Hacker News new | past | comments | ask | show | jobs | submit login

Well, trying to build AI using failed papers from the 1980s have near zero value today.

What Tesla is doing is like building a modern app but refusing to do A/B tests or code review or version control, generally universally accepted methods in the current industry, and going back to re-try software development patterns of the 1980s that failed.




>>Well, trying to build AI using failed papers from the 1980s have near zero value today.

I think if you had picked any other example you'd have been correct, but many current-day AI advances are based on decades old techniques. One caption from an article about Geoff Hinton states: "This publication from the mid-1980s showed how to train a neural network with many layers. It set the stage for this decade’s progress in AI."[1]

[1] https://www.technologyreview.com/s/608911/is-ai-riding-a-one...


Cheers for the correction!




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: