Emad Mostaque
speaker
1,351 appearances
1 recordings
1 series
first heard Dec 2024
last heard Dec 2024
Emad Mostaque’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
You just need one entity to do it.
Like
M meta doing llama is good enough for almost most things, right?
And they did twenty million odd H one hundred hours for the wild dispersion of lava.
And that's what let's say forty million dollars spent for Lava seventy B three.
And then everyone could just take that and fine tune it, but to tune it requires a hundred and twenty eight to two hundred and fifty six eight one hundreds.
And the order of magnitude costs is a hundred thousand, two hundred thousand dollars to create an Arabic version of that.
And so again, we're blinded by the massive compute and the edge stuff when actually it doesn't require that much to do the tuning.
The thousands enables you to actually give access to your people to the technology, especially if, again, you don't require rate limits or you're satisficing a level.
And it also allows you to
Do the tuning.
So you probably start out with a few hundred and then you build up from there.
And again, when you look at the disparity of compute and energy and other things, it's very interesting.
Like I was in South Africa, I think the fastest cluster there was like fifty or a hundred H one hundreds.
But then the clouds are building even larger ones.
This is why it's very achievable if we're coordinated.
But then do you want to be giving your people that level of healthcare AI or that level of healthcare AI?
Do you want to have your companies in your country be able to call on these experts or even more experts?
So this is why I think there'll be a very interesting capital stock question.
And when we look at things like 5G.
Showing 761–780 of 1,351 · page 39 of 68
← Previous
Next →