Scaling AI, Future Possibilities
Dwarkesh and Dario discuss the potential outcomes if scaling plateaus before achieving human-level intelligence in AI. They explore the possibility of running out of data, using up all available compute, or not having the right architecture. Dario expresses his belief that the current scaling laws are unlikely to stop, while acknowledging the challenges of representing certain abilities without the attention capabilities of transformers.In this clip
From this podcast
Dwarkesh Podcast
Dario Amodei (Anthropic CEO) - $10 Billion Models, OpenAI, Scaling, & AGI in 2 years
Related Questions
Will large language models scale all the way to artificial general intelligence (AGI) as discussed in the episode Dario Amodei (Anthropic CEO) - $10 Billion Models, OpenAI, Scaling, & AGI in 2 years and the clip Grasping Language, Scaling AGI?
Will large language models scale all the way to artificial general intelligence (AGI) as discussed in the episode Demis Hassabis - Scaling, Superhuman AIs, AlphaZero atop LLMs, Rogue Nations Threat and the clip Scaling Challenges?
Will large language models scale all the way to artificial general intelligence (AGI) as discussed in the episode Demis Hassabis - Scaling, Superhuman AIs, AlphaZero atop LLMs, Rogue Nations Threat and the clip Scaling Hypothesis Insights?