Mitchell Barrington Center for AI Safety University of Michigan University of Southern California Abstract This paper argues that training AI systems with absolute constraints--which forbid certain acts irrespective of the amount of value they might produce--may make considerable progress on many AI safety problems in principle. First, it provides a guardrail for avoiding the very worst outcomes of misalignment: An AI attempting to commit mass murder might have correctly deduced that doing so maximizes expected value, but more likely, the system is severely misaligned. Second, it could prevent AIs from causing catastrophes for the sake of very valuable consequences, such as replacing humans with a much larger number of beings living at a higher welfare level. Third, it makes systems more corrigible, allowing creators to make corrective interventions in them, such as altering their objective functions or shutting them down. And fourth, it helps systems explore their environment more safely by prohibiting them from exploring especially dangerous acts. I offer a decision-theoretic formalization of an absolute constraints, improving on existing models in the literature, and use this model to prove some results about the training and behavior of absolutist AIs. I conclude by showing that, although absolutist AIs will not maximize expected value, they will not be susceptible to behave irrationally, and they will not (contra coherence arguments) face environmental pressure to become expected-value maximizers. Introduction Advanced AI systems are expected to be dangerous because of the opacity of their goals: We may know that they will effectively pursue their goals but fail to know what those goals are.

The growth of artificial intelligence and machine learning will create millions of new jobs over the next few years. Artificial intelligence (AI) and machine learning (ML) specialists are in the most "in demand" employees across all industries, according to the World Economic Forum's new Future of Jobs report, 2020, yet there are only a few hundred thousand engineers trained in AI and ML worldwide. As more organizations embrace digital transformation by moving to the cloud to make machine learning a reality, they're using creative training initiatives to close the machine learning skills gap. Capital One's AWS DeepRacer League is one excellent example of this. As an organization, Capital One has been on a multi-year journey to transform into a technology company that also happens to be a bank, instead of a bank that uses technology.

Machine-learning techniques, best known for teaching self-driving cars to stop at red lights, may soon help researchers around the world improve their control over the most complicated reaction known to science: nuclear fusion. Fusion reactions are typically hydrogen atoms heated to form a gaseous cloud called a plasma that releases energy as the particles bang into each other and fuse. Getting these reactions under better control could create huge amounts of environmentally clean energy from nuclear reactors in fusion power plants of the future. "The connection between machine learning and fusion energy is not obvious," said Sandia researcher Aidan Thompson, principal investigator for a $2.2 million, three-year DOE Office of Science award to make that connection. "Simply put, we have pioneered machine-learning's use to improve simulations of the reactor's wall material as it interacts with the plasma. This has been beyond the scope of atomic-scale simulations of the past."

Engineers have developed a robotic shopping trolley for elderly customers based on a drawing by a boy who simply wanted to help his grandmother. Aidan McCann, 13, dreamed up a push cart with height adjustment features to help his grandmother Lydia who'isn't very strong'. He witnessed how the 4ft 11in 76-year-old finds it difficult to carry groceries from the shops and perform other physically demanding tasks due to her height. Bosses at engineering giant Doosan Babcock were so impressed with Aidan's design they selected it as their overall winner at the Scottish Engineering Special Leaders Award 2015 Users can make the trolley go up and down by the flick of a switch. The idea is to lift bags of shopping towards the users so that they don't have to bend down and lift it up themselves.

According to experts, 2016 may finally be the year that artificial intelligence comes into its own -- not in the science fiction "robots will take over humanity" sense, but in a much more practical and useful way. AI is already excellent at problem-solving -- when it comes to finding patterns, it can usually solve a problem much faster than its human counterpart. For the most part, though, AI still very limited in scope, and the dreams of a general intelligence are still far off. To some, AI being able to execute nearly any task that humans can perform today may sound like a worst-case scenario. But, in actuality, this future will bring about a new era of creativity and innovation.