Those who thought that day would be sometime in the far distant future would be wise to think again. To be sure, DeepMind has made inroads on this goal before, specifically with their work on Psychlab and Differentiable Neural Computers. However, Impala is their largest and most successful effort to date, showcasing a single algorithm that can learn 30 different challenging tasks requiring various aspects of learning, memory, and navigation.

But enough preamble; let’s look under the hood and see what makes Impala tick. First, Impala’s based on reinforcement learning, an AI technique that has its origins in behaviorism. It parallels the way humans build up an intuition-based skill, such as learning to walk or riding a bicycle. Reinforcement learning has already been used for some amazing achievements, such as endowing an AI with emotions (see video below) and learning complex games like Go and Poker.

Eoin Treacy's view

Relational learning is a big step for artificial intelligence because unlike humanity, computers have no limit on the number of relationships they can form and how much experience can be relied on to draw conclusions.

Of course, that also creates the very real potential for issues with limited data being used to draw general conclusions. The problem Facebook had with its AI being unable to differentiate between people who are not white is an example of that issue. I would consider that the teenager problem. I see it with my 12-year old who thinks she knows a lot about the world based on limited experience and can’t be convinced otherwise.

Making mistakes and learning from them is about the most human of all characteristics and it is that mannerism which general artificial intelligence is solving for. This is a major innovation for Google’s Deepmind and is likely to find its way into search engines before long.

Alphabet/Google broke successfully upwards to new highs in July and has been consolidating above $1200 since. The pullback in February was the largest in at least the last decade and represented an inconsistency for what had been one of the most consistent anywhere. If the damage is to be repaired then the share needs to hold the move above the trend mean.