We keep misjudging pivotal tech: GPUs went from games to deep learning, Transformers from language to coding
bendee983 · x · 2026-09-19
The author argues we're bad at projecting the best applications of pivotal technologies. GPUs were built for gaming, but parallel matrix multiplication powered deep learning; Transformers were framed as language models, but sequence prediction turned out most valuable for coding. They wonder what diffusion models and world models (generative and JEPA-style) will really be used for — many of today's assumptions about AI will prove wrong.
More from AGI Musings
- Dario and Sam said we must slow down last week — so do they believe it? — _TheWolfOfWalmart_ · 2026-09-20
- How AI-Era Intermediaries Could Become Systemically Dangerous Power Brokers — iamtrask · 2026-09-20
- Ethan Mollick: policy should actively guide us toward the good AI world — emollick · 2026-09-20
- Ethan Mollick: AI likely ends broadly well like past GPTs, but a von Neumann-style singularity is possible — emollick · 2026-09-20
- iamtrask endorses betterpath.ai framework: compete on narrow AI, slow down general AI — iamtrask · 2026-09-20
- Researcher's New Routine: Most Time Now Spent Reading Papers Written by His Own Models — generativist · 2026-09-20