4 ms·

For how many generations have we said that Artificial General Intelligence will be "solved in our lifetime". Seems like it is alway a couple of decades away.

For how many generations have we said that Artificial General Intelligence will be "solved in our lifetime". Seems like it is alway a couple of decades away. But, if the definition is narrowed, there might be parts of the field that can be solved.

ggchappell 13y ago
> For how many generations have we said that Artificial General Intelligence will be "solved in our lifetime". Ah, but it isn't the same "we". 60 years ago, I guess your basic computing researcher thought human-level AI was just around the corner. And why not? Suddenly there was this wonderful machine that could do amazing feats of reasoning and computation in an eyeblink. And computers were constantly being improved; who knew what they would be capable of in a few more years. Nowadays, it's mainly Ray Kurzweil and a few others like him. And ... well, they're basically paid to say it. I could get a group of 1000 people together and give them a speech about how 2040 will be much like today, except for cooler phones, more expensive gas, and faster pizza delivery, and they'll all be bored and go home disappointed. Get Ray K. in front of the same group telling them the future is going to be indescribably different, and they're interested. Some people do write-ups on the speech. It gets discussed. It's something you hear about. And Ray K. is the one who gets invitations to other speaking engagements. In short, when you hear that super-AI is on the horizon, remember that, however far-fetched that might be, this is an interesting thought. The idea that it's a long way off, is not nearly so catchy. (See also memes, etc.)
- orionblastar 13y ago
  It is a matter of building data structures and algorithms to teach a computer the meaning of words, and then use logic of how those words fit together along with parts of speech. Then make ones for logic, reason, critical thinking, and then a conscience to screen out any bad 'thoughts' (Yes you have to make a data structure for thoughts and algorithms for them as well and then a conscience function to determine if they are good or evil, like 'slice loaf of bread' is good 'slice finger off person' is evil.) All I've been able to do is use a Thesaurus to paraphrase words at random. This Artificial genuine Intelligence will need a Thesaurus database to keep track of words and words that are like them for fuzzy logic comparisons. https://github.com/orionblastar/blastarparaphrase https://github.com/orionblastar/blastarparaphrase My PHP/MYSQL code is just an alpha test proof of concept and I need help with it as it just does random replacements. Someone I know has made some AI chatbots with Ruby and Python etc over here: http://subbot.org/ http://subbot.org/ They too need work, but the source code is open and you can look at it and contact the developer.
  - eli_gottlieb 13y ago
    It is a matter of building data structures and algorithms to teach a computer the meaning of words, and then use logic of how those words fit together along with parts of speech. Then make ones for logic, reason, critical thinking, and then a conscience to screen out any bad 'thoughts' (Yes you have to make a data structure for thoughts and algorithms for them as well and then a conscience function to determine if they are good or evil, like 'slice loaf of bread' is good 'slice finger off person' is evil.) This describes something an AGI would be able to do. This is nowhere near an accurate definition of an AGI. I'm about to write a top-level comment on this subject pointing to the actual current science on the subject, so go read ;-).

ggchappell 13y ago

> For how many generations have we said that Artificial General Intelligence will be "solved in our lifetime". Ah, but it isn't the same "we". 60 years ago, I guess your basic computing researcher thought human-level AI was just around the corner. And why not? Suddenly there was this wonderful machine that could do amazing feats of reasoning and computation in an eyeblink. And computers were constantly being improved; who knew what they would be capable of in a few more years. Nowadays, it's mainly Ray Kurzweil and a few others like him. And ... well, they're basically paid to say it. I could get a group of 1000 people together and give them a speech about how 2040 will be much like today, except for cooler phones, more expensive gas, and faster pizza delivery, and they'll all be bored and go home disappointed. Get Ray K. in front of the same group telling them the future is going to be indescribably different, and they're interested. Some people do write-ups on the speech. It gets discussed. It's something you hear about. And Ray K. is the one who gets invitations to other speaking engagements. In short, when you hear that super-AI is on the horizon, remember that, however far-fetched that might be, this is an interesting thought. The idea that it's a long way off, is not nearly so catchy. (See also memes, etc.)

orionblastar 13y ago
It is a matter of building data structures and algorithms to teach a computer the meaning of words, and then use logic of how those words fit together along with parts of speech. Then make ones for logic, reason, critical thinking, and then a conscience to screen out any bad 'thoughts' (Yes you have to make a data structure for thoughts and algorithms for them as well and then a conscience function to determine if they are good or evil, like 'slice loaf of bread' is good 'slice finger off person' is evil.) All I've been able to do is use a Thesaurus to paraphrase words at random. This Artificial genuine Intelligence will need a Thesaurus database to keep track of words and words that are like them for fuzzy logic comparisons. https://github.com/orionblastar/blastarparaphrase https://github.com/orionblastar/blastarparaphrase My PHP/MYSQL code is just an alpha test proof of concept and I need help with it as it just does random replacements. Someone I know has made some AI chatbots with Ruby and Python etc over here: http://subbot.org/ http://subbot.org/ They too need work, but the source code is open and you can look at it and contact the developer.
- eli_gottlieb 13y ago
  It is a matter of building data structures and algorithms to teach a computer the meaning of words, and then use logic of how those words fit together along with parts of speech. Then make ones for logic, reason, critical thinking, and then a conscience to screen out any bad 'thoughts' (Yes you have to make a data structure for thoughts and algorithms for them as well and then a conscience function to determine if they are good or evil, like 'slice loaf of bread' is good 'slice finger off person' is evil.) This describes something an AGI would be able to do. This is nowhere near an accurate definition of an AGI. I'm about to write a top-level comment on this subject pointing to the actual current science on the subject, so go read ;-).