Google's DeepMind is a company working on what could be one of the most significant scientific advances of our era. They focus on the development of AI systems that could have a broad range of applications from cybersecurity to healthcare. The primary goal of DeepMind is to develop artificial intelligence (AI) that can solve the most complex problems without even first having to be taught how.
A major training/proving ground for DeepMind's software has been gaming. The company has developed AI that can play 49 different Atari games at expert levels. Also, in a world-first development, they created software called AlphaGo which challenged the world champion of the ancient Chinese game of Go, and won.
The latest news out of DeepMind comes back to that historic victory. As a means of testing some upgrades to AlphaGo, the company secretly unleashed the AI on some unwitting Go players. AlphaGo completely dominated the competition. More than 50 games were played and AlphaGo won every single one.
Demis Hassabis, the co-founder and CEO of Google DeepMind came clean yesterday that AlphaGo was behind the juggernaut. "We've been hard at work improving AlphaGo, and over the past few days we've played some unofficial online games at fast time controls with our new prototype version, to check that it's working as well as we hoped." So, if you happened to get bested while playing Go against players named Magister or Master, don't feel too bad.
The development of AI that can tackle huge issues with the same ferocity as AlphaGo takes on the game of Go may not be too far off. Even just looking at DeepMind, and not other companies who are making major strides in the field of AI, there is some very impressive and awe inspiring work being done.
Even the processes by which the systems are being built to learn are fascinating. For example, DeepMind is taking a page out of the Westworld playbook by teaching its software "to dream." Video games are also being deployed in this area to create auxiliary tasks, which ended up boosting the rate at which the software learns. Deepmind is also teaching AI how to learn similarly to humans and other living organisms – software is being made able to learn by doing and having that reinforced with a system of rewards and punishments.
The applications of software being created at the company are already causing much excitement. An AI was able to teach itself to identify a unique condition that could lead to blindness, much like a well-trained eye doctor. DeepMind has also partnered with the National Health Service (NHS) in the UK to work together on transforming healthcare.
AI advancements are not strictly being made in the field of healthcare. Google has employed AI to even help them cut down on their utility bills.
AI has already come a long way and there is no end in sight to the benefits it will likely have in store for the future of humanity. The rapid pace of discovery could mean that such a future is right around the corner.