(1/5) GenAI reached half of US adults in 3 years – the fastest adoption ever measured. Faster than the PC and the internet.
True stat; wrong story.
The real story is a cascade: reach ~55% → habit ~15% → value ~5%
(1/3) Everyone's reading this chart as a bet on when AGI arrives – take your pick. It isn't a disagreement about when. It's a disagreement about what "intelligence" even is. Benchmarks, evals, reasoning traces or something more broader?
@GaryMarcus Raw intelligence was necessary but not sufficient. Tools, accumulated & shared context and social fabric also critical and the "loop" that it created. We are still focused on which model, what benchmark!
https://t.co/VM8zvHjbku
(1/3) Everyone's reading this chart as a bet on when AGI arrives – take your pick. It isn't a disagreement about when. It's a disagreement about what "intelligence" even is. Benchmarks, evals, reasoning traces or something more broader?
@VraserX Whether we’re in the singularity or not depends on how you define it. Raw intelligence is necessary but not sufficient
https://t.co/VM8zvHjbku
(1/3) Everyone's reading this chart as a bet on when AGI arrives – take your pick. It isn't a disagreement about when. It's a disagreement about what "intelligence" even is. Benchmarks, evals, reasoning traces or something more broader?
(1/3) Everyone's reading this chart as a bet on when AGI arrives – take your pick. It isn't a disagreement about when. It's a disagreement about what "intelligence" even is. Benchmarks, evals, reasoning traces or something more broader?
@davidpattersonx Whether we’re in the singularity or not depends on how you define it. Raw intelligence is necessary but not sufficient
https://t.co/VM8zvHjbku
(1/3) Everyone's reading this chart as a bet on when AGI arrives – take your pick. It isn't a disagreement about when. It's a disagreement about what "intelligence" even is. Benchmarks, evals, reasoning traces or something more broader?
@haider1 Whether we’re in the singularity or not depends on how you define it. Raw intelligence is necessary but not sufficient
https://t.co/VM8zvHjbku
(1/3) Everyone's reading this chart as a bet on when AGI arrives – take your pick. It isn't a disagreement about when. It's a disagreement about what "intelligence" even is. Benchmarks, evals, reasoning traces or something more broader?
@KentonVarda Another analogy might be "outsourcing" coding vs. "delegating" coding (esp. for non-engineers).
However, AI agents are still local optima maximizers and for outsourcing like usage, you need them to follow the spirit of the goal, not letter of the prompt.
Working with agents = working with new analysts who are smarter than you. They knw their technicals, are resourceful, tireless, and eager to please. They stumble adjusting goals – zooming from detail to big-pict & back.
End-output is bottlenecked by reviewer bandwidth. That's you
(3/3) Another view: Perez's surge cycle runs ~half a century. ~20-30 yrs of installation before deployment pays broadly. A 20-25 yr onset-to-boost lag is installation from outside.
Place AI on that clock. We're not yet halfway.
(1/3) 5 general purpose techs - every time, the productivity payoff arrives decades after launch.
The lag compressed as each wave landed on the last one's finished infra - grid made the moving assembly line possible, internet shipped chatbot overnight.
But there's a floor: ~20-25 yrs
(2/3) Why a floor? Two clocks.
1st the tech iterates – 1995 internet ≠ what reorganized daily life. Frontier is still a benchmark argument; memory, context, goal adjustment getting started.
2nd institutions reorganize – workflows, org charts, careers. Not bolted on. Rebuilt around it
@rand_longevity Depends on how you define it? 2035+
Raw intelligence was necessary but not sufficient. Tools, accumulated & shared context and social fabric also critical and the "loop" that it created. We are still focused on which model, what benchmark!
https://t.co/VM8zvHjbku
(1/3) Everyone's reading this chart as a bet on when AGI arrives – take your pick. It isn't a disagreement about when. It's a disagreement about what "intelligence" even is. Benchmarks, evals, reasoning traces or something more broader?
@kimmonismus Whether we’re in the singularity or not depends on how you define it. More important: how has it actually changed how you live?
https://t.co/VM8zvHjbku
(1/3) Everyone's reading this chart as a bet on when AGI arrives – take your pick. It isn't a disagreement about when. It's a disagreement about what "intelligence" even is. Benchmarks, evals, reasoning traces or something more broader?