Comment by joe_the_user | Hacker News Reader

Comment by joe_the_user | Hacker News Reader

Parent Full thread

Machine Learning algorithms are only as good as the data they are trained on.

Machine Learning algorithms are never better than the data they're trained-on. But they can easily be "worse".

Specifically, an ML algorithm trained on one data set can have a hard time operating on a different data set with some similarities and some fundamental differences. This is related to algorithms generally not having an idea how accurate their predictions/guesses are. This in turn relates to current ML as being something like statistical prediction with worries about bias tossed-out (which isn't to dismiss it but to illuminate it's limitations).

For tasks like self-driving or spotting cancer in x-rays, they are producing novel result because these kinds of tasks are amenable to reinforcement. The algorithm crashed the car, or it didn't. The patient had cancer, or they didn't.

Ironically, both those applications have been failures so-far. Self-driving is far more complex than a binary crash-or-not scenario (usually the road is predictable but just about anything can wander into the roadway occasionally and you need to deal with this "long tail"). It also requires extreme safety to reach human levels. Diagnosis by image has problems of consistence, of doctors considering more than just an image and other things possibly not understood.

Agreed. ML (a component of AI) faces the same issue today as it always has, outliers, i.e. the real world.

Alphago, and similar, are all examples of ML systems exceeding their training data and human abilities. It's clearly possible to beat human capability with ML.

>Alphago, and similar, are all examples of ML systems exceeding their training data

Alphago basically generated its own training data (hence, "exceeding" is not correct) because the domain was amenable to a specific form of reinforcement learning. Programming isn't such a domain.

An early version of Alphago was trained on expert human games and did go on to defeat the human champion. Later versions of Alphago are stronger still and developed entirely through selfplay to become superhuman. I would say these versions did exceed their training data because they started training on basically random games and got increasingly better as they learned. What is that if not exceeding what they started from?

Alphago is proof that ML systems can exceed human abilities in at least some domains. Other domains, like identifying animal breeds/species by sight, I believe ML models are also super human. It's just not correct to say models are limited by their training data to something less than human quality. We have multiple counterexamples already.

>I would say these versions did exceed their training data

Well, this is obviously incorrect because the training data included all the games it played against itself (good luck replicating that in the domain of programming though).

We were not talking about "exceeding human abilities in at least some domains", we were talking about whether ML (DL really) systems were able to "go beyond" their training data - but your deflection and going off point is noted.

Their training data is also part of their output. The training games produced last exceed the earlier games. So, yes, obviously the model can exceed its training data.

If you're trying to argue that the model can't be better than it is, i.e. the last training data it produces, then that's true but tautologically so. ML has conquered some domains, like Go or Chess or birdwatching, and is making progress in others - like programming. There is no reason to assume ML systems can't exceed human ability at programming or any other task.

>Their training data is also part of their output.

More deflection and going off point.

>is making progress in others - like programming

I already explained that programming is not amenable to the approach that allowed alphago to learn from the games it played against itself.

This is already getting pointless and tedious, should in the future know better than to respond to people that are obviously in the "scale is all you need" nonsense camp, nothing meaningful comes out of it.

I should have add "in unstructured interaction with the 'real world'", 'cause that's the context here.

Computers exceeded human chess capacities without any AI as such in 1999 or so. Computer exceeded human addition and multiplication capacities as soon as they were built - that's what they were built for.

>> Computers exceeded human chess capacities without any AI as such in 1999 or so.

To be pedantic, that was without any machine learning. Deep Blue was absolutely considered an AI system at the time. Tree search as used in Deep Blue is also still central to Alpha* family of game-playing systems.

Reply on news.ycombinator.com