5 ms·

Why? The machine learns via essentially trial-and-error to produce a model that best fits the data provided to it. The hope is that the training data is represe

Why? The machine learns via essentially trial-and-error to produce a model that best fits the data provided to it. The hope is that the training data is representative of "unseen" data the model later encounters in order to be useful. Seems pretty accurate given the broadness of the topic and variety of use cases.

abhinav22 6y ago
Because it’s regression and statistical analysis. Nothing special about ML except applying it to large(r) datasets, particularly in areas where it works well (various recommendation based models for example).
- KingOfCoders 6y ago
  How does this differ from you "learning" words of a foreign language?
  - loopz 6y ago
    Try to prove electronic computers/TPU works like an organic human.
    - drdeca 6y ago
      Why does it have to? If you define “learning” as “being like an organic human”, then sure perhaps it isn’t “learning” under that definition, but that seems like a poor definition, and I don’t think it is how people use the word. What does “learning” mean?
      - loopz 6y ago
        Good. Ponder on that question.
  - abhinav22 6y ago
    It’s not “learning” anything though. Difference is causation vs correlation: If I learn how to boil water for example, there is close to 100% causation that heating water causes it to boil (notwithstanding we do not know everything about Physics, eg before we would say the earth is flat because we learnt it so, but that was incorrect, ie not everything we learn is true, it’s just a super high level of correlation) However if a machine learning model says the probability of water boiling is 99% after heating water, that is actually correlation. So I would distinguish it from learning In this example there’s not much different. But when you apply to more complex examples, the difference between causation and correlation become clearer
    - DubiousPusher 6y ago
      You think people's ability to learn is really distinguished by an ability to discern causation and correlation? That does not comport with modern behavioral science.
      - abhinav22 6y ago
        To some degree yes. For example, can the fields of mathematics and advanced physics or programming be studied to the degree they are with an understanding of causation? For some things, correlation can be sufficient, so I can see where you are coming from. As a better counterpoint: if we could not distinguish between causation and correlation, why is general AI so hard?
        DubiousPusher 6y ago
        > can the fields of mathematics and advanced physics or programming be studied to the degree they are with an understanding of causation? Yes, pretty much. Newton's laws are a perfect example of something that establishes a correlation but are useful absent accurate causation. They did just fine until we needed something better. Einstein provided this but can we be so sure gravity as a result of curved space is "true" and not just a really good model? Perhaps the hunt for what quantum gravity will one day tell. > As a better counterpoint: if we could not distinguish between causation and correlation, why is general AI so hard? I think general AI is hard because for one it might be a fiction. For general AI to be real, we'd have to assume people posses "general intelligence". Or at the very least, were it beyond us, we'd have to assume we could tell if we had created it. What we are really trying to do with "general intelligence" is reproduce something that took 3.5 billion years of natural selection to design. We have only recently become capable of even creating machines with the kinds of part counts that biology comes with. Were we to produce "general intelligence" in the next thousand years I'd be surprised it came so easy.
- laichzeit0 6y ago
  A lot of current ML/AI algorithms come from the computer science departments in universities. They use different words for the same ideas that exist in the statistics community. There is more of a focus on prediction versus inference. The interpretability of these algorithms come second. Here’s a little translation mapping: https://3qeqpr26caki16dnhd19sv6by6v-wpengine.netdna-ssl.com/wp-content/uploads/2018/04/Glossary-Mapping-Terms-in-Statistics-to-Terms-in-Machine-Learning-972x1024.png https://3qeqpr26caki16dnhd19sv6by6v-wpengine.netdna-ssl.com/... Nothing really to get one’s panties in a knot about though.
- perlgeek 6y ago
  Your neurons also learn by processes similar to regression and statistical analysis. Why would it be OK to call it learning in a human or an animal, but not in software?
  - abhinav22 6y ago
    A human can learn maths and do pure analysis (if it was smart enough). A machine would struggle with simply machine learning to answer pure math questions. Although the M1 chip from Apple.... ;)
- st1x7 6y ago
  I don't see a problem with machine learning as a term. Especially since it incorporates more than regression methods - it's useful to have a general term without having to specify the type of algorithm each time.
  - abhinav22 6y ago
    Fair enough. Although I would prefer a less sensational term, I liked “Big Data” a lot in that regard.
    - st1x7 6y ago
      The terms big data and machine learning describe very different things though. You can't use them interchangeably. I would also argue that big data is the more sensational one of the two. Machine learning can be easily defined as a group of algorithms. Big data on the other hand is a vague notion that doesn't make it clear what is happening at all and this ambiguity allows sensationalizing.
BoiledCabbage 6y ago
The question is: Is the machine learning, or is the machine intelligent. The answer is its neither. But we've taken them to be useful analogies and loosely designated them to mean different things. Neither one really applies because in truth, the machine isn't learning. Because we've deemed some types of organs optimizers, others dynamic programing and others learning. It's all arbitrary.
- DubiousPusher 6y ago
  I don't believe it is entirely arbitrary. Just as there are categorizations of machine, turing machine vs push down automata vs finite state machines, which determine their ability to solve certain problems, there are differences between categories of algorithm and their ability to solve certain problems. There are many "learning" algorithms beyond what is popularly called ML today but they have in common that they use information about their own success or failure to inform future results.

abhinav22 6y ago

Because it’s regression and statistical analysis. Nothing special about ML except applying it to large(r) datasets, particularly in areas where it works well (various recommendation based models for example).

KingOfCoders 6y ago
How does this differ from you "learning" words of a foreign language?
- loopz 6y ago
  Try to prove electronic computers/TPU works like an organic human.
  - drdeca 6y ago
    Why does it have to? If you define “learning” as “being like an organic human”, then sure perhaps it isn’t “learning” under that definition, but that seems like a poor definition, and I don’t think it is how people use the word. What does “learning” mean?
    - loopz 6y ago
      Good. Ponder on that question.
- abhinav22 6y ago
  It’s not “learning” anything though. Difference is causation vs correlation: If I learn how to boil water for example, there is close to 100% causation that heating water causes it to boil (notwithstanding we do not know everything about Physics, eg before we would say the earth is flat because we learnt it so, but that was incorrect, ie not everything we learn is true, it’s just a super high level of correlation) However if a machine learning model says the probability of water boiling is 99% after heating water, that is actually correlation. So I would distinguish it from learning In this example there’s not much different. But when you apply to more complex examples, the difference between causation and correlation become clearer
  - DubiousPusher 6y ago
    You think people's ability to learn is really distinguished by an ability to discern causation and correlation? That does not comport with modern behavioral science.
    - abhinav22 6y ago
      To some degree yes. For example, can the fields of mathematics and advanced physics or programming be studied to the degree they are with an understanding of causation? For some things, correlation can be sufficient, so I can see where you are coming from. As a better counterpoint: if we could not distinguish between causation and correlation, why is general AI so hard?
      - DubiousPusher 6y ago
        > can the fields of mathematics and advanced physics or programming be studied to the degree they are with an understanding of causation? Yes, pretty much. Newton's laws are a perfect example of something that establishes a correlation but are useful absent accurate causation. They did just fine until we needed something better. Einstein provided this but can we be so sure gravity as a result of curved space is "true" and not just a really good model? Perhaps the hunt for what quantum gravity will one day tell. > As a better counterpoint: if we could not distinguish between causation and correlation, why is general AI so hard? I think general AI is hard because for one it might be a fiction. For general AI to be real, we'd have to assume people posses "general intelligence". Or at the very least, were it beyond us, we'd have to assume we could tell if we had created it. What we are really trying to do with "general intelligence" is reproduce something that took 3.5 billion years of natural selection to design. We have only recently become capable of even creating machines with the kinds of part counts that biology comes with. Were we to produce "general intelligence" in the next thousand years I'd be surprised it came so easy.
laichzeit0 6y ago
A lot of current ML/AI algorithms come from the computer science departments in universities. They use different words for the same ideas that exist in the statistics community. There is more of a focus on prediction versus inference. The interpretability of these algorithms come second. Here’s a little translation mapping: https://3qeqpr26caki16dnhd19sv6by6v-wpengine.netdna-ssl.com/wp-content/uploads/2018/04/Glossary-Mapping-Terms-in-Statistics-to-Terms-in-Machine-Learning-972x1024.png https://3qeqpr26caki16dnhd19sv6by6v-wpengine.netdna-ssl.com/... Nothing really to get one’s panties in a knot about though.
perlgeek 6y ago
Your neurons also learn by processes similar to regression and statistical analysis. Why would it be OK to call it learning in a human or an animal, but not in software?
- abhinav22 6y ago
  A human can learn maths and do pure analysis (if it was smart enough). A machine would struggle with simply machine learning to answer pure math questions. Although the M1 chip from Apple.... ;)
st1x7 6y ago
I don't see a problem with machine learning as a term. Especially since it incorporates more than regression methods - it's useful to have a general term without having to specify the type of algorithm each time.
- abhinav22 6y ago
  Fair enough. Although I would prefer a less sensational term, I liked “Big Data” a lot in that regard.
  - st1x7 6y ago
    The terms big data and machine learning describe very different things though. You can't use them interchangeably. I would also argue that big data is the more sensational one of the two. Machine learning can be easily defined as a group of algorithms. Big data on the other hand is a vague notion that doesn't make it clear what is happening at all and this ambiguity allows sensationalizing.

BoiledCabbage 6y ago

The question is: Is the machine learning, or is the machine intelligent. The answer is its neither. But we've taken them to be useful analogies and loosely designated them to mean different things. Neither one really applies because in truth, the machine isn't learning. Because we've deemed some types of organs optimizers, others dynamic programing and others learning. It's all arbitrary.

DubiousPusher 6y ago
I don't believe it is entirely arbitrary. Just as there are categorizations of machine, turing machine vs push down automata vs finite state machines, which determine their ability to solve certain problems, there are differences between categories of algorithm and their ability to solve certain problems. There are many "learning" algorithms beyond what is popularly called ML today but they have in common that they use information about their own success or failure to inform future results.