Eric Schmidt on Navigating an AI-Enabled Future (TransformX 2022)

Eric Schmidt in conversation with Scale AI’s Alexandr Wang at TransformX 2022.

Why it was kept

Kyle’s notes: transformers took over, then diffusion; AI was always about analyzing patterns, so what can be analyzed can eventually be generated; incumbents struggle to integrate new modalities, and their data is often not normalized; open source follows fast behind large labs; OpenAI kept GPT-2 less accessible for a while; bad actors in any large community; and Alexandr Wang’s line that technologists need geopolitical training.

As Kyle captured it

• (6) Eric Schmidt on Navigating an AI-Enabled Future | TransformX 2022 with Alexandr Wang at Scale AI and Eric Schmidt
  • Transformers took over, in terms of architecture
  • Diffusion came next
  • AI was always about analyzing patterns, and if you can analyze it, eventually you can generate it.
  • Every incumbent says they need to stay up to speed, but very few of them do it. Large companies haven’t been able to integrate these new modalities into their existing flows
  • Sometimes its they don’t know where their data is, or its not normalized.
  • Eric Schmidt: A lot of this innovation has happened in these large companies and super well-funded startups; but now you have open source community following very quickly
  • Many of these companies require access to these technologies as quickly as possible
  • OpenAI kept GPT2 less accessible for a while, but people’s demand for these models keep pushing them into the world.
  • It’s naive to assume this technology won’t be used by evil people. Any community that is large enough will have at least one bad actor.
  • “Technologists need to go through geopolitical training, just like they go through any other training.” Alexandr Wang