3 ms·
That's not what ML is. You have some input, which may include labels or not. If it does have labels, ML is the automatic creation of a function that maps the
by ben_w 2y ago
That's not what ML is.
You have some input, which may include labels or not.
If it does have labels, ML is the automatic creation of a function that maps the examples to the labels, and the function may be arbitrarily complex.
When the function is complex enough to model the world that created the examples, it's also capable of modelling any other input; this is why LLMs can translate between languages without needing examples of the specific to-from language pair in the training set.
The data may also be synthetic based on the rules of the world; this is how AlphaZero beats the best Go and chess players without any examples from human play.