AI & Tech
AI Researcher Claims Modern Go Bots Match AlphaGo for $3K Using LLM Coding Assistance
Dwarkesh Patel Podcast
Eric Jang – Building AlphaGo from scratch
"What took a whole team of research scientists at DeepMind and millions of dollars of research and compute can now be done for a few thousand dollars of rented compute. I spent maybe the first $4K doing exploratory research and then about $3K on the final run."
Eric Zhang, former VP of AI at 1X Technologies and ex-Google DeepMind senior research scientist, replicated AlphaGo-level performance on a budget of approximately $7,000 using modern compute and LLM-assisted coding. This represents a dramatic reduction from the estimated millions DeepMind spent, suggesting massive efficiency gains through better hardware, simplified architectures, and AI-assisted research. Zhang argues many of DeepMind's original algorithmic tricks are now unnecessary with modern GPUs and proper initialization against existing strong bots like Katago.
From this episode
Dwarkesh Patel Podcast