ipsum2

31 ms·

361.
by ipsum2 2y ago
Probably not. Cerebras chip only has 16bit and 32bit operators.
362.
by ipsum2 2y ago
You can't expect a 1B model to perform as well as 7B or chatGPT, probably the best use case is speculative decoding or to use to fine tune for a specific use case.
363.
by ipsum2 2y ago
Those are different approaches afaict.
364.
by ipsum2 2y ago
It's meant as both humorous and a nerd snipe :)
365.
by ipsum2 2y ago
`equal(0.3, 0.2 + 0.1); // true` how is this wizardry possible?
366.
by ipsum2 2y ago
> Provide so much syntactic sugar as to cause any C developer a diabetes-induced heart attack. seems like its obvious to me that its a joke
367.
by ipsum2 2y ago
Yes, many non-technical people thought they were autonomous.
368.
by ipsum2 2y ago
They already are, and aren't very good.
369.
by ipsum2 2y ago
They are advertising a ridiculously low price for their GPUs that can't be rented. If a store advertised $0.50 burgers, but when you visit, they say they're not for sale, wouldn't you consider that a scam?
370.
by ipsum2 2y ago
I signed up and don't have access currently. My point that the prices are low because demand is limited because of lack of users still stands. Once people sign up and hear about it, the price will increase substantially.
371.
by ipsum2 2y ago
The average consumer cannot. Only those who have access to sfcompute's private beta can access those prices. Once it opens up to the public, the price will increase.
372.
by ipsum2 2y ago
Okay, but you can't actually buy it at that price, it's a pure marketing ploy.
373.
by ipsum2 2y ago
sfcompute is a scam. You can't buy GPUs at that price. They're running a "private beta" where people can bid for a spot GPU, but they let a limited number of people into the beta, so the prices are artificially low.
374.
by ipsum2 2y ago
Only if they're networked with Infiniband.
375.
by ipsum2 2y ago
It's definitely teleoperated. The hardware is there, but the software isn't.
376.
by ipsum2 2y ago
Having built teleoperated robots, I can guarantee that the non-dancing ones are teleoperated.
377.
by ipsum2 2y ago
Title is such clickbait, it does not rewrite the rules of 3d vision, it is a marginal improvement on existing models, and does not work for video, only images. However, Apple open sourced the model weights, which is amazing for research.
378.
by ipsum2 2y ago
Definition of fraud, from Google: * wrongful or criminal deception intended to result in financial or personal gain. * a person or thing intended to deceive others, typically by unjustifiably claiming or being credited with accomplishments
379.
by ipsum2 2y ago
https://x.com/shinboson/status/1832933747529834747 is the best summary. if you don't have a twitter account: https://threadreaderapp.com/thread/1832933768031588622.html
380.
by ipsum2 2y ago
No. He served the "weights" (actually Claude) for over 24 hours. It's practically impossible to have served the "correct weights" and just have lost them.
381.
by ipsum2 2y ago
No. CoT has been around for several years (Jan 2022) https://arxiv.org/abs/2201.11903 . And so has Reflection (March 2023) https://arxiv.org/abs/2303.11366 . This approach that was taken by Reflecti
382.
by ipsum2 2y ago
Sahil Chaudhary of GlaiveAI perpetuated fraud, where he replaced the model that he "trained" with other backend ML providers. He still has not given a reason why "Claude" the string would be missing, just magically happe
383.
by ipsum2 2y ago
"Microsoft has spare cloud capacity" definitely not, if they're actively building out new datacenters to keep up with demand.
384.
by ipsum2 2y ago
I, and several others, had no problem running on PyTorch on AMD GPUs, with no code changes from CUDA. Check out MosaicML's blog posts: https://www.databricks.com/blog/training-llms-scale-amd-mi25...
385.
by ipsum2 2y ago
Big tech mergers seem to be halted with the current SEC chair. Maybe next year if the administration changes.
386.
by ipsum2 2y ago
The camera is manual focused, so 1/8mm would make it out of focus.
387.
by ipsum2 2y ago
Previous discussion (2022): https://news.ycombinator.com/item?id=33908576
388.
by ipsum2 2y ago
You worked at FB, but you call yourself an SRE, not a PE? ;)
389.
by ipsum2 2y ago
You're wrong here. Meta has released state of the art open source ML models prior to ChatGPT. I know a few successful startups (now valued at >$1b) that were built on top of Detectron2, a best-in-class image segmentation model.
390.
by ipsum2 2y ago
The first thing the demo told me was that it was in a dark and scary forest.

More ›