What is AlphaGo Zero and why should you care?


What is AlphaGo Zero and why should you care?

Rafe Blandford

Towards the end of October, Google DeepMind’s supercomputer AlphaGo Zero mastered the game of Go within 40 days. In doing so it discovered game principles that had taken humans thousands of years to master. All without any human input. Rafe Blandford, Senior Mobile Strategist at Digitas, explains why all this is so important.  

What's happening?

In what has been hailed as one of the biggest recent landmarks in the development of AI, AlphaGo Zero recently taught itself to play the ancient and notoriously complex strategy game Go - without any human input. Previous approaches had depended on huge databases of previously played games to train the AI. The new approach gave it the first principles (rules) then had it play itself millions of time.

Go is considered a difficult game for computers to master because the number of possible moves, many more than chess at 10, is greater than the number of atoms in the universe.

In the case of AlphaGo Zero, it took just 40 days to surpass 3,000 years of human knowledge and what’s more significant it was not constrained by human knowledge. It was able to create knowledge itself from first principles - from a blank slate. 

"It starts from a blank slate and figures out only for itself, only from self-play, and without any human knowledge, or any human data, or features, or examples, or intervention from humans. It discovers how to play the game of Go from first principles," noted DeepMind's professor David Silver. 

Why does it matter for marketers?

The achievements of Alpha Go Zero represent a significant moment for the reinforcement learning approach to building machine learning models. This is because, unlike the more common supervised learning approach, it does not require an existing set of data to train the model. Put another way, Alpha Go Zero provides an elegant example of how it’s possible for a self-learning system to “create knowledge” without relying on past information. It also means that machine learning is opened up to domains where human knowledge may be too expensive, or simply unavailable.

This ability to create knowledge from first principles is especially important with new regulations such as GDPR on the horizon. In a world where consumer data is increasingly heavily regulated being able to build machine learning models without large data sets will undoubtedly be very attractive.

What’s also notable about Alpha Go Zero is the removal of any need for human expertise in the system and that the strategies it came up with were previously unknown - the implication being that Alpha Go Zero was able to discover knowledge that humans were unable to do so.

This is suggestive of a truly creative approach – something that previously didn’t seem possible with AIs trained on existing knowledge – and is something that will particularly resonate with creative and marketing industries, where we are frequently encouraged to free ourselves from previous constraints. Might we start to consider the implications of AI that is quite literally able to create unimaginable? 

What are the implications for the future?

While Alpha Go Zero lives in the rule-based world of games, the approach can theoretically be extended to any application that is based on a fixed set of physical rules from chemistry and biology, to traffic management and logistics. These systems would need no outside data and therefore no data ingestion or structure problems. What’s more there are fewer human derived bottlenecks in the learning process.

These developments have implications for numerous aspects of the digital landscape. Perhaps most obviously we can expect to see AI have a fundamental impact on programmatic advertising in the coming years. Less obviously the creative world could be transformed. With the emergence of applications like Adobe Sensei, users will be able to simply define what’s pleasing and leave AI to come up with ways to achieve that. 

Put simply, AI is set to have a fundamental impact on our industry. The case of AlphaGo Zero is a timely reminder that businesses must be ready to face the future. 

Suggestions for further reading 

AlphaGo Zero: Learning from scratch (DeepMind blog)

The big brains at DeepMind outline what makes AlphaGo Zero such a leap forward.

Mastering the game of Go without human knowledge

The full Nature magazine research paper featuring very detailed explanation of the AlphaGo Zero project.

Rafe Blandford

Rafe Blandford

Chief Product Officer

As Chief Product Officer at Digitas, Rafe leads a blended team of strategy, creative and technology experts, leveraging his extensive mobile knowledge and working with the agency’s clients to help develop mobile strategies and creative solutions.

Prior to joining Digitas in 2014, Rafe was an independent mobile consultant, advising brands including Nokia, Orange and Ericsson on their business strategies. 

Rafe is co-host of mobile technology podcast 361 Degrees, and the founder and editor of the All About Windows Phone and All About Symbian sites, which have a combined reach of more than a million.


What do people really think about AI?

Like what you see?

Have questions? Thoughts? Requests?


Like what you see?

Have questions? Thoughts? Requests?


Make a Connection