Machine Learning on Twitch, the Next Big Thing?

The first instance of AI/Machine Learning (ML) that I came across was Salty’s Dream Cast Casino on Twitch.  This channel and it’s accompanying website SaltyBet allows viewers to earn fake currency  to bet on fighting game matches.  The matches are unique in that they’re powered by the M.U.G.E.N engine that allows importing of assets from various fighting games (such as characters and background locations) and let the engines AI battle it out.  This channels been running 24/7 since it’s creation back which looks to be in 2013.

Example of MUGEN from Twitch.tv/saltybet
Screen shot of PacMan vs Terra (Final Fantasy VI) from Twitch.tv/SaltyBet

CHATBOTS AND DRIVING AROUND LOS SANTOS

In 2017 for about two weeks in January, most of Twitch was glued to their screen watching two Google Homes talk, express their love, fight, and inspire everything from memes to fan art. You can read more about them and OpenAI’s efforts in Machine Learning in video games in my post here.

Before jumping into the latest ML project that’s come up on Twitch (which has an open source framework available to play with).  There was Charles the Python powered AI that was learning how to drive through the streets of Los Santos.

Sadly as of this post being published, it looks like Charles hasn’t been live for a while.  You can still check out the collection of clips on the channel or follow his creator Sendex on Twitter.

THE LATEST IN MACHINE LEARNING TRAINING STREAMS

Finally, the project that inspired me to write this post Serpent AI.  This seems to be a project inspired by the shortcomings and limits of the OpenAI project as described in the introduction post by Nicholas aka Serpent (Twitter).

Screen shot from the Serpent AI Labs Twitch channel
Screen shot of Serpent AI in action from the Serpent AI Labs Twitch channel

Serpent AI is open source, a framework that runs on Windows, Linux, and OS X (I refuse to call it macOS), and is designed to work with Steam, games in web browsers, and independent executables.

You can read more about Serpent IA at their website and Github repository.

NEXT

Now, this got me to thinking, is there an audience in broadcasting content of training AIs or even having AIs performing in the name of creating content.  While the training of Charles was broadcasted, there was a sense of waiting to see what happens next or what crazy unexpected event will occur while no one was at the preverbal wheel.  There’s another aspect to showing the general public just what AI, Machine Learning, and this “black magic” is that we as a society are adopting more and more in our daily lives.

As much information as Tesla, nvidia, Google, and other companies talk about the progress of their autonomous car projects.  Seeing first hand the technologic concepts and tools used  to “teach” a computer to “think” gives one a more concrete understanding of just what’s happening when these systems are taken from a virtual environment such as in a video game and applied to the real world.

Have you come across other projects or interesting projects being streamed live online?  Leave your thoughts if this is the future of entertainment, projects you’ve seen, and related links in the comments below.

Click to post a comment with your Twitter or Facebook account