Why did it take so long for machine learning to take off?

Question

I hope this questions doesn't offend anyone.

I know that there are very advanced and complex algorithms in machine learning I'd probably never fully understand. However it seems to me that the concept of machine learning as such is actually quite simple. Basically you classify some things, try to find correlations and make predictions based on that correlation. Even on Sololearn there are working example codes that fit on a mobile screen.

PCs have been around for 40 years now, the internet for 25, people invented much more complex things than machine learning algorithms (as far as I can judge). Why did it take such a long time for machine learning to take off?

Answer 1

Maybe the fact that it is only recently that large masses of people started using smart devices and getting online. So today there's a large online global consumer base for information, products and services. This also means that there's a massive amount of data/information that global companies like Google, Amazon and Facebook can utilise and profit from. You probably need this critical mass of information, consumers and resources to make the end product of using AI and ML to become profitable to organisations. That's probably why it's only recently that AI/ML has marched out of the research labs and into the mainstream and the global market place. Just a theory....

Answer 2

Many things we use today are allready been tought of many years ago. Most of the time they needed more processing power. Or in case of machine learning processing power and input/training data for the algorythm.

Answer 3

Maybe this podcast can give you some answers. http://omegataupodcast.net/259-maschinelles-lernen-und-neuronale-netze/

It's been a while since I heard the episode. But I think it was also about knowledge linking. e.g. Statistics / Probability etc. Where it is usually needed, there was no interest in machine learning or neural networks. While the computer scientists did not know that they needed it.

Answer 4

Maybe it was there but we didn't knew?

There has been definitely some data science from many years used in business, not sure can be category of ML, though.

See major developments here… https://en.m.wikipedia.org/wiki/Timeline_of_machine_learning

See, many algorithms are old and statistical work is even older.
Internet has been for 25 years, but

1) It was 2G till 3G invented in 2001, distributing and standardising prices took 4-5 years(in developed countries), so speed was issue.

2) So, getting data was hard, see IBM had data so could make Watson. Without having reliable data, there's no application of ML.

3) I know Python is quite old, but not quite sure about its popularity, they're Java days I guess. And all this ___ about using Py and R over others because of their easiness must not be all false.

4) Price: As population grew and PC - internet came in developing countries, …
LOL this is endless, I know I've deviated from the question and I've gone about AI sometimes but ML is part of it BTW…