Is Artificial Intelligence Real?

The short answer? Not yet. But Google is probably really, really close.

The short answer? Not yet. But Google is probably really, really close.

This week in an op-ed for BBC, Eric Schmidt, the executive chairman for Alphabet, the company that owns Google, famously said that they are “starting to see real progress”.

And if Schmidt is saying that, we have good reason to believe him.

In 2014, Google acquired Nest Technologies and invested more than $400 million in the UK-based artificial intelligence researching firm, DeepMind.

It’s not surprising that Google is at the forefront of AI research. AI technology would help fuel Google’s advertising and marketing business, as well as play an integral role in mobile services, autonomous cars and even robotics.

But what is AI anyways?

Basically, AI is when machines can manage intellectual tasks at a comparable level to humans. So it’s not only processing, but also analysing, reasoning, planning, and perhaps most importantly, learning.

Basically, AI is when machines can manage intellectual tasks at a comparable level to humans. So it’s not only processing, but also analysing, reasoning, planning, and perhaps most importantly, learning. The ability to communicate with language is also highly relevant.

Behavioral or “bottom-up” AI is when the researches construct systems with very simple behaviors, and showed how each behavior worked in a different context. When it worked, the machine would be “rewarded” with points. So as they are trained, these bottom-up systems can develop behaviors - learn.

And the best part? Behavioral AI is based on real-world neuroscience; artificial neural networks simulate the actual neural networks in the human brain.

All the best tech you can think of - from the Roomba to Siri to Facebook, are all behavioral AI.

What is Deep Learning?

Well, I have discussed what deep learning is before. 

Basically, deep or machine learning is based on a set of algorithms that creates high level abstractions in data using different model structures. So the idea is that with each additional layer of nonlinear processing units, each successive layer that uses the output from the previous layer can be unsupervised. Which means it can analyze (unsupervised) patterns, instead of merely classify (supervised) them.

So, now you see the relation.

So now that Google has DeepMind, what next?

But what are the ramifications for the real world? Well, frankly, the possibilities are endless. But more importantly, the reality is closer than we think.

Well, DeepMind is concerned with the same thing. DeepMind wants these unsupervised systems to learn unlabeled data, and then draw its own conclusions from the data. These systems will then receive feedback from other, supervised systems, which then has a much faster data exchange, because the unsupervised systems can remember.

So what’s the big deal? Well, you keep layering these systems on top of one another, one after the other, and you get programs that can manage massive amounts of unlabeled data, which is exactly what Google does.

In his op-ed, Schmidt took a shot at Apple Music, saying that companies should be able to create algorithms to determine what the users will want to listen to next, instead of artisanally crafting curated music by tastemakers. Tech wise, he might have a point. But numbers wise? Spotify, based on an algorithm has done exceedingly well. But Apple, even though the numbers are still quite early, is also hitting the ground running, with their curation.

So we can at least say that algorithm isn’t the only way to success. But what are the ramifications for the real world? Well, frankly, the possibilities are endless. But more importantly, the reality is closer than we think.