Google's DeepMind AI learns like a human to overcome 'catastrophic forgetting'

Posted March 17, 2017

While this may not sound particularly impressive, it represents a major step forward for AI and brings us one step closer to the holy grail/terrifying existential threat of artificial general intelligence (AGI), sometimes referred to as strong AI.

"When a new task is introduced, new adaptations overwrite the knowledge that the neural network had previously acquired", the DeepMind Google team said on its blog.

"We do not know - and have no power to find out - what Google and DeepMind are really doing with NHS patient data, nor the extent of Royal Free's meaningful control over what Google and DeepMind are doing", it points out, and notes that "any assurances about use of the dataset come from public relations statements, rather than independent oversight or legally binding documents". The inability to do that has been a key failure in machine learning. To test its effectiveness, the researchers used a programme called Deep Q-Network (DQN), which had previously mastered a series of Atari games from scratch without being taught the rules.

Their work was published February in the Proceedings of the National Academy of Sciences of the United States of America. This new formula gives importance to all the input events the machine perceives and, in turn, is able to remember and learn about them simultaneously.

They were able to show that it is possible to overcome "catastrophic forgetting" and train networks that can "maintain expertise on tasks that they have not experienced for a long time". One program resorted to violent tactics as a way of a gaining a competitive edge over its opponent.

"The greed motivation reflects the temptation to take out a rival and collect all the apples oneself."
Leading AI academic Nick Bostrom said in 2016 that he believed Google is leading the global race to develop human-level artificial intelligence.

Philosopher Nick Bostrom at the University of Oxford argued that developers would have to focus on how to establish ethical oversight over AI research projects that take humans into an uncertain future where AI will play increasing roles in human affairs.

This is comparable to Hitler focusing on the "Jewish problem" obsessively and obtaining the "final solution" which involved committing the crime of genocide against an entire ethnic group.

Neural networks, software which is loosely based on the structure of synapses in the human brain, are considered the best machine learning technique for language translation, image classification and image generation. After several days playing each game, the computer was as good as a human player at an average of seven of them.

"The creation of AI with the same powerful learning and planning abilities that make humans smart will be watershed moment".