Young U.S. Workers' Career Optimism Plummets, Seniors Improve
I'm usually skeptical of 'generational' trends, but young Americans do indeed seem unusually pessimistic about their careers right now 🧵 A long running survey finds satisfaction for 18-24 year olds has fallen since 2023, despite rising for older workers. https://t.co/CBcrAUi9Bb
Just a Few Percent of Tasks Sustain High Wages Post‑AGI
You only need a couple of percent of tasks to be things we want humans for to result in high wages post-AGI.
Long‑term AI Agents Revive Classic Alignment Challenges
With chatbots, AI alignment looked easier than expected. But with the shift to ever smarter longer-horizon agents, the classic reasons for concern come back. New primer: four reasons why AI won't do what we want 🧵 https://t.co/xTFjPXWnK8
Apollo Was Vanity; Real Progress Means Automating Economies
The Apollo Program was basically a vanity project, dropped a few years after hitting its goal. Totally different beast to automating your entire economy.
Debating Machine Consciousness: Confident Denials Meet Ongoing Philosophy
Amazed at how many people will confidently assert machines will never be conscious. It's maybe the most contested question in all philosophy.
Most Philosophers Think AI Could Be Conscious, 11% Remain Agnostic
On AI consciousness: 1. Functionalism is the most popular view of philosophy of mind, which basically says sufficiently complex machines *will* be conscious. 2. Most other views are also compatible with AI consciousness (e.g. identity theory, panpsychism). 3. Eliminativists say humans...
Limited Chip Exports Preserve US Edge, Fuel China’s Future AI
Jensen's arguments against chip export controls didn't make a ton of sense to me, but here's a steelman: selling a limited quantity of n-1 generation chips to China could undercut their domestic industry, while retaining a 10x American compute advantage. Totally...
Plan for AI: Fast Takeoff, Slow Rise, or Plateau
3 scenarios for AI: 1) Automated coder in 2028, 1yr intelligence explosion to superintelligence in 2029 2) Automated coder in 2030, slower takeoff to ASI ~2036 3) AI plateaus around 2030. AI agents everywhere but don't reach true AGI for decades. What should you...
Only Two Years
You may have two years left to prepare for the intelligence explosion. How are you going to spend them?
Current Trends Point to AI Explosion by 2030
Everything's unfolding exactly as you'd expect if there will be an intelligence explosion around 2028-2030.
Claude Opus 4.5 Codes Hour for $0.65 vs $21
In the METR benchmark: Cost for Claude Opus 4.5 to do 1h coding task: $0.65 Cost to pay humans to do the same: $21 AI is extremely cheap.
Top AI Forecasters Assign 10% Chance of Next-Year R&D Automation
In Feb, I was shocked to find some of the best AI forecasters think there's a 10% chance AI R&D gets automated *the next year*.