The DeepMind Ethics Council was carefully guardedsecret since Google acquired this artificial intelligence company in 2014. However, the Halist Hodson of The Economist sheds light on how the ethics council came about and why it is needed. He writes that DeepMind and Google signed an “Ethics and Security Audit Agreement”, drafted by lawyers in London a year before the company's purchase.
What does the DeepMind Ethics Council do?
This agreement states that if DeepMindwill ever succeed in its primary task of building a general artificial intelligence (AGI) - which sometimes means a machine that can successfully perform any intellectual task that a person can perform - control of this machine will belong to those who are on the governing council - the Council on ethics.
Ethics Council, in fact, allows DeepMindlegally maintain a certain level of control over the technologies it creates, no matter how valuable or dangerous they become.
It is also worth noting that a machine with general artificial intelligence may not want to listen to those who "control" it.
Despite repeated requests, DeepMind did not disclose who is on the Ethics Council, but The Economist names all three founders of DeepMind (Demis Hassabis, Mustafa Suleiman and Shane Legg).
In December 2016, Suleiman announced atthe TechCrunch conference that “the ethics council is working and it has to monitor a number of our initiatives from within”. He added: “We have always said that he will be largely focused on full-fledged general-purpose learning systems, and I think this is a very, very distant prospect. We are at a distance of dozens of years from the risks that the council originally envisaged. We are implementing many mechanisms in place that focus on the immediate consequences. ”
The year before, at a conference held at the London office of Bloomberg in 2015, Suleiman said: “We will announce the names, but that's not all. This is one of the components of the whole apparatus. ”
DeepMind is often called one of the leaders of the AI race, along with Facebook, Microsoft, OpenAI, Baidu and several other companies.
But for now, artificial intelligence systemsable to win people only in some games, as well as perform other niche tasks in certain areas. It’s impossible to call them omnipotent and omnipotent, and it will be many more years before the situation changes.
DeepMind did not respond to comments. What do you think? Tell us in our chat in Telegram.