Re:Microsoft bids $44.6B to buy Yahoo (2008)
by cbsmith on Tuesday May 04, 2021 @07:59PM
(#27044403)
> If you look at the queries in GPT-3 that are used to extract structured data, or to synthesize answers, the technical difference is huge, but it just has a high failure rate.
Yup. Though there are other systems that arguably had more technical success.
> There's also the general switch from searching for information, to doing things. We've gone from 10 blue links, to answers, to assistive functionality. The current state of assistive functionality is pretty shitty and has a long way to go IMHO.
Yes, I agree there's a great deal of opportunity for technical innovation. I expect to see a lot of it from Google themselves. Just don't kid yourself about the impact of those innovations on market share.
If you look at the queries in GPT-3 that are used to extract structured data, or to synthesize answers, the technical difference is huge, but it just has a high failure rate.
The various vertical search systems are examples of this, but the key to a next-gen system isn't to hand-code such verticals, but to have the AI be able to create an infinite number of them on the fly. Vertical based on Julia code? Yes. Vertical based on Pokemon cards? Yes. Today these things are hand-curated collections.
There's also the general switch from searching for information, to doing things. We've gone from 10 blue links, to answers, to assistive functionality. The current state of assistive functionality is pretty shitty and has a long way to go IMHO.