Again and again, the name “DeepMind” appears in the media in connection with artificial intelligence (AI). What exactly is behind it and why do industry giants and billionaires like Elon Musk (Tesla, SpaceX), Peter Thiel (Paypal, Facebook), Jaan Tallinn (Skype) and Li Ka-Shing (Horizon Ventures) invest in it – and not least Google, which has owned the company since 2014? The answer is: because DeepMind as an AI seeks and develops a completely different approach to solutions than previous AI concepts have done. The difference lies in the situational flexibility.
The corporate goal of Google DeepMind is “Solve Intelligence”. This open approach differs from more targeted projects such as IBM’s Deep Blue, which aimed to understand and win chess games. Google DeepMind does not deal with pre-formulated goals but has a flexible, therefore, and is therefore suitable for various applications.
Using resources intelligently
Google is already using one of these flexible applications to optimize the use of energy in its data centers. For example, DeepMind analyzes and controls the use of cooling systems, which has already reduced consumption by around 40%. The neuronal networks in combination with a short-term memory act like an artificial memory that records the interrelationships of the different influences on energy consumption and draws independent (optimizing) conclusions from them.
Efficiency in all areas
But also outside the Google world, DeepMind makes headlines. Since February 2016, DeepMind has been working with the National Health Service to reduce both workload and costs at three London hospitals. Google DeepMind was given access to the data of 1.6 million NHS insured and data from the past five years to identify (disease) patterns. In this case, the aim was to use financial resources more efficiently, anticipate risks and secure and improve the overall health care situation.
Success is rewarded
So what is the “intelligence” of the AI? The crucial point is that AI actually learns by (randomly) trying out actions and activities, whereby success is rewarded after feedback loops. Such actions are applied and refined more often – and over time more efficiently. It goes without saying that far-reaching projects such as project management for hospitals or energy management for data centers do not start from scratch. However, the results prove that Google DeepMind can actually react flexibly and independently to changing conditions.
The connection between “general intelligence” and “artificial intelligence” lies in so-called meta-learning. Every human being can confirm that our world and our environment have a specific characteristic: We never experience the same situation twice. But we also never experience a completely new situation. Every intelligent system must, therefore, have the ability to use the experiences of the past in order to transfer them to new tasks in an elegant and fast way. This is the same prerequisite for man and machine.
A head start for the future
As our environment and our lives are not getting any easier, AI applications are becoming more and more important. Since AI concepts use the power of powerful data centers, they can run through almost countless scenarios within a very short time in order to arrive at the most effective solution. In contrast to human intelligence, these solutions can be understood emotionlessly and transparently – and thus also tested from the human side. Here lies gigantic potential for the future, because every lead promises profit – that is an approach that every type of intelligence understands.