An AI Is Playing Pictionary To Figure Out How the World Works (technologyreview.com) 31
Researchers at the Allen Institute for AI (Ai2) believe that Pictionary could push machine intelligence beyond its current limits. To that end, they have devised an online version of the game that pairs a human player with an AI program. MIT Technology Review reports: In case you've never played it before, Pictionary involves trying to draw an image that conveys a written word or phrase for your teammates to guess. This tests a person's drawing skills but also the ability to convey complex meaning using simple concepts. Given the phrase "wedding ring," for example, a player might try to draw the object itself but also a bride and groom or a wedding ceremony.
That makes it the perfect vehicle to help teach machines. The team developed an online version of the game, called Iconary, that pairs a user with an AI bot called AllenAI. Both take turns as the artist and the guesser. Playing as artist, a user is given a phrase and then has to sketch things to convey it. The sketches are first turned into clip-art icons using computer vision; then the computer program tries to guess the phrase using a database of words and concepts and the relationship between them. If the program gets only part of the phrase, it will ask for another image to clarify. The AI program uses a combination of AI techniques to draw and guess. Over time, by playing against enough people, AllenAI should learn from their common-sense understanding of how concepts (like "books" and "pages") go together in everyday life, Fahadi says. It will also help the researchers explore ways for humans and machines to communicate and collaborate more effectively.
That makes it the perfect vehicle to help teach machines. The team developed an online version of the game, called Iconary, that pairs a user with an AI bot called AllenAI. Both take turns as the artist and the guesser. Playing as artist, a user is given a phrase and then has to sketch things to convey it. The sketches are first turned into clip-art icons using computer vision; then the computer program tries to guess the phrase using a database of words and concepts and the relationship between them. If the program gets only part of the phrase, it will ask for another image to clarify. The AI program uses a combination of AI techniques to draw and guess. Over time, by playing against enough people, AllenAI should learn from their common-sense understanding of how concepts (like "books" and "pages") go together in everyday life, Fahadi says. It will also help the researchers explore ways for humans and machines to communicate and collaborate more effectively.
In other news... (Score:3)
...that same AI read Slashdot for a month and promptly shot itself when it saw what goes on in AC-land.
You know what I"m talking about. There's a whole underbelly of political .. ahem... "discourse".. taking place at 0 and below.
I wish the lot of them would just... go away. It's not entertaining anymore. It's just children yelling at each other, just like the "grownups" they see on TV, yelling at each other.
Re: (Score:2)
Can I break it? (Score:2)
Could be fun.
There's so much innuendo with humans (Score:2)
(Assuming I could draw at all legibly) I could draw a ring in a pictionary game, then add a ball and chain to narrow the ring type down.
Humans can make intuitive leaps that might baffle machine learning algorithms.
Re: (Score:1)
It doesn't matter what you draw, the computer doesn't care is there a chain or a wedding dress. Of course if training data does not contain something, then it can't guess it. E.g. I could draw a tree, car and a triangle and claim that it is a wedding ring and the AI would never guess it. But if you train it with enough many people and some people draw also a chain, then the AI learns that. If none or very few draw it, then it is so insignificant that it doesn't need that information.
Mind you that my kids wo
Look the AI can do games (Score:2)
Re: (Score:2)
let's play missile command
Re: (Score:2)
Quick, Draw (Score:1)
Elsewhere (Score:2)
I'm teaching an AI beer pong and spin the bottle. I expect to take over the world pretty soon.
ugh (Score:2)
"SYNTAX ERROR" (Score:2)
?SYNTAX ERROR
>_
..and sits there with the cursor blinking, waiting for a new command.
As usual, I'm not impressed.
Finite search spaces (Score:2)
Machine learning operates within a model, influenced by implementer's bias. Anything it'll ever be able to do will be constrained to lie within the model. There is no way "an AI" will be able to understand human concepts the same way humans do, and in part that's because humans don't learn just by playing Pictionary.
When will this hype subside, so that we can actually focus on the real stuff machine learning can do, without constant conflation with the idea that "general AI" could even ever exist?