Video Games Are The Fruit Fly Of AI Research

Video games are the fruit fly of artificial intelligence research. There's a fair amount to unpack from that💯 statement, so let's just dive right in.

Artificial intelligence - or AI - is undoubtedly a buzzphrase (haha, because flies) at the moment, and it's only gaining more momentum as research grows. As that research grows and more ꦓpeople recognize its potential, it's probably not too outlandish to think AI will be wr🐲iggling its way into more and more corners of our daily lives.

Within this field, researchers and AI developers alike are continuously looking to create AI systems - or "agents" - that are as capable and intelligent as pos💦sible, an endeavor referred to as the ‘Policy Problem.’ According to by some researchers at DeepMind - Google's AI company - describes this concept further:

In short, the ‘Policy Problem’ entails the search for (super) human-level AI behavior in the vast space of all possible strategies (also known as policies in the reinforcement learning literature) that can be learned 🌳by an artificial agent.

💞And in order to do this, researchers are tappin🦋g into the glory of video games.

But Why A Fruit Fly?

Here's a brief biology/genetics/neur꧋oscience lesson - bear with me. You may have come across that "cloning fruit flies" phrase which often crops up in pop culture. I for one never questioned wh﷽y it was a fruit fly of all things, but it turns out they occupy a major role in scientific research.

Commonly referred to by their scientific name Drosophila melanogaster, these tiny little guys serve as a really useful experimental model for a . This includes human research, thanks to the surprising fact that there are a fair amount of genetic and anatomical similarities between us and ♛these flies. As such, they - along with mice - have become the established go-to as waysཧ of studying human conditions, thereby strengthening our understanding thereof.

Yet Another Reason Why Video Games Are Great

So back to games. Much like fruit flies have been established as a basic model with which we can explore a range of deeper, larger (literally and figuratively) matters, video games have become of further exploring the capabilities of artificial intelligence agents. As the put it:

We saw early success in computer games, which researchers often use to test AI. One of our programs learned how to play 49 different Atari games from scratch just from seeing the pixels and score on the screen, and our AlphaGo program was the first to beat a professional player at the ancient game of Go, a feat experts described as a decade ahead of its time.

[caption id="attachment_725967" align="alignnone" width="1710"]

(Via: Medical Daily)[/caption] 168澳洲幸运5开奖网:We recently co💧vered the story of Agent57, a DeepMind AI agent who - in just four hours - went from being totally incompetent at playing a bunch of Atari games, to performing better than humans in a bunch of Atari games. This particular collection of 57 Atari games are known in the field as , and is a "long-standing benchmark to gauge [AI] agent performance across a wide range of tasks."

The Significance Of Chess & Go

If you follow the trail of AI research and the role played therein by games back to the 1960s, you'll see that good ol' fashioned chess was actually considered for quite some time. , much like the fruit fly, chess is:

an accessible, familiar, and relatively simple experimental technology that nonetheless can be usedܫ productively to produce valid knowledg💯e about other, more complex systems.

The ancient Chinese board game called Go is widely considered to be one of the - apparently having - and as such is a great test for AI, too. In 2016, the AI agent AlphaGo - developed by DeepMind - beat a legendary Go master in South Korea, "the first computer program to beat a professional human Go player, the first to defeat a Go world champion, and is arguably the strongest Go player in history." Not bad. [caption id="attachment_726041" align="alignnone" width="1710"]

(Via: Twelvesec)[/caption]

StarCraft & Dota 2

Other complex games of a different ilk have also been singled out as the perfect playgrounds wherein potential AI agents can flex their algorithms. Both Dota 2 and 168澳洲幸运5开奖网:StarCraft 2 are two such games, the latter making the headlines in the AI field thanks to the incredible feats performed by DeepMind's agent AlphaStar. AlphaStar has managed not only to defeat professional StarCraft 2 players but also As for Dota 2, the AI agent OpenAI Five (developed by Elon Musk's AI company, OpenAI) - the Dota 2 world champions of 2019 demonstrating superhuman performance in the game. It learnt to do so by as one does. The use of games as testbeds within AI research has been a thing for decades, but the most exciting thing is that there's no sign of these fascinating efforts slowing down any time soon. Just when you thought video games couldn't get any cooler.

Source: