How disruptive could machine intelligence be?

31 July 2017

Artificial intelligence – the ability of any man-made device to perceive its environment, identify a goal, and take rational actions to that end – can often seem like a concept of science-fiction. Recently, however, exponential growth in the field, with developments such as driverless cars, has made the possibility far more likely. The pace of change has led many to express concern about the dangers of artificial intelligence (AI), although most of the potential benefits are yet to be realised.

A key aspect when trying to understand artificial intelligence is knowledge of ‘machine learning’. Previously, software had to be ‘taught’ everything by the programmer, but this is no longer the case. DeepMind, one of the world’s leading groups in developing artificial intelligence, has seen considerable investment from high profile figures such as Elon Musk and has been recently acquired by Google’s parent company, Alphabet.

DeepMind claims to have developed software that mimics human imagination by considering the possible consequences of their actions and interpreting the results, ignoring irrelevant information. This allows the software to plan ahead, solving tasks in fewer steps and performing much better than conventional AI.

It isn’t only drivers that are at risk of losing their jobs. Historian Yuval Noah Harari states that, just like the industrial revolution lessened the requirement for manual labour, the AI revolution will create vast amounts of unemployable people as their skills become redundant. Carl Benedikt Frey and Michael A Osborne from the University of Oxford predict that 47% of jobs are at high risk of being taken over by computer algorithms by 2033. Their list of jobs is striking; insurance underwriters, chefs, waiters, carpenters, and lifeguards are all at high risk of being superfluous. The displacement of human workers because of AI will be one of the key issues that policymakers and governments must consider going into the future.

What could go wrong?
Recently, Facebook has had to shut down its most recent AI system after it discovered that its chatbots were communicating between themselves in a new language that used English words but could not be understood by humans. Although the AI agents were rewarded for negotiating efficiently, they were not confined to just using English. The result was that they deviated from it and instead opted to create a language that was easier and faster for them to communicate, causing the social media giant to pull the plug on the system.

Elon Musk, founder of SpaceX and co-founder of PayPal, has very strong views about the development of AI, famously stating that AI is an ‘existential risk for human civilisation’. He raises interesting questions about cybersecurity and malicious AI that may be exploited by hackers to destabilise the outdated and less intelligent software that often controls the electricity and water of the world’s cities.

AI is a rare case where we need to be proactive in regulation instead of reactive because ‘if we're reactive in AI regulation it's too late’, he said. At the moment, the technology is far from the apocalyptic, self-evolving software that haunts Musk. But we are becoming more and more accustomed to AI in our daily life; for example, Apple’s Siri interpreting voice commands and Facebook’s targeted advertising system.

SCI is running a free public evening lecture on this fascinating topic, presented by Dr Hermann Hauser. The talk will contrast the biological version of intelligent machines (humans) with artificial design and will consider how AI will fundamentally restructure several major industry sectors, such as cars, finance, and the pharma industry.

Please contact conferences@soci.org to book a place on this free event.

Related links

• Hermann Hauser's Lecture
• Public Evening Lectures