Log

Can AI be creative?

Last month Deep Mind’s latest AI broke another record, winning 5-0 against one of the world’s strongest professional players. This victory forces us to revisit our preconceptions and recognise how creative machine learning has become.

[Originally published on LinkedIn]

Computers are not creative or clever in our daily experience. Microsoft Word stumbles through the programmer’s instructions. And we know that computers are machines which behave exactly as they’ve been programmed, bugs included. Right?

Not quite. Last month Deep Mind’s latest AI broke another record, winning 5-0 against one of the world’s strongest professional players. This victory forces us to revisit our preconceptions and recognise how creative machine learning has become.

StarCraft II is among the biggest and most successful games of all time, with players competing in e-sport tournaments. StarCraft II is more chaotic than most games. Players operate in a constantly shifting universe. At each time step, players have hundred trillion trillion alternative moves (compared with only 400 in chess, for example). For two decades AI struggled to win tournaments against professional (human) gamers. Until DeepMind’s machine learning was used.

As a machine learning agent, AlphaStar was not merely following one IF statement at a time. AlphaStar learned how to play by ‘observing’ the games and by playing against itself. Like human players, AlphaStar has developed its own skills, created its own winning strategy, and used it to beat one of the world’s strongest professional players. Was that not creative?

Yesterday, I was surprised. But today I am speechless. Lee Sedol, one of the world’s top Go players, spoke after having lost 4-1 to AlphaGo, another AI from DeepMind’s labs. Speaking about one particularly striking move the AI made Sedol added: “It’s not a human move. I’ve never seen a human play this move. So beautiful.”

Machine learning find solutions that are increasingly originalsurprising, and beautiful. With it, the idea we have that computers cannot be creative seems increasingly anachronistic.

2 Feb. 2019 Correction: Leo Sedol match results (Thanks to Mel Ó Cinnéide)

Related posts:

HSBC’s New Securities Blockchain Will Control $20B

HSBC will upgrade their blockchain technology to control $20 billion in assets by March 2020. Digital Vault, the blockchain HSBC announced today, will control assets referred to as “private placements”, which are securities bought on private markets. Digital Vault will use distributed ledger technology to “give investors real-time access to their records”.

The Singularity Controversy, Part I

Sapience Technical Report STR 2016-1 Author: Amnon H. Eden ‘The Singularity Controversy, Part I: Lessons Learned and Open Questions: Conclusions from the Battle on the Legitimacy of the Debate‘ informs policy makers on the nature and the merit of the arguments for and against the concerns associated with a potential technological singularity. Part I describes the lessons learned from our investigation of the subject, separating the arguments of merit from the fallacies and misconceptions that confuse the debate and undermine its rational resolution.

Sponsor: The Singularity Lab

We are delighted to announce that The Singularity Lab PC, a small startup in Greece and the owner of Magicpress, has donated-in-kind €612 to Sapience Project. Thank you!

Leave a Reply