ago
> Third observation Why would people not support Wagner over the regular army, and specifically over Shoigu and Gerasimov that he alleges to target? I can explain. None of the reasoning below is meant to excuse or praise Wagner or the re
42.
by airgapstopgap 3y ago
The vast majority of human data can be refined into something approximating high quality textbooks, which is what happened here too.
43.
by airgapstopgap 3y ago
Bloom and OPT are weak, predictably so. Nobody uses them (except for research sometimes, and even then rarely). It doesn't make sense to look for hardware to run a 176B model when a 13B outperforms it across the board.
44.
by airgapstopgap 3y ago
Not OpenAI. But AI in principle? Perhaps. This would have to be solved by actual nonprofits, though.
45.
by airgapstopgap 3y ago
> Nothing in your reply now, nor in any of your other comments, supports that claim Singer's insistence that a utilitarian doctor is morally bound to kill a patient to save others is sufficient. > Your claim does not follow from
46.
by airgapstopgap 3y ago
> 15.6”(39.62cm) FHD (1920x1080) 16:9 panel, 3.2mm I've been disgusted with this stuff for well over a decade. I remember how excited I was when Microsoft showed they can see and made a glitchy 2-in-1 with bright, high-resolution
47.
by airgapstopgap 3y ago
> That's a misreading of the paper and a misrepresentation of the position that Singer holds It's not. However, utilitarians are inevitably compelled to argue that it is, because their efficacy depends on it. This amounts to ga
48.
by airgapstopgap 3y ago
This is comically naive and literal. Have you read any moral philosophy before? The point of his argument is not that under certain exceptional conditions surgeons should kill people to harvest their organs to save more lives, but precisely
49.
by airgapstopgap 3y ago
The most important of Singer's writings, in my opinion, is "Secrecy in Consequentialism"[0], mentioned in this podcast too, though not the crucial part, which I will present without further comment except that this paper shou
50.
by airgapstopgap 3y ago
There definitely is – when there is. See the new paper on what exactly Transformer reasoning entails. https://twitter.com/bohang_zhang/status/1664695084875501579
51.
by airgapstopgap 3y ago
Always has been, it's only blocked in EU and a few more countries.
52.
by airgapstopgap 3y ago
Frankly I think anything crypto-related is very unpleasant to discuss on HN (by the way why is this site called HN? Why not just "American tech news"?) Any sign of damage done to crypto industry is met with fierce jubilation you c
53.
by airgapstopgap 3y ago
LLMs are the apex of NLP research to date, and NLP is obviously a branch of AI. You may have some sophisticated notion of AI, AGI, human-level or human-like AI or whatever, but NLP has been considered AI for generations now.
54.
by airgapstopgap 3y ago
No longer trying, I suppose.
55.
by airgapstopgap 3y ago
Interesting release from Apple but there's been a similar (and I think more promising architecturally) paper from Meta lately: > MEGABYTE: Predicting Million-byte Sequences with Multiscale Transformers https://arxiv.org&#
56.
by airgapstopgap 3y ago
> note that it is not much higher 1,5 (or 1,2) vs 2 is a lot. It is close to 3 so confidently above replacement for the whole secular+non-Haredim religious group (reminder that nominally the US is nearly completely "religious")
57.
by airgapstopgap 3y ago
What motivates people to provide sources that refute their own argument? The Bedouin (Muslim), Google says, make up 200,000-250,000 out of Israel's 9,3 million people. Or as your source puts it: > Of Israel’s population of 9.5m, Isr
58.
by airgapstopgap 3y ago
Well Russian charts do show it, because since 2015 Russia includes Crimea in internal stats. As you can see it's also growing in 2022, despite there being consistently more deaths than births for years on end. I suppose we can infer
59.
by airgapstopgap 3y ago
Russian population decline has been compensated for with a great number of immigrants from Central Asia (plus occupied Crimea; that's not recognized internationally but is obviously a factor for their economy). It'd probably be mo
60.
by airgapstopgap 3y ago
Those economic explanations are searching for keys under the lamp post. As Tanner Greer notes[1]: > Structurally the arguments in both countries go like “life is so hard, and things generally so depressing, that I have no desire to bring

More ›