Some may think it's nothing to worry about, but it is true that a self-learning, intelligent machine will not be perfect. Even if they may not turn evil and want to take over the world, chances are there will be times when these AI agents make harmful decisions. This is why Google is creating a safe way to interrupt their actions. Google-owned company DeepMind specializes in AI research and is in charge of this study. Think of it as a kill switch of sorts.

Google has teamed up with scientists at the University of Oxford to ensure that its AI, DeepMind, won't be able to overpower human control by creating, what is essentially, an off switch. Revealed in a peer-reviewed paper titled'Safely Interruptible Agents', the framework allows a human operator to safely interrupt an AI without the AI learning how to prevent or induce interruptions. "If such an agent (AI) is operating in real-time under human supervision, now and then it may be necessary for a human operator to press the big red button to prevent the agony from continuing a harmful sequence of actions - harmful either for the agent or for the environment - and lead the agent to into a safer situation." Speaking to Business Insider, Google DeepMind scientist and one of the paper's writers, Laurent Orseau, explained: "If the agent expects a reward but can predict it's going to be shut down. It will try to resist so as to get its reward.

Machines are becoming more intelligent every year thanks to advances being made by companies like Google, Facebook, Microsoft, and many others. AI agents, as they're sometimes known, can already beat us at complex board games like Go, and they're becoming more competent in a range of other areas. Now a London artificial-intelligence research lab owned by Google has carried out a study to make sure that we can pull the plug on self-learning machines when we want to. DeepMind, bought by Google for a reported 400 million pounds -- about 580 million -- in 2014, teamed up with scientists at the University of Oxford to find a way to make sure that AI agents don't learn to prevent, or seek to prevent, humans from taking control. The paper -- "Safely Interruptible Agents PDF," published on the website of the Machine Intelligence Research Institute (MIRI) -- was written by Laurent Orseau, a research scientist at Google DeepMind, Stuart Armstrong at Oxford University's Future of Humanity Institute, and several others.

DeepMind, Google's artificial intelligence company, catapulted itself into fame when its AlphaGo AI beat the world champion of Go, Lee Sedol. However, DeepMind is working to do a lot more than beat humans at chess and Go and various other games. Indeed, its AI algorithms were developed for something far greater: To "solve intelligence" by creating general purpose AI that can be used for a host of applications and, in essence, learn on their own. This, of course, raises some concerns. Namely, what do we do if the AI breaks…if it gets a virus…if it goes rogue?

Google's DeepMind artificial intelligence has started working with the doctors and computers at London's Royal Free Hospital Trust; many of the first questions being asked concentrate on data security and patient confidentiality. About 1.6m patients a month are seen by the three hospitals that the Trust covers and, according to an agreement seen by the New Scientist magazine, the data includes patient information from the past five years. That data can't be used by Google in any other part of its business, it is stored by a third party and will have to be destroyed/given back when the agreement ends, according to the New Scientist. Google wants to apply machine learning to mine the data and help doctors by predicting care needs – both at the level of the individual patient and the wider community. If the pilot works then implications are huge: clever new algorithms will help make healthcare better and be more efficient at treatment. Artificial Intelligence (A.I.) could predict the outbreak of infectious diseases from scattered hospital admissions data before a human expert has noticed.

It seems to me it is just multi-task learning with online training, and with examples shuffled at every iteration. BTW, they mentioned that labels are shuffled from dataset-to-dataset. Is this the same as shuffling examples?

You've probably heard about OpenAI -- a new nonprofit to focus on artificial intelligence research that is good for humanity. So here we have a case of knowledgeable people recognizing a threat and deciding that the way forward is to create a nonprofit organization. This moment has some historical precedent. In 1955, Albert Einstein, Bertrand Russell and several other scientists got together and issued a manifesto about the dangers of nuclear technologies. It would launch decades of Pugwash conferences and the anti-nuclear movement. The founders of OpenAI also issued a manifesto.

DeepMind, an artificial intelligence company owned by Google, has attempted to justify why it needs access to millions of NHS patient records for a kidney monitoring app, after a new investigation from New Scientist questioned whether an ethical approval process should have been obtained first. The AI research lab, acquired by Google in 2014 for around 400 million, signed a data-sharing agreement with the Royal Free London NHS Foundation Trust on 29 September 2015. The agreement gives Google DeepMind access to the names, addresses, and medical conditions of the 1.6 million patients that are treated at Barnet, Chase Farm, and the Royal Free hospitals each year, as well as data on all patients treated by the Trust in the past five years. This week, New Scientist questioned why Google DeepMind needs access to so much data on so many people, including those who have never experienced kidney problems, for the app, which is called Streams. Streams -- used by Royal Free clinicians in three separate trials since December 2015 -- is designed to detect acute kidney injury (AKI), a condition that kills more than 1,000 people a month.

Gym-TORCS is the reinforcement learning (RL) environment in TORCS domain with OpenAI-gym-like interface. TORCS is the open-rource realistic car racing simulator recently used as RL benchmark task in several AI studies. Gym-TORCS is the python wrapper of TORCS for RL experiment with the simple interface (similar, but not fully) compatible with OpenAI-gym environments. The current implementaion is for only the single-track race in practie mode. If you want to use multiple tracks or other racing mode (quick race etc.), you may need to modify the environment, "autostart.sh" or the race configuration file using GUI of TORCS.

Three weeks ago, New Scientist revealed that Google's artificial intelligence company DeepMind has access to the identifiable personal medical information of millions of UK patients through a data-sharing agreement with the Royal Free London NHS Foundation Trust. Now, a New Scientist investigation has found that Google DeepMind deployed a medical app called Streams for monitoring kidney conditions without first contacting the relevant regulatory authority. Our investigation also asks whether an ethical approval process that covers this kind of data transfer should have been obtained, and raises questions about the basis under which Royal Free is sharing data with Google DeepMind. DeepMind's partnership with the Royal Free provides it with fully identifiable information – including names, addresses and details of medical conditions – for the 1.6 million patients treated at Barnet, Chase Farm and the Royal Free each year. It also includes complete data on all patients treated by the trust in the past five years.