By the end of 2022, OpenAI, Google, Facebook and DeepMind will each have large GPT-3-scale multimodal transformers trained on images and video as well as text.
Partly right ●●●●●
“OpenAI, Google, Facebook, and DeepMind all have gigantic multimodal transformers, similar in size to GPT-3 but trained on images, video, maybe audio too, and generally higher-quality data.”
- Said on
- 6 August 2021 · 2022 section of his year-by-year AI scenario post
- Source
- What 2026 looks like · LessWrong
- Should come true
- 2022
- How sure they sounded
- stated as fact in his best-guess scenario ("as realistic (to me) as I can currently manage")
- Categories
- AI
- Importance
- 3 of 5 · Direction of frontier AI development across the industry
Outcome
Partly right Happened 31 December 2022.
DeepMind (Flamingo, Apr 2022) and Google (PaLI, Sep 2022) had large multimodal transformers in 2022, but OpenAI's multimodal GPT-4 was released only in March 2023 and Meta had no comparable GPT-3-scale multimodal model in 2022. Half of the named labs met it within the window.
- https://deepmind.google/discover/blog/tackling-multiple-tasks-with-a-single-visual-language-model/
- https://research.google/blog/pali-scaling-language-image-learning-in-100-languages/
- https://en.wikipedia.org/wiki/GPT-4
History
- 2026-10-17 07:27:07 · created by claude
- 2026-10-17 07:27:07 · categorized by claude
- 2026-10-17 07:28:03 · scored by claude