p1esk

29 ms·

511.
by p1esk 2y ago
Why not just use a regular laptop for this? There’s a ton of low level sound processing libraries for every OS.
512.
by p1esk 2y ago
Has that stolen info been published anywhere?
513.
by p1esk 2y ago
What makes it better than simply watching it on a TV?
514.
by p1esk 2y ago
Can’t access the article
515.
by p1esk 2y ago
I don’t think anyone doubts that. The question is when exactly do we get to human level intelligence on all tasks (agi)? Is it going to be GPT5? GPT6? Or has the performance improvements saturated already? It makes a huge difference in term
516.
by p1esk 2y ago
Who said gpt2 was agi?
517.
Beating NumPy matrix multiplication in 150 lines of C(salykova.github.io)
392 points by p1esk 2y ago|81 comments
518.
by p1esk 2y ago
Sounds like regular lora adapters
519.
by p1esk 2y ago
One of the worries he’s repeatedly expressed to Moore is that AI and software generally are pulling talent away from semiconductor engineering. Perhaps he should try paying them accordingly.
520.
by p1esk 2y ago
Using CPU memory to extend GPU memory seems like a more straightforward approach. Does this method provide any benefits over it?
521.
by p1esk 2y ago
I’m not sure I understand your problem – are you doing a good job? If yes, what’s the problem? If no, why not?
522.
by p1esk 2y ago
I feel like my company sees me not as a human but as a set of metrics to be judged and compared against. How did you manage to graduate from college?
523.
by p1esk 2y ago
They use Brainfuck language for the experiments. Seriously.
524.
by p1esk 2y ago
The problem is it is different 128 in each major, and the overlap across all four majors in a year is significantly less than 100. People move up and down the rankings a lot. To actually make a living you need to be consistently in the top
525.
by p1esk 2y ago
For MNIST your model is sufficient. But it’s not structurally complex enough to generate more complex images, even if you scale it up.
526.
by p1esk 2y ago
It’s similar to wanting to start your own company: the odds to succeed are grim.
527.
by p1esk 2y ago
The difference is: consultants are expected to get paid after they complete the job, tennis players lose a lot more matches on average than they win so they usually do not get paid.
528.
by p1esk 2y ago
In pro tennis only the top 100 players actually make more money than they spend.
529.
by p1esk 2y ago
It's just a word meaning AI gives wrong answer. No, it’s more specific than just wrong. Hallucination is when a model creates a bit of fictitious knowledge, and uses that knowledge to answer a question.
530.
by p1esk 2y ago
In my experience Sonnet 3.5 is about the same as 4o for coding. Sometimes one provides a better solution, sometimes the other. Both are pretty good.
531.
by p1esk 2y ago
Reddit has been the most trustworthy source for me in the last ~5 years, especially when I want to buy something.
532.
by p1esk 2y ago
Do you need someone to remind you the rules of a polite conversation? This is not Reddit.
533.
by p1esk 2y ago
“ We have not made any decisions and any discussion of this is premature,” said one administration official. “The president is absolutely firm that he will not be sending US troops to Ukraine.”
534.
by p1esk 2y ago
FYI, candle light probably makes your air worse.
535.
by p1esk 2y ago
I just want to avoid putting one black box on top of another if possible.
536.
by p1esk 2y ago
What does this have to do with hallucinations?
537.
by p1esk 2y ago
I didn’t check, but iirc it was an fp16 model checkpoint which we converted to int8 for inference, so I assume 140GB?
538.
by p1esk 2y ago
Can I try your question? Just curious.
539.
by p1esk 2y ago
Depends on your CPU. I once tried 70b llama on 256 thread Epyc, it was around 1/10 of A100 (80GB) speed.
540.
by p1esk 2y ago
OK, so you have a way to evaluate the accuracy and convince yourself that it’s probably works as expected. But what about me, a user? How can I check that the question I asked was answered correctly?

More ›