Predictions Scorecard, 2025 January 01
Rodney Brooks’s latest “prediction scorecard” on AI, robotics, self‑driving cars, and flying cars draws mixed reactions, with many readers arguing his time windows are so loose and criteria so malleable that he can retroactively claim success. Much of the debate centers on whether services like Waymo constitute true autonomous driving, how far LLMs and deep learning have really advanced toward general intelligence, and whether current hype meaningfully misallocates capital and jobs. Several commenters still find value in his role as a skeptic of overconfident tech forecasts, but others say the exercise feels self‑congratulatory and does little to improve how we reason about technological timelines.
Prediction methodology & tone
- Many readers find the prediction scheme (e.g., “NET20XX”) too loose, allowing wide time windows and post-hoc interpretation. Accusations of goalpost moving, especially on self‑driving and flying cars.
- Others defend the approach as a counterweight to 2017‑era hype and executive overconfidence. What seems “obvious” now was not in 2018.
- Several note the piece feels self-congratulatory and focused on proving past correctness rather than honestly reassessing errors.
- A minority appreciate the detailed reasoning and annual self‑audit as intellectually valuable despite the style.
Self-driving cars and Waymo
- Big dispute over whether current robo‑taxis mean previous “driverless taxi in a major US city” predictions should be counted as “hit” or “miss”.
- Some argue limited-area, operator-assisted services with occasional remote help are still not “self-driving” in the originally understood sense (buy a car that drives anywhere, no human backup).
- Others say that if operators rarely intervene (tens–hundreds of miles per intervention) and cars complete rides safely, this is practically self-driving.
- Frequent claim that the article downplays Waymo’s progress, nitpicks rare failures, and ignores that some riders find it safer than humans.
- Coverage and economics are major concerns: service still tiny relative to total miles driven; Alphabet’s modest investment vs. buybacks cited as evidence of perceived risk/limited upside.
- Actual remote-intervention rate is unknown; commenters emphasize this makes strong claims (from either side) speculative.
LLMs, AI hype, and deep learning
- Several think the article mischaracterizes LLMs as mere “lookup in weights”, ignoring clear evidence of novel reasoning and scenario handling. This undermines trust in its AI claims.
- Others agree with its warning against “exponentialism” and the assumption that scaling deep learning alone will deliver everything.
Flying cars & eVTOL
- Debate over whether high-end eVTOLs already satisfy “flying car for the wealthy” predictions.
- Constraints highlighted: energy density, safety, maintenance, pilot skill, weather, noise, and lack of autorotation/glide for many eVTOL designs.
- General consensus: demos and niche services are coming, but mass adoption remains unlikely or far off.
AI, jobs, and capital allocation
- Some see current AI as unlikely to be the main driver of recent tech layoffs; macroeconomic and tax changes are blamed instead, with “AI” used as PR cover.
- Others report real pressure on occupations like copywriting and foresee future displacement of drivers.
- Tension between hype as useful risk-taking vs. hype as misallocation of capital and source of human misery (e.g., billions on robo‑taxis vs. social needs).
Other technologies & meta
- Discussion of EV limits (recycling, insurance, grid capacity) and slower-than-hyped adoption.
- Battery futures: solid‑state, grid-scale chemistries, hydrogen for aviation; recognition of hard physical constraints.
- Several criticize the essay’s length and rambling structure; others see the background context as necessary to understand the critiques of hype.