5 ms·

like other commenters, I fail to see the AI here. How are matching algorithms absent any machine learning considered to be AI? I thought AI was all about solvin

like other commenters, I fail to see the AI here. How are matching algorithms absent any machine learning considered to be AI? I thought AI was all about solving probabilistic problems where deterministic ones didn't exist?

kgwgk 8y ago
AI can be anything involving computers... a few years ago they might have mentioned big data, I’m sure they have databases with thousands of people. “This series explores the promise of AI to personalize, democratize, and advance medicine—and the dangers of letting machines make decisions.” You have to work with what you have, I guess.
YeGoblynQueenne 8y ago
Out of curiousity, how did you come by this definition? I can't say I recognise it! Like I say in another comment, most of AI is not about machine learning, but rather about reasoning in complex domains, very often using hand-crafted procedures. For example, most of the AI systems of the '80s and '90s, which were in fact hailed as great successes of AI, were of the expert system variety: in short, a large, hand-crafted database of production rules elicited from domain experts, coupled with an inference procedure. For instance, an early expert system was MYCIN, used to diagnose bacterial infections. It was one of the earliest AI systems to perform at least as good, and even better than, experts: https://en.wikipedia.org/wiki/Mycin https://en.wikipedia.org/wiki/Mycin
- rexreed 8y ago
  The sentence you're referring to is not a definition of AI, but rather it's most suitable intent. If you have truly deterministic systems, you don't need any sort of intelligence, other than the intelligence of the human that put together the system. The machine is simply executing the intelligence of the human and not providing any of its own or adding to the picture, other than providing leverage by means of automation. Waffle irons aren't intelligent, even tho they automate the process of making waffles. If your definition of intelligence renders the machine to be simply an executor of intelligence from humans, then that's not particularly useful and then you might as well just say "computing" whenever you say AI. I tend to believe that AI is best suited to solve problems that can't be easily solved with deterministic, programming approaches. Machines can be trained to provide value on the logic and decision-making side rather than just on the decision-executing side.
QML 8y ago
If you go to slide 70 of the CMU link posted above [1], it covers an algorithm called FutureMatch [2], which attempts to assess or predict the quality of a matching. I wouldn't call it AI as the article author calls the algorithm, but more of a subset of ML -- reinforcement learning (RL). [1] http://www.cs.cmu.edu/~arielpro/15896s16/slides/896s16-14.pdf http://www.cs.cmu.edu/~arielpro/15896s16/slides/896s16-14.pd... [2] https://www.cs.cmu.edu/~sandholm/futurematch.aaai15%20with%20appendix.pdf https://www.cs.cmu.edu/~sandholm/futurematch.aaai15%20with%2...

kgwgk 8y ago

AI can be anything involving computers... a few years ago they might have mentioned big data, I’m sure they have databases with thousands of people. “This series explores the promise of AI to personalize, democratize, and advance medicine—and the dangers of letting machines make decisions.” You have to work with what you have, I guess.

YeGoblynQueenne 8y ago

Out of curiousity, how did you come by this definition? I can't say I recognise it! Like I say in another comment, most of AI is not about machine learning, but rather about reasoning in complex domains, very often using hand-crafted procedures. For example, most of the AI systems of the '80s and '90s, which were in fact hailed as great successes of AI, were of the expert system variety: in short, a large, hand-crafted database of production rules elicited from domain experts, coupled with an inference procedure. For instance, an early expert system was MYCIN, used to diagnose bacterial infections. It was one of the earliest AI systems to perform at least as good, and even better than, experts: https://en.wikipedia.org/wiki/Mycin https://en.wikipedia.org/wiki/Mycin

rexreed 8y ago
The sentence you're referring to is not a definition of AI, but rather it's most suitable intent. If you have truly deterministic systems, you don't need any sort of intelligence, other than the intelligence of the human that put together the system. The machine is simply executing the intelligence of the human and not providing any of its own or adding to the picture, other than providing leverage by means of automation. Waffle irons aren't intelligent, even tho they automate the process of making waffles. If your definition of intelligence renders the machine to be simply an executor of intelligence from humans, then that's not particularly useful and then you might as well just say "computing" whenever you say AI. I tend to believe that AI is best suited to solve problems that can't be easily solved with deterministic, programming approaches. Machines can be trained to provide value on the logic and decision-making side rather than just on the decision-executing side.

QML 8y ago

If you go to slide 70 of the CMU link posted above [1], it covers an algorithm called FutureMatch [2], which attempts to assess or predict the quality of a matching. I wouldn't call it AI as the article author calls the algorithm, but more of a subset of ML -- reinforcement learning (RL). [1] http://www.cs.cmu.edu/~arielpro/15896s16/slides/896s16-14.pdf http://www.cs.cmu.edu/~arielpro/15896s16/slides/896s16-14.pd... [2] https://www.cs.cmu.edu/~sandholm/futurematch.aaai15%20with%20appendix.pdf https://www.cs.cmu.edu/~sandholm/futurematch.aaai15%20with%2...