Jeff Dean: AI isn't as smart as you think -- but it could be TED
Interesting talk Jeff Dean. He gave a TED talk two years ago about AI, in which he highlighted the development of neural networks. (On the left like now, and on the right where to dig)
- Separate models -> General purpose models
- Image or text or speach -> All modalities at once
- Dense models -> Sparse models Interestingly, these areas are now on everyone’s ears, however, Open AI with its ChatGPT became the leader. This may be due to the end of the performance, where Chris Anderson (head of TED ) I asked Jeff Dean about taking a responsible approach to AI. Jeff went on to explain how Google is responsible for AI, but this responsible approach seems to affect T2M. (time to market). But Google began to actively reduce its AI Ethics team and focused on reducing the gap from competitors.
P.S. I'm personally interested in following the career of Jeff, who has had a hand in many of Google's infrastructure projects:
- Google Search and whitepaper
- MapReduce and whitepaper
- BigTable and whitepaper
- Google Spanner and whitepaper
- Protocol Buffers (protobuf)
- And a lot of stuff.
Jeff is now Google’s Chief Scientist focusing on AI for Google DeepMind and Google Research. Interestingly, in the video I gave, Jeff says that he wrote a study paper on neural networks in the distant past. 1990 year:) Anyway, I think Jeff will do well and I’m very excited to learn the next AI whitepapers. (last 2022 year)
#ML #AI #DataScience