| Computing desk | ||
|---|---|---|
| < December 23 | << Nov | December | Jan >> | December 25 > |
| Welcome to the Wikipedia Computing Reference Desk Archives |
|---|
| The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages. |
December 24 Information
Is all AI, at least, all practical applications of AI, nowadays just some flavor of ML? -- 92.191.143.129 ( talk) 12:56, 24 December 2018 (UTC)
- Artificial intelligence is a very broad field. Machine learning is a subset of artificial intelligence. 216.59.42.36 ( talk) 14:25, 24 December 2018 (UTC)
- Same OP, another IP here.
- I'm slightly confused still about the whole picture surrounding AI, ML and things like deep learning, neuronal networks.
- According to the box on the right of Artificial intelligence, machine learning is a major goal, but not an approach. As approaches it lists: symbolic, deep learning, bayesian networks, evolutionary algorithms.
- On the other hand, Outline of artificial intelligence lists machine learning as just one approach. That is, it divides the approaches in symbolic (called Good Old Fashioned AI), subsymbolic (with a load of subdivisions, including ML) and statistical.
- And why isn't ML classified into statistical? Isn't it somehow a way of generating stats about what to do in context or after X happens?
- Why don't we have a rule-based approach in the AI article or outline? Wouldn't that be feasible and a complete different animal than the approaches mentioned above? --
83.39.57.239 (
talk)
17:09, 24 December 2018 (UTC)
- The problem is that machine learning has one meaning in computer science and another outside of computer science. In computer science, it is strictly a subset of artificial intelligence where a thing (not strictly defined, but usually a neural network) is trained to provide a certain response for each category of input. It is pattern matching or clustering. Outside of computer science, machine learning is a machine that learns. How? Magic.
71.12.10.227 (
talk)
23:09, 24 December 2018 (UTC)
- Indeed, one of the great troubles with these fields is that the terminology is so poorly defined, and it is used so widely that it is difficult to unify all the different lines of effort under a single umbrella definition.
- Here is a review article: A Collection of Definitions of Intelligence, which was presented to the ACM at the a 2007 AI conference.
- Nimur ( talk) 18:20, 25 December 2018 (UTC)
- The problem is that machine learning has one meaning in computer science and another outside of computer science. In computer science, it is strictly a subset of artificial intelligence where a thing (not strictly defined, but usually a neural network) is trained to provide a certain response for each category of input. It is pattern matching or clustering. Outside of computer science, machine learning is a machine that learns. How? Magic.
71.12.10.227 (
talk)
23:09, 24 December 2018 (UTC)
- The overlap between ML and AI changes with time because we define them in a very different way.
- ML's definition is an operational one. ML is a computer system that's able to improve itself through experience. Even if we use neuronal networks for that nowadays, if we find something better in the future, we will keep calling it ML probably as long as it operates under the same scheme. And that does not have to be an intelligent task.
- Some ML solutions are not specially intelligent. It can be something that even non-humans are able to perform like walking. ML has created some walking robots that mimic this.
- AI is the intelligence created by humans, whatever your definition of intelligence may be. That includes when human learn from the data and design the system and when humans define the system that learns from the data. Doroletho ( talk) 01:02, 26 December 2018 (UTC)
| Computing desk | ||
|---|---|---|
| < December 23 | << Nov | December | Jan >> | December 25 > |
| Welcome to the Wikipedia Computing Reference Desk Archives |
|---|
| The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages. |
December 24 Information
Is all AI, at least, all practical applications of AI, nowadays just some flavor of ML? -- 92.191.143.129 ( talk) 12:56, 24 December 2018 (UTC)
- Artificial intelligence is a very broad field. Machine learning is a subset of artificial intelligence. 216.59.42.36 ( talk) 14:25, 24 December 2018 (UTC)
- Same OP, another IP here.
- I'm slightly confused still about the whole picture surrounding AI, ML and things like deep learning, neuronal networks.
- According to the box on the right of Artificial intelligence, machine learning is a major goal, but not an approach. As approaches it lists: symbolic, deep learning, bayesian networks, evolutionary algorithms.
- On the other hand, Outline of artificial intelligence lists machine learning as just one approach. That is, it divides the approaches in symbolic (called Good Old Fashioned AI), subsymbolic (with a load of subdivisions, including ML) and statistical.
- And why isn't ML classified into statistical? Isn't it somehow a way of generating stats about what to do in context or after X happens?
- Why don't we have a rule-based approach in the AI article or outline? Wouldn't that be feasible and a complete different animal than the approaches mentioned above? --
83.39.57.239 (
talk)
17:09, 24 December 2018 (UTC)
- The problem is that machine learning has one meaning in computer science and another outside of computer science. In computer science, it is strictly a subset of artificial intelligence where a thing (not strictly defined, but usually a neural network) is trained to provide a certain response for each category of input. It is pattern matching or clustering. Outside of computer science, machine learning is a machine that learns. How? Magic.
71.12.10.227 (
talk)
23:09, 24 December 2018 (UTC)
- Indeed, one of the great troubles with these fields is that the terminology is so poorly defined, and it is used so widely that it is difficult to unify all the different lines of effort under a single umbrella definition.
- Here is a review article: A Collection of Definitions of Intelligence, which was presented to the ACM at the a 2007 AI conference.
- Nimur ( talk) 18:20, 25 December 2018 (UTC)
- The problem is that machine learning has one meaning in computer science and another outside of computer science. In computer science, it is strictly a subset of artificial intelligence where a thing (not strictly defined, but usually a neural network) is trained to provide a certain response for each category of input. It is pattern matching or clustering. Outside of computer science, machine learning is a machine that learns. How? Magic.
71.12.10.227 (
talk)
23:09, 24 December 2018 (UTC)
- The overlap between ML and AI changes with time because we define them in a very different way.
- ML's definition is an operational one. ML is a computer system that's able to improve itself through experience. Even if we use neuronal networks for that nowadays, if we find something better in the future, we will keep calling it ML probably as long as it operates under the same scheme. And that does not have to be an intelligent task.
- Some ML solutions are not specially intelligent. It can be something that even non-humans are able to perform like walking. ML has created some walking robots that mimic this.
- AI is the intelligence created by humans, whatever your definition of intelligence may be. That includes when human learn from the data and design the system and when humans define the system that learns from the data. Doroletho ( talk) 01:02, 26 December 2018 (UTC)