4 ms·

AI doesn't exist. There is nothing inherently intelligent about deep NNs which are merely very good pattern matching algorithms, statistics really... The term A

AI doesn't exist. There is nothing inherently intelligent about deep NNs which are merely very good pattern matching algorithms, statistics really... The term AI suggests something is thinking which is false marketing at best, abused by thousands of consulting companies for a decade now. Unless of course your opinion is that thinking is just statistics which makes for an interesting metaphysical discussion.

autokad 5y ago
some of the new techniques in natural language processing that uses transformers is getting closer to understanding. Understanding may be an illusion though. I wrote a game in 2002 for a class I was in, you have a space ship and you shoot alien ships. To make the game challenging, when the user was far away from the alien ships they would move randomly, if you moved in on them they moved away in a random direction, and once you got too close they would vector strait for the ship in a kamakaze attack. I also needed to figure out when the alien ships would fire back, so I just made them randomly fire when the user also fired. Given that your ship's firing range was limited, this created a dangerous scenario where you tried to chase down a ship, and at the last moment it would come right at you and fire. the point of all of this was that when people played the game, they stated "wow, the AI of this game is really good". People believed the AI was thinking when it was just a few rules. I dont suppose its impossible to be tricked into thinking something had understanding even if that something is itself. if something is complicated enough and good enough at making decisions, it seems possible its not understanding but just a sufficiently complicated process, or maybe a sufficiently complicated process is understanding
YeGoblynQueenne 5y ago
"Pattern matching" is what regular expressions do. What you are alluding to is "pattern recognition", which is an older name for so-called "statistical" machine learning, particularly used to perform "perception" tasks (such as machine vision). "Statistical" machine learning is also not very much about statistics, it's just that similar technologies and concepts are sometimes used as in statistics, for example probabilities, function optimisation etc. "Artificial Intelligence" (AI) is the name of the field of research that studies approaches to artificial intelligence. It's a bit confusing but when someone says "I study AI" they don't, today, mean that they study artificially intelligent entities, rather they mean they study approaches that might one day allow us to create such entities.

autokad 5y ago

some of the new techniques in natural language processing that uses transformers is getting closer to understanding. Understanding may be an illusion though. I wrote a game in 2002 for a class I was in, you have a space ship and you shoot alien ships. To make the game challenging, when the user was far away from the alien ships they would move randomly, if you moved in on them they moved away in a random direction, and once you got too close they would vector strait for the ship in a kamakaze attack. I also needed to figure out when the alien ships would fire back, so I just made them randomly fire when the user also fired. Given that your ship's firing range was limited, this created a dangerous scenario where you tried to chase down a ship, and at the last moment it would come right at you and fire. the point of all of this was that when people played the game, they stated "wow, the AI of this game is really good". People believed the AI was thinking when it was just a few rules. I dont suppose its impossible to be tricked into thinking something had understanding even if that something is itself. if something is complicated enough and good enough at making decisions, it seems possible its not understanding but just a sufficiently complicated process, or maybe a sufficiently complicated process is understanding

YeGoblynQueenne 5y ago

"Pattern matching" is what regular expressions do. What you are alluding to is "pattern recognition", which is an older name for so-called "statistical" machine learning, particularly used to perform "perception" tasks (such as machine vision). "Statistical" machine learning is also not very much about statistics, it's just that similar technologies and concepts are sometimes used as in statistics, for example probabilities, function optimisation etc. "Artificial Intelligence" (AI) is the name of the field of research that studies approaches to artificial intelligence. It's a bit confusing but when someone says "I study AI" they don't, today, mean that they study artificially intelligent entities, rather they mean they study approaches that might one day allow us to create such entities.