The use of machine learning and computer intelligence is becoming increasingly important in emerging technology fields, from big data to self driving cars. However, Mustafa Suleyman, CEO of AI company DeepMind which Google bought back in 2014, still believes that we’re still a long way away before general artificial intelligence becomes a reality.
Speaking at the Disrupt London event today, Mustafa suggests that computer learning technologies will continue to focus on solving specific problems in the immediate future, while technology presumable improves processing and compute capabilities. He also raised a point about developers potentially introducing their own judgement flaws and biases into machine learning algorithms, and that designers need to take a new approach when building these new systems to avoid human error.
According to the talk, Mustafa believes it could be 20 years or so before general AI capable of broader understanding emerges, and even then he believes that it won’t bear much resemblance to the type of AI that we’re used to seeing in films. At least not based on the research that DeepMind is undertaking.
“when we say it’s 20 years out or decades way, we say that it so far out, we can’t really measure it … When it comes to imagining what the future will be like, a lot of that is fun and entertaining, but it doesn’t bear a great deal of resemblance to the systems that we are building,” – DeepMind CEO, Mustafa Suleyman
Those worried about a post-apocalyptic future ruled by computer overlords won’t take much comfort from Mustafa’s refusal to name those who sit on Google’s mysterious AI ethics board. He has previously stated that he wants to disclose details to the public. When pressed on the issue, Mustafa suggested that we’re still decades away from the kinds of risks that the board initially envisages, and he repeated that DeepMind’s mission statement is to “solve intelligence and make the world a better place.”
On the more productive side, DeepMind has also just opened up its Lab platform to researchers. It features a 3D game-like engine tailored for agent-based AI research. This allows researchers to test navigation, memory, and other strategies in a 3D environment. Code for the DeepMind Lab platform is going to be published to Github later this week.