Why is a wary Elon Musk playing with AI fire?3 min read . Updated: 03 Jul 2018, 10:56 PM IST
While Elon Musk has expressed fears about AI, he is backing its advancement via OpenAI and his firm, Neuralink
Mumbai: AI bots just beat humans at the video game Dota 2. That’s a big deal, because their victory required teamwork and collaboration—a huge milestone in advancing artificial intelligence (AI)," tweeted Microsoft Corp. co-founder Bill Gates on 27 June. That Gates no longer fears AI is evident from the tweet.
What remains unexplained, though, is why Tesla CEO Elon Musk—who has often expressed fears that AI-powered software and robots could rule mankind some day if not reined in and has likened AI’s progress to “...summoning the demon"—backing AI advancements through OpenAI. After all, AI bots are from the OpenAI project, co-founded by Musk.
But first, what is so exciting about five AI bots beating human players at Dota 2? This is not the first time that computers have done so.
A decade ago, International Business Machines Corp.’s (IBM) supercomputer Deep Blue defeated the then world chess champion, Gary Kasparov. In March 2016, Alphabet-owned AI firm DeepMind’s computer programme, AlphaGo, beat Go champion Lee Sedol. On 7 December 2017, AlphaZero—modelled on AlphaGo—took just four hours to learn all chess rules and master the game enough to defeat the world’s strongest open-source chess engine, Stockfish.
The AlphaZero algorithm is a more generic version of the AlphaGo Zero algorithm. It uses reinforcement learning, which is an unsupervised training method that uses rewards and punishments. AlphaGo Zero does not need to train on human amateur and professional games to learn how to play the ancient Chinese game of Go.
Further, the new version not only learnt from AlphaGo—the world’s strongest player of the Chinese game Go—but also defeated it in October 2017.
AI bots, though, will face a bigger challenge when attempting to beat professional players at Dota 2. Published by Valve Corp., Dota 2 is a free-to-play multiplayer online battle arena video game and is one of the most popular and complex e-sports games. The reason: professionals train throughout the year to earn part of Dota’s annual $40 million prize pool that is the largest of any e-sports game.
Dota 2 has been actively developed for over a decade, with the game logic implemented in hundreds of thousands of lines of code. This logic takes milliseconds per tick to execute, versus nanoseconds for chess or Go engines. The game is updated about once every two weeks.
The Open AI bots have already achieved a measure of success. A team of five neural networks—the OpenAI Five—defeated amateur human teams at Dota 2 on 25 June. “While today we play with restrictions, we aim to beat a team of top professionals at The International in August (20 August) subject only to a limited set of heroes," the OpenAI researchers acknowledged on their blog.
OpenAI will host a match with the top Dota 2 players on 28 July as a precursor to the main event. It will be a tough man versus machine battle OpenAI Five plays 180 years worth of games against itself every day, learning via self-play. It trains using a scaled-up version of Proximal Policy Optimization (a new class of reinforcement learning algorithms similar to how AlphaGo Zero works) running on 256 GPUs (graphic processing units) and 128,000 CPU (central processing units) cores.
Regardless of whether the OpenAI bots beat the professional Dota 2 players in August or not, the question is: If AI is getting increasingly powerful, why is Musk backing it through OpenAI and his other company, Neuralink—which is developing ultra high bandwidth brain-machine interfaces to connect humans and computers?
Elon Musk is “an incredibly smart guy" and “amazing entrepreneur", said Andrew McAfee, co-director of the Massachusetts Institute of Technology (MIT) Initiative on the Digital Economy, recently. He added, though, that “he (Musk) is doing the world a huge disservice by going around and scaremongering about [AI and machine learning]".
McAfee is right. Musk can’t have his AI cake and eat it too.