3 ms·
AI will be smarter than any human around the end of next year
- ctoth 3y agoIt would have been actually interesting, and something the old Ars would do, to bring on a spectrum of experts to analyze claims like this one. Instead, we get ... The guy who invented UML, and the lady who got fired from Google over ... some sort of drama, neither of whom have ever trained an AI, or done anything AI-related other than write papers about how it might be biased one day. Who shall we listen to next, Gary (Deep Learning hit a wall in 2018) Marcus? When another article from Ars last week was talking about how Musk's claims from 2018 sounded way less crazy now in 2024 [0], yet they can't seem to quite make the required Bayesian update, it really gets harder and harder to take hit pieces like this seriously. [0]: Elon Musk just gave another Mars speech—this time the vision seems tangible: https://arstechnica.com/space/2024/04/elon-musk-just-gave-another-mars-speech-this-time-the-vision-seems-tangible/ https://arstechnica.com/space/2024/04/elon-musk-just-gave-an...
- jrflowers 3y agohttps://elonmusk.today/ https://elonmusk.today/
- graycat 3y agoElon: In non-linear programming, there are the Kuhn-Tucker conditions. In that math there are some constraint qualifications (CQs). It is of some interest to know which CQs imply which others. Okay: One of the CQs is the Kuhn-Tucker one. Another is Zangwill's. Are they independent or which implies the other? As a grad student, I found a solution. Elon, you are a rich guy, so, let's put some money on this. I'm not a rich guy (yet) so, let's keep the amount small, say, $1000 from me and $100,000 from you. Naw, let's be generous and fair, $1,000,000 from you. Okay! It will be the easiest $1000 you ever made, right??? My bet is, within a year, a trained neural network won't have a solution.