Evan Kervella
speaker
257 appearances
3 recordings
1 series
first heard Jul 2026
last heard 9 Sep
Evan Kervella’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 3 in all, peaking in Sep 2026 with 1.
Appearances
In fact, I'm not really not because the LLM are the architecture the most optimal world.
But force to constatate that it was a little bit more less than what you would imagine.
And the demonstration mathematics are really one of the things.
What you
l'intelligence can say that it's intelligence, because it was extremely bad to reformular.
Reformular shows a little like if I had a compromise that was the surface, alors que that.
But I think that it would be modellizer a distribution and that distribution.
I constate that approach to the project token and post-training with your enforcement learning for that at part of that capacity of intelligence line.
arrive to demand in the world and generate data original and generation intelligent, but in fact it's a little bit more less than what I think.
The model is not too much, but it's not lower.
say.
It's super positive.
But in fact, as I said, the model is when we schematically, we have the infrastructure loose, the power, the data center, and all that, the compute, there's the models and then the application layers.
But the models are commodities, as Valentine.
I'm not done.
They have distilled.
And in fact, there are two uses: front, it's Anthropic and Chat GPT, which are at a moment the intelligent.
You
not for instance, but in six months, in two months, we're not distilling Astra.
So okay, there are models that have their proposition of value, and that's what the Chinook into three and six months at the front.
Showing 61–80 of 257 · page 4 of 13
← Previous
Next →