oersted

38 ms·

601.
by oersted 3y ago
> Meta already sells the Quest systems basically at production cost, and just ignores the development costs It is a bit ambiguous, but it sounds like they sell the Quest at hardware cost and take the loss in all software costs and R&
602.
by oersted 3y ago
Both Apple and Nvidia use TSMC for chip manufacturing. Actually Apple might do a lot more of the hardware themselves, and more varied. Nvidia is just a design house, chips at TSMC and the surrounding electronics by many partners like Asus o
603.
by oersted 3y ago
Thanks for the Space Shuttle factoid, quite relevant to the OP article https://www.astrodigital.org/space/stshorse.html
604.
by oersted 3y ago
They actually have a lot of difficulty in hiring good directors for this exact reason, they have widened the gap several times and are planning to keep widening it.
605.
by oersted 3y ago
I don’t understand this, we had 5h of classes a day in CS for 4 years in Spain, it’s still like that, and it was not particularly strenuous. There was plenty of time left to start my first company and a reasonable amount of leisure time, wh
606.
by oersted 3y ago
Agreed, but it's wild that even one benchmark shows this. Based on what we knew just a few months ago, these models should be so far from each other in every benchmark.
607.
by oersted 3y ago
It's easy to miss: select English in the dropdown. The scores are quite different in Overall and in English for LMSYS. As I've stated in other comments, yeah... Agreed, I'm stretching it a bit. It's just that any indicat
608.
by oersted 3y ago
If anything this is good for them. Apple's play here has always been getting their devices ready for running LLMs locally. This makes it way easier.
609.
by oersted 3y ago
In the English category: GPT-4-0314 (ELO 1166), Llama 3 8B Instruct (ELO 1161), Mistral-Large-2402 (ELO 1151), GPT-4-0613 (ELO 1148). You are right, I toned down the language, I got a bit overexcited, and I missed the difference in the vers
610.
by oersted 3y ago
Right thanks for the reminder, I added it
611.
by oersted 3y ago
Incredible, rivals Llama 3 8B with 3.8B parameters after less than a week of release. And on LMSYS English, Llama 3 8B is on par with GPT-4 (not GPT-4-Turbo), as well as Mistral-Large. Source: https://chat.lmsys.org/?leaderb
612.
by oersted 3y ago
Well I have an internal monologue and I think in language a lot, but I wouldn't say the sentence is on my mind before I write it. It's more like I redirect the stream of language to the page, so instead of saying the words to myse
613.
by oersted 3y ago
Agreed, that's an extreme but reasonably plausible scenario. In general, I don't think there's much you can do on this earth that could resist such a coalition. I guess some encryption could hold even if the whole world agree
614.
by oersted 3y ago
You can't "double spend" without anyone noticing, even a minimal amount is in theory detectable by the exchange and the word would spread like fire. Perhaps the exchange would let it go if the amount is small enough, so that
615.
by oersted 3y ago
Off the top of my head: it seems unlikely that you can recoup the enormous investment required to get to 51% by making fraudulent transactions without anyone noticing and crashing the value of bitcoin, or agreeing to switch to a clean fork.
616.
by oersted 3y ago
His "comic slides" are fun :) https://eurocrypt.iacr.org/2017/slides/A04-constraint-hiding...
617.
by oersted 3y ago
Not really, right now it's within +14/-15 ELO with 95% probability, which barely changes the ranking. It's well ahead of Gemini Pro and squarely within the top 5 with Claude 3 Opus and a few versions of GPT-4-Turbo. More impr
618.
by oersted 3y ago
The ranking has changed significantly after 6 hours: Llama 3 70B Instruct (ELO 1198) is now behind even Claude 3 Sonnet (ELO 1202) and well behind the latest GPT-4-Turbo (ELO 1259), but still above the last GPT-4 (ELO 1189). EDIT: My bad, I
619.
by oersted 3y ago
Where did you find this number? Not doubting it, just want to get a better idea of how precise the estimate may be.
620.
by oersted 3y ago
I see it more as an indirect signal for how good Llama 3 8B can get after proper fine-tuning by the community.
621.
by oersted 3y ago
It's likely true that they didn't change the model, same for the many claims of GPT-4 getting worse. But they do keep iterating a lot on the "safety" layers on top: classifiers to detect dangerous requests, the main syst
622.
by oersted 3y ago
For the instruction tuned models, Llama 3 8B is even significantly better than Llama 2 70B!
623.
by oersted 3y ago
Better than Claude 3 Sonnet but Claude 3 Opus is significantly more powerful, albeit I'm not sure how they compare accounting for parameter size.
624.
by oersted 3y ago
Look at the electron microscope images! I expected it would roughly look like a fractal, but it is so clean, it looks artificial. https://www.nature.com/articles/s41586-024-07287-2/figures/1
625.
by oersted 3y ago
> This means the sensor can tell you where you bent it, with a predefined (and coarse) resolution. > OptiGap’s application is mainly within the realm of soft robotics, which typically involves compliant (or ‘squishy’) systems, where t
626.
by oersted 3y ago
LinkedIn mostly, you can use the free version, but it's easier if you use Recruiter for much more precise filtering, and/or set up a job post. You could try looking for related projects in GitHub if that makes sense for your proje
627.
Show HN: MirrorThink AI for Professional Science and Engineering(mirrorthink.ai)
1 points by oersted 3y ago|1 comments
628.
by oersted 3y ago
The video embedded in the article is a great and more brief explanation of what's written: https://www.youtube.com/watch?v=LgPDOSou1tw
629.
by oersted 3y ago
Looking at the leaderboard shows a clearer picture: https://tatsu-lab.github.io/alpaca_eval/ - GPT-4-Turbo: 50.00% - Snorkel (current 2nd, Mistral 7B fine-tune): 34.86% - Yi 34B Chat (current 6th): 29.66% - GPT-4: 23.5
630.
by oersted 3y ago
Thank you for teaching me a new phrase :)

More ›