Keep pulling the thread on Epoch AI Researchers.
The speaker predicts that an AI could solve a major unsolved math problem, such as the Riemann Hypothesis, within the next five years.
Anthropic is considered the most likely candidate to operate the first gigawatt-scale AI data center.
The "Project Rainier" data center, a joint project between Anthropic and Amazon in New Carlisle, is on track to come online in January.
Microsoft is building a large-scale data center in Mount Pleasant, called "Microsoft Fairwater," which will be used at least in part by OpenAI.
The Microsoft Fairwater data center is projected to consume an amount of power equivalent to more than half of New York City's usage.
Dario Amodei of Anthropic predicted in March 2024 that AI would write 90% of all code within six months of that date.
Dario Amodei of Anthropic predicted that AI systems equivalent to a "country of geniuses" could exist in a data center as soon as 2026 or 2027.
The joint Anthropic-Amazon data center project in Indiana is expected to consume nearly as much power as the state's capital city, Indianapolis.
Extrapolating NVIDIA's current revenue growth suggests the economic value generated by AI could lead to an increase of approximately 1% of GDP by 2030.
The speaker believes Anthropic's bullish AI timelines are based on the theory that automating R&D will lead to a rapid, self-improving "takeoff" in AI capabilities.
The speaker has successfully used ChatGPT Agent for data center research to navigate county-level databases and find specific documents like air pollution permits and tax abatement filings.
Large-scale AI data centers like xAI's Colossus 2 are being constructed on timelines of two years or less.