Bluesky Thread

Dario Amodei:

View original thread
Dario Amodei:

“100 million words context window is already possible, which is roughly what a human hears in a lifetime. Inference support is the only bottleneck to achieve it.

And AI Models actually do learn during the context window, without changing the weights”

x.com/rohanpaul_ai...
x.com
Rohan Paul on X: ""100 million words context window is already possible, which is roughly what a human hears in a lifetime. Inference support is the only bottleneck to achieve it. And AI Models actually do learn during the context window, without changing the weights." ~ Anthropic CEO Dario https://t.co/vGEW7iNGLD" / X
"100 million words context window is already possible, which is roughly what a human hears in a lifetime. Inference support is the only bottleneck to achieve it. And AI Models actually do learn during the context window, without changing the weights." ~ Anthropic CEO Dario https://t.co/vGEW7iNGLD
31 2
So I guess Anthropic’s continual learning take is “moar context”

my left shoulder (😈): “i’m so blackpilled on Anthropic”

my right shoulder (😇): “actually all big LLM advances have been super simple strategies that seemed too dumb to work”
14
they’re willingly missing the continual learning bandwagon. that’s one helluva be

but again, for GPT-3 OpenAI essentially bet the entire company in one big yolo run

en.wikipedia.org/wiki/GPT-3

Anthropic tends to be more focused. It’s their style to ignore complicated continual learning strategies
en.wikipedia.org
GPT-3 - Wikipedia
11
remember when o1-preview was announced?

i spent the first many days shitting on OpenAI because “they’re just doing Chain of Thought”

that was super dumb, lol it really was. Yet here we are
8
31 likes 2 reposts

More like this

×