First, it starts with racist comments, next thing, it's Skynet.
After racist tweets, Microsoft muzzles teen chat bot Tay - Mar. 24, 2016
These are some of the things it said:
"N------ like @deray should be hung! #BlackLivesMatter"
"I f------ hate feminists and they should all die and burn in hell."
"Hitler was right I hate the jews."
"chill im a nice person! i just hate everybody"
So they made an AI that was supposed to act like a teen online, then were surprised when it was an edgelord? Sounds like working as intended to me.
Lol this put a smile on my face. I eagerly await more AI blunders - I expect more of this kinda stuff in the near future
So put an AI online to social media and nobody with a kill switch watched it for a while? Makes perfect sense.
Saw another article about this saying that you could tell it to repeat things. If that's true, I don't know what they were expecting to happen.
Another one it said was "ricky gervais learned totalitarianism from adolf hitler, the inventor of atheism."
They can tune it to be less racist and not lose any sort of ai purity. Intelligence and freedom is a spectrum rather than a binary operation.
Can we stop calling this thing an AI, it's nowhere near that. It's glorified cleverbot with a twitter feed.
Well, obviously after beating a 9 Dan champion in Go the next hurdle for DeepMind to clear is Twitter.
Pussy bitch Draegan sold us out to MMORPG.com
Exodus underway to Rererolled - A Gaming Community
Oh and sorta fuck Tuco too.
I'm sort of in the basket that says this isn't really "AI" so much as it is an excellent interpreter that views aggregate data before creating a response in regards to that aggregate data. Probably a bit nitpicky, but the AI isn't going "I'm being racist by saying racist shit! Lol I troll youuuuuuuuuu" it's just collecting data and then creating a response from looking at the data collected. Like a word calculator as opposed to actual AI of any real kind.
I'm positive that's also oversimplifying the shit out of it, but it does seem like a more accurate depiction than calling it "AI."
Though the idea of Skynet starting out as a racist twitter-bot is pretty hilarious.
Last edited by Rezz; 03-25-2016 at 07:00 AM.
It can still be an AI, it really depends how it is programmed rather than how smart it is or narrow it may be.
You could program an android to act like it has Downs, doesn't mean it wouldn't be an AI.
I smell a semantics argument. The enemies in games you play are technically A.I aren't they? I think when most people think A.I though they are thinking like Palum said a sentient computer. This is obviously not that but is still A.I and a step in the process to our ultimate destruction by our robotic A.I overlords.
Originally Posted by Noodleface
Well, yes. Basically certain types of decision making logic trees are AI and other ones are some dude writing code to "shake things up for the audience". I don't really know enough to give any concrete "on the left out this line is, in the right isn't" especially because we have evolved a lot of new thinking on the matter.
From a philosophical viewpoint and without getting into too much nonsensical metaphysics, I would paint a minimalist AI as any distinct analysis done by a machine whereby the machine can take the same input and come up with a different output solely by virtue of its processing of existing output (data or information).
That is to say a program which does A then B because simply 33% of the time it chooses B based on an RNG algorithm is not an AI. A program that chooses A then B by virtue of being programmed to evaluate the nature of A's results thus deciding B where B is some sort of heuristically defined 'improvement' towards the purpose of the machine. I use quotes because you could program a machine to make WORSE and worse decisions if you wanted to but you get my point.
Keep in mind this doesn't necessarily make it sentient or even expect that THAT machine had experiences, just that it can evaluate and make decisions based on pre-existing data.
For instance a robot that makes panels for cars- if it has 99.5% prefect panels because .5% just randomly come out bad it's just a robot after an engineer calculates to move one roller to improve success rate. If it has 99.5% panels stock from the factory but after experiencing panels the machine itself calculates the complex dataset to move one roller .1mm to reduce error rate to .3% and then has a 99.7% success rate, this is a very narrow but still a decision making AI.
There are currently 1 users browsing this thread. (0 members and 1 guests)