Artificial Intelligence Will Do What We Ask. That’s a Problem.

[cguy]

How can an algorithm have "emergent behaviour" that is not intrinsic to that algorithm?

The behavior of Machine Learning algorithms is determined by both the algorithm and the data that is used to train it. The level of learning and types of behavior that the algorithm will learn from the data is not explicitly coded anywhere, and can be very surprising and complex.

 

[ToxicBunny]

It seems to fit you quite well.
Play the ball, and not the player.

Problem is that its hard to just play the ball when the player is monumentally out of their depth.

 

[Ponderer]

The behavior of Machine Learning algorithms is determined by both the algorithm and the data that is used to train it. The level of learning and types of behavior that the algorithm will learn from the data is not explicitly coded anywhere, and can be very surprising and complex.

The behaviour of every/all programs is determined by data.
A "machine learning algorithm" is an algorithm - its a particular set of instructions the machine executes.
A "machine learning algorithm" does not and cannot "learn" - its just an algorithm.

It is silly to argue that the level of learning and behaviour is not "explicitly coded anywhere".

 

[cguy]

The behaviour of every/all programs is determined by data.

But not all programs are trained by data.

A "machine learning algorithm" is an algorithm - its a particular set of instructions the machine executes.
A "machine learning algorithm" does not and cannot "learn" - its just an algorithm.

It's an algorithm that learns. Hence, why it is called machine learning.

It is silly to argue that the level of learning and behaviour is not "explicitly coded anywhere".

It's silly to argue that it is when you so clearly know very little about it.

 

[Ponderer]

But not all programs are trained by data.



It's an algorithm that learns. Hence, why it is called machine learning.



It's silly to argue that it is when you so clearly know very little about it.

Not sure what it is that you don't understand.
A "machine learning algorithm" is nothing other than an algorithm.
A machine cannot learn - a machine can only execute an algorithm.

 

[cguy]

Not sure what it is that you don't understand.
A "machine learning algorithm" is nothing other than an algorithm.
A machine cannot learn - a machine can only execute an algorithm.

A machine can learn, via an algorithm that learns. You seem to have a limited understanding of what algorithms are capable of.

 

[cguy]

Adapts based on data is probably a purer way to say it, but the industry uses the term "learn". ML is based on neural networks which are based on how scientists think our brains work.

Neural networks are common, but there are many other techniques, such as linear regression, support vector machines, regression trees, Bayesian networks, Markov models, combinations of these, etc.

I would say “learn” is a pretty accurate description. All learning is essentially adapting to data.

 

[cguy]

The difference is its built in to our DNA but computers are programmed. But we're having a philosophical discussion here, not technical. It's about whether we should use the pure definition of the term learn or not.

I don’t think that “built into DNA” is in the pure definition, so it works.

 

[Ponderer]

A machine can learn, via an algorithm that learns. You seem to have a limited understanding of what algorithms are capable of.

The machine "will learn" how/what it is instructed to "learn" via an algorithm.

 

[cguy]

Ergo you are of the camp who are cool with saying "learn". Most people are. You're in good company, I'm just pointing out that it's not a winnable argument, it's a point of view.

Honestly, this is the first time I’ve heard of there being two camps (apart from Ponderer vs everyone else).

 

[cguy]

The machine "will learn" how/what it is instructed to "learn" via an algorithm.

That’s not how it works.

 

[Ponderer]

That’s not how it works.

That is exactly how it works.

 

[cguy]

That is exactly how it works.

Well, if you say so, it must be true. Well done.

 

[cguy]

Yet you agreed with him?
Seems were on the same page then.

The real disagreements are in the AGI Vs narrow AI camps.

Well, subject to a large amount of sarcasm, sure.

The actual consciousness argument is indeed a lot more involved. Personally, I don’t think that any AI in existence is conscious, but I do believe that obtaining consciousness is just a matter of scale.

 

[ToxicBunny]

That is exactly how it works.

Hold on... you're actually trying to tell someone who does this sort of thing for a living how it works? Thats pure comedy gold that is.

 

[Ponderer]

Hold on... you're actually trying to tell someone who does this sort of thing for a living how it works? Thats pure comedy gold that is.

What "sort of thing" you referring to?
Programming?

The reason for your post.
You disputing anything I said, or you just making a noise.

 

[ToxicBunny]

What "sort of thing" you referring to?
Programming?

The reason for your post.
You disputing anything I said, or you just making a noise.

Cute son, very cute...

I'm guessing you have zero clue what "programming" entails... or algorithms, or AI, or anything complicated in these fields.

 

[cguy]

What "sort of thing" you referring to?
Programming?

The reason for your post.
You disputing anything I said, or you just making a noise.

He’s referring to my 15 years of experience building state of the art, large scale machine learning technology.

 

[ToxicBunny]

He’s referring to my 15 years of experience building state of the art, large scale machine learning technology.

Only 15 years?
Oh I thought you were actually an expert or something

 

[Ponderer]

Cute son, very cute...

I'm guessing you have zero clue what "programming" entails... or algorithms, or AI, or anything complicated in these fields.

Your reasoning = guessing?
Wow - fantastic stuff dude.