m3kw9

29 ms·

751.
by m3kw9 2y ago
To them they are the apex predators, just like how humans think they are
752.
by m3kw9 2y ago
Why not on commercial aircraft?
753.
by m3kw9 2y ago
Just add 2 to each char.
754.
by m3kw9 2y ago
Like train it like “ umm….. uhhhhh… emmmmmm….. “ right?
755.
by m3kw9 2y ago
Same amount of money now is worth more than in future
756.
by m3kw9 2y ago
It’s 3x cheaper and faster
757.
by m3kw9 2y ago
JavaScript never supported support time zones is news to me
758.
by m3kw9 2y ago
You don’t think we will have AGI or even ASI by then to have significantly improved tech? I would be we will
759.
by m3kw9 2y ago
Sorry to dampen the news but 4o-mini level isn’t really a useful model other than talk to me for fun type of applications.
760.
by m3kw9 2y ago
Snoop dogg smokes it, not that I’m endorsing it
761.
by m3kw9 2y ago
It would have been scary even 2 years ago but by that time AI would solve that, maybe even turn it into some sort of mining operation
762.
by m3kw9 2y ago
So if OpenAI didn't have these outputs for distillation, Deepseek wouldn't exist?
763.
by m3kw9 2y ago
This is more like an experiment more than practical use with the stated 6 toks/s. Paying 6gs for that and days of setup when the next model may come in a month
764.
by m3kw9 2y ago
Better be Mr Beast level to get that
765.
by m3kw9 2y ago
He still right even with deepseek, they are way more than 10m
766.
by m3kw9 2y ago
What does distil qwen 32b mean? It uses qwen for what?
767.
by m3kw9 2y ago
Yes and is easy
768.
by m3kw9 2y ago
He has bias from writing the book regarding black swans which almost always implies market explosions. So there is 2 valuable cents you can take
769.
by m3kw9 2y ago
By that time you can have your personal humanoid robot clean your gutter. You just do nothing but prompt, what a life that’s gonna be.
770.
by m3kw9 2y ago
What people are betting on by selling is there would be a short term pause or deceleration in the investment while they try to squeeze more from their current batch, could be some efficiency catchup
771.
by m3kw9 2y ago
Everything is overvalued if you think in terms of earning multiples, p/s, p/e should be 1:1
772.
by m3kw9 2y ago
Maybe they mistake recreation for the cp command
773.
by m3kw9 2y ago
go read the privacy policy and my thought is that your data is worth a lot to deepseek so they can charge less. They even track your keystroke rhythm.
774.
by m3kw9 2y ago
Was reading the privacy policy of their ios APP, I hate that they collect your keystroke rhythm to biometrically track you.
775.
by m3kw9 2y ago
The chinese gv would be cutting spending on AI according to your logic, but they are doing opposite, and they’d love to get those B200s I bet you
776.
by m3kw9 2y ago
Like the cloud compute we all use right now to serve most of what you use online?
777.
by m3kw9 2y ago
You are looking at one model and also you do realize it isn’t even multimodal, also it shifts training compute to inference compute. They are shifting the paradigm for this architecture for LLMs, but I don’t think this is really new either
778.
by m3kw9 2y ago
The 500b isn’t to retrain a model with same performance as R1, but something better and don’t forget inference. Those servers are not just serving/training LLMs, it training next gen video/voice/niche subject and it’s equival
779.
by m3kw9 2y ago
lol “for those who didn’t realize yet”. Please stop with the hype, we all use it
780.
by m3kw9 2y ago
What this means is that OpenAI can serve even cheaper models in a month using this technique for their updated models

More ›