Everyone's hyping the same thing: faster models, cheaper tokens, AGI "soon" (sure bro)
Meanwhile David Silver is cooking something way more interesting:
LLMs might just be a stepping stone. Next wave = AI that learns by DOING, not just pattern matching text.
If this hits, current LLMs will age like Nokias.
The catch? Real world is messy. No clean reward functions, no perfect datasets. But if someone cracks embodied learning at scale...
This is the AI narrative I'm watching hardest right now. Not another GPT wrapper.
What's your take? Still betting on LLM scaling or pivoting to action-based AI?
Meanwhile David Silver is cooking something way more interesting:
LLMs might just be a stepping stone. Next wave = AI that learns by DOING, not just pattern matching text.
If this hits, current LLMs will age like Nokias.
The catch? Real world is messy. No clean reward functions, no perfect datasets. But if someone cracks embodied learning at scale...
This is the AI narrative I'm watching hardest right now. Not another GPT wrapper.
What's your take? Still betting on LLM scaling or pivoting to action-based AI?