Nimitz14

5 ms·

1.
by Nimitz14 23d ago
Great question. I think these people are good at spinning narratives _and_ the people above are not doing the work they should be doing to check whether the narrative checks out: I think good leaders (and every manager below) should be doin
2.
by Nimitz14 1mo ago
You have some fundamental misunderstandings on how LLMs work.
3.
by Nimitz14 1mo ago
You’re not getting it. The probabilities do not change at all. The only change is given some probabilities there is a deterministic method for determining which symbol was sampled from that distribution. The distribution or sampling process
4.
by Nimitz14 1mo ago
Uh isn’t google known to have the best tooling in the world
5.
by Nimitz14 2mo ago
Ai slop
6.
by Nimitz14 3mo ago
He said it nudges it to be more deterministic. Your comment is not correct.
7.
by Nimitz14 4mo ago
This is mostly slop. But you may be directionally correct
8.
by Nimitz14 4mo ago
Really really cool. If they can serve this efficiently it would disrupt a lot of things.
9.
by Nimitz14 5mo ago
I think your own math leads to the conclusion the public apis are not serving models of that size. They couldn’t afford to
10.
by Nimitz14 6mo ago
3/4 are actually negative value for most customers
11.
by Nimitz14 6mo ago
Your answer presumes we cannot hold people accountable. I think that is incorrect.
12.
by Nimitz14 7mo ago
Most of these errors will not be meaningful. Real speech is full of ambiguities. 3% is low
13.
by Nimitz14 10mo ago
I think you're bang on the money fwiw. But also worth mentioning that it's OK to ask rather than trying to predict and feeling that having to ask means failure
14.
by Nimitz14 10mo ago
This is clearly a partisan website (see its name). Read with a large grain of salt.
15.
by Nimitz14 10mo ago
Yann was in charge of FAIR which has nothing to do with llama4 or the product focussed AI orgs. In general your comment is filled with misrepresentations. Sad.
16.
by Nimitz14 11mo ago
Sora is about data gathering. This take is too simplistic.
17.
by Nimitz14 11mo ago
There's still too much fat, or rather people who just add net negative value. My org would benefit from firing easily 50%.
18.
by Nimitz14 1y ago
I'm surprised at the model dim being 2.8k with an output size of 200k. My gut feeling had told me you don't want too large of a gap between the two, seems I was wrong.
19.
by Nimitz14 1y ago
This is normal. Rope was introduced after bert/gpt2
20.
by Nimitz14 1y ago
What is it like living a life completely ignorant of reality? If you'd like to learn, feel free to ask chatgpt on the leftish pushback against abundance. Or historical examples of leftists blocking housing projects. Or environmentalist
21.
by Nimitz14 1y ago
We're all gonna be PMs.
22.
by Nimitz14 1y ago
Think the microsoft terminal saga shows a pretty clear track record.
23.
by Nimitz14 1y ago
This is very good. I've also noticed I do the polite engagement sometimes and how its roboticsness is obvious to any person with not low EQ. Some people seem to just be naturally happy in almost any settings and its a very valuable tra
24.
by Nimitz14 1y ago
[flagged]
25.
by Nimitz14 1y ago
So this is not the case for https://gemini.google.com ?
26.
by Nimitz14 1y ago
If your friends are the sort that stick to tribes instead of thinking independently get smarter friends.
27.
by Nimitz14 1y ago
Not the same thing at all. Corporations care about their data a lot and would cancel deals over this. Noone cares if some authors get upset, they have no leverage. Disappointing how people will make confident statements while being so clear
28.
by Nimitz14 2y ago
Any continual learning papers you're a fan of?
29.
by Nimitz14 2y ago
If you had spent the time writing this and other comments googling/chatgpting you could have learned why returning to the gold standard is a joke.
30.
by Nimitz14 2y ago
You're looking at the start and the end of the plots and totally ignoring what happens in between. If Austin had followed the linear trendline from Q4 2019 to Q2 2022 (increase of 22.7/Q) then it would have been at 800 in Q4 2024

More ›