Talk about your sensationalist titles.
The article itself describes the program being something more of simulating our ability to anticipate what happens next, specifically in the context of cooking and catering and being ready to assist a colleague/customer by trying to know what they'll want next. Interesting stuff regardless.
the smartest people to have ever lived on planet earth couldnt even look 1 minute into the future what hope does this "program" have
I must be a genius then because last night I was all like "I'm going to wake up tomorrow", and I did.
Humans: 1
Computers: 0
I keep thinking the opposite, but it just won't happen.
le funny suicide meme
just to clear my name by association with Seth2492 and therefore ComicSansler i don't condone suicide, so don't do it.
Science headlines are always sensationalist but damn, this one is quite impressive.
Accuracy was over 40 percent for short forecast periods
Must be quite the amazing thing. /s
Honestly I'm not even really impressed by what the thing actually is. Mostly because it uses neural networks.
Neural networks are always so hyped and they can be quite impressive but the biggest downside is that they need tons of data and time to learn.
That is for one task. Now they want to use it for something as adaptive and can anticipate variable events?
They admit that without good data and having to identify the steps from video itself (a whole different field) it gets a lot worse:
Especially since the algorithm performs noticeably worse if it has to recognize on its own what happens in the first part of the video, instead of being told.
Feels like a lot of what they want to create you can just replace by dumber automatic methods.
The automatic vacuum cleaner meanwhile knows that it has no business in the kitchen at that time, and instead takes care of the living room.
Do I really need a fancy neural network based learning AI vacuum cleaner for that? Why not just make it avoid the kitchen based on simple sensor data like an active stove, activity on the fridge etc.
https://www.youtube.com/watch?v=xMNYRcVH_oI
Maybe I'm too harsh on it but it heavily relies on a lot of training data, and correct detection of information.
Fully realizing this for anything that isn't a specified task in a fairly controlled environment will be tough.
Sorry, you need to Log In to post a reply to this thread.
