3 ms·
> It's not "the new machine can gin cotton" or "the new machine can weave cloth", it's "the new machine can handle an extremely wide range of problems, and we'r
by moxious 10y ago
> It's not "the new machine can gin cotton" or "the new machine can weave cloth", it's "the new machine can handle an extremely wide range of problems, and we're moving from having to teach it each problem individually to having it teach itself."
It's funny you mention this because in the past few days on HN there have been a number of articles whose general point was that the machine learning stuff presently in use is well tailored to very narrow, specific problems, and that the AI tech that would tackle a broad range of problems is still very elusive.
This new machine that tackles a very wide range of problems without massive programming investment is something I've been wanting for 20 years, and I still don't see anyone with plans to bring it to market.
- Gargoyle 10y ago> the machine learning stuff presently in use is well tailored to very narrow, specific problems It's more accurate to say the machine learning in use now works on narrow classes of problems. But even that is a giant leap forward from teaching the machine to solve each problem individually. Teaching a car to drive very specific routes is one thing (and has been effectively done for decades now). Teaching a car to drive entire classes of routes is something else, and where we are now. In a decade or so cars will drive all classes of routes better than humans, and that will be something else again.