DeepMind`s AI Can Now Learn On Its Own

DeepMind`s AI Can Now Learn On Its Own

Google-owned DeepMind`s Go-playing system that uses artificial intelligence can reportedly learn on its own.

Microsoft & Amazon To Link Their Voice Assistants Cortana & Alexa
Artificial Intelligence to Take Over The Banking Sector
Squirro Opens An Office In Singapore To Help AI Development

Google-owned DeepMind`s Go-playing system that uses artificial intelligence can reportedly learn on its own.

The first version of AlphaGo had to be trained on data from more than 100,000 human games. However, AlphaGo Zero can learn to play without any training. Moreover, Zero is also able to learn faster than the older version of the software.

“The most striking thing for me is we don’t need any human data anymore.“, according to Demis Hassabis, the CEO and co-founder of DeepMind, who spoke to Wired.

As David Silver,  the main programmer on DeepMind’s Go project, said, the original AlphaGo defeated the world champion Lee Seadol after several months of training, while AlphaGo Zero needed just 72 hours of training to show the same level of performance. This achievement was discussed in the research called “Mastering the game of Go witout human knowledge”.

This is said to be a step forward when it comes to working on the “long-standing goal of artificial intelligence”, as the research paper stated.

“Our results comprehensively demonstrate that a pure reinforcement learning approach is fully feasible, even in the most challenging of domains: it is possible to train to superhuman level, without human examples or guidance, given no knowledge of the domin beyond basic rules.”, as the research concluded.

However, the challenge that remains is to apply this to other problems and situations in areas like medicine,  drug discovery, chemistry, and others.  Using what has been learned from the game of Go in real-world scenarios is still difficult. For example, the system can`t be used to translate language, as Satinder Singh,  the director of the University of Michigan’s AI lab, told Wired.

Future will show whether and how artificial intelligence advancements can be implemented in different areas to deliver better services to their users.

COMMENTS

WORDPRESS: 0