Debating AI's Future: Hawking's Prediction

It is not often that you are obliged to proclaim a much-loved genius wrong, but in his alarming prediction on artificial intelligence and the future of humankind, I believe Stephen Hawking has erred. To be precise, and in keeping with physics – in an echo of Schrödinger’s cat – he is simultaneously wrong and right. Asked how far engineers had come towards creating artificial intelligence, Hawking replied: “Once humans develop artificial intelligence it would take off on its own and redesign itself at an ever increasing rate. Humans, who are limited by slow biological evolution, couldn’t compete, and would be superseded.” In my view, he is wrong because there are strong grounds for believing that computers will never replicate all human cognitive faculties. He is right because even such emasculated machines may still pose a threat to humankind’s future – as autonomous weapons, for instance. Such predictions are not new; my former boss at the University of Reading, professor of cybernetics Kevin Warwick, raised this issue in his 1997 book March of the Machines. He observed that robots with the brain power of an insect had already been created. Soon, he predicted, there would be robots with the brain power of a cat, quickly followed by machines as intelligent as humans, which would usurp and subjugate us.

Triple trouble

This is based on the ideology that all aspects of human mentality will eventually be realised by a program running on a suitable computer – a so-called strong AI. Of course, if this is possible, a runaway effect would eventually be triggered by accelerating technological progress – caused by using AI systems to design ever more sophisticated AIs and Moore’s law, which states that raw computational power doubles every two years. I did not agree then, and do not now. I believe three fundamental problems explain why computational AI has historically failed to replicate human mentality in all its raw and electro-chemical glory, and will continue to fail. First, computers lack genuine understanding. The Chinese Room Argument is a famous thought experiment by US philosopher John Searle that shows how a computer program can appear to understand Chinese stories (by responding to questions about them appropriately) without genuinely understanding anything of the interaction. Second, computers lack consciousness. An argument can be made, one I call Dancing with Pixies, that if a robot experiences a conscious sensation as it interacts with the world, then an infinitude of consciousnesses must be everywhere: in the cup of tea I am drinking, in the seat that I am sitting on. If we reject this wider state of affairs – known as panpsychism – we must reject machine consciousness. Lastly, computers lack mathematical insight. In his book The Emperor’s New Mind, Oxford mathematical physicist Roger Penrose argued that the way mathematicians provide many of the “unassailable demonstrations” to verify their mathematical assertions is fundamentally non-algorithmic and non-computational.

Not OK computer

Taken together, these three arguments fatally undermine the notion that the human mind can be completely realised by mere computations. If correct, they imply that some broader aspects of human mentality will always elude future AI systems. Rather than talking up Hollywood visions of robot overlords, it would be better to focus on the all too real concerns surrounding a growing application of existing AI – autonomous weapons systems. In my role as an AI expert on the International Committee for Robot Arms Control, I am particularly concerned by the potential deployment of robotic weapons systems that can militarily engage without human intervention. This is precisely because current AI is not akin to human intelligence, and poorly designed autonomous systems have the potential to rapidly escalate dangerous situations to catastrophic conclusions when pitted against each other. Such systems can exhibit genuine artificial stupidity. It is possible to agree that AI may pose an existential threat to humanity, but without ever having to imagine that it will become more intelligent than us.

Elon Musk has branded artificial intelligence “a fundamental existential risk for human civilisation”.

He says we mustn’t wait for a disaster to happen before deciding to regulate it, and that AI is, in his eyes, the scariest problem we now face.

He also wants the companies working on AI to slow down to ensure they don’t unintentionally build something unsafe.

The CEO of Tesla and SpaceX was speaking on-stage at the National Governor’s Association at the weekend.

“I have exposure to the most cutting-edge AI and I think people should be really concerned about it,” he said. “I keep sounding the alarm bell but until people see robots going down the street killing people, they don’t know how to react because it seems so ethereal.

“I think we should be really concerned about AI and I think we should… AI’s a rare case where I think we need to be proactive in regulation instead of reactive. Because I think by the time we are reactive in AI regulation, it’s too late.

“Normally the way regulations are set up is that a whole bunch of bad things happen, there’s a public outcry, and then after many years, a regulatory agency is set up to regulate that industry. There’s a bunch of opposition from companies who don’t like being told what to do by regulators. It takes forever.

“That, in the past, has been bad but not something which represented a fundamental risk to the existence of civilisation. AI is a fundamental risk to the risk of human civilisation, in a way that car accidents, airplane crashes, faulty drugs or bad food were not. They were harmful to a set of individuals in society, but they were not harmful to society as a whole.

“AI is a fundamental existential risk for human civilisation, and I don’t think people fully appreciate that.”

However, he recognises that this will be easier said than done, since companies don’t like being regulated.

Also, any organisation working on AI will be “crushed” by competing companies if they don’t work as quickly as possible, he said. It would be up to a regulator to control all of them.

“When it’s cool and regulators are convinced that it’s safe to proceed, then you can go. But otherwise, slow down.”

He added: “I think we’d better get on [introducing regulation] with AI, pronto. There’ll certainly be a lot of job disruption because what’s going to happen is robots will be able to do everything better than us. I’m including all of us.”

Earlier this year, Mr Musk said that humans will have to merge with machines to avoid becoming irrelevant.

Ray Kurzweil, a futurist and Google’s director of engineering, believes that computers will have “human-level intelligence” by 2029.

However, he believes machines will improve humans, making us funnier, smarter and even sexier.

Using Artificial Intelligence To Fix Healthcare

The healthcare industry should be using Artificial Intelligence (AI) to a far greater degree than at present, but progress has been painfully slow. The same factors that make the healthcare system so attractive to AI developers - fragmented or non-existent data repositories, outdated computer systems and doctor shortages - are the same things that have stopped AI from providing the gains that should be created.

The healthcare sector also presents unique obstacles for AI: data must flow freely through AI systems to achieve real results, but extracting data from handwritten patient files or PDFs is cumbersome for us, and difficult for AI. Despite technical and operational challenges, new research suggests that the arrival of the tech giants into the industry may provide the data and the capital required to digitize this fairly untapped market.

Where AI can help now

Severe fragmentation between different branches of healthcare, and life-threatening miscommunication within institutions (in 2016, ~10% of all US deaths were caused by medical errors), presents an opportunity for AI to ease the burden on doctors in more creative, less intrusive ways. Mabu is a humanoid robot developed by Catalia Health and works with the American Heart Association to help patients keep on top of at-home treatment for congestive heart failure. Acting as a personal health assistant, Mabu asks patients how they are feeling, makes activity suggestions and provides medication reminders. ‘There are key points we make sure Mabu covers,’ says Catalia Health founder Cory Kidd, ‘but the conversation is adaptive to what is going on with that patient at that moment,’ much like a home nurse’s visits might be scripted to a certain degree while relying on some human intuition.

Mabu is a promising step towards integrating AI into the healthcare system without disturbing doctors within facilities - the data Mabu gathers can be fed into Electronic Medical Records (EMRs) via email or text, and ‘daily conversations’ with the device mean that Catalia Health can collect patient information consensually ‘without depending on access to their medical data.’ The implementation of AI throughout healthcare institutions or an entire country will remain a huge task even for data-rich multi-nationals, but solutions like this may help to improve outpatient care and reduce readmission rates for long-term conditions without setting foot in a hospital.

The grand scheme

The move towards integrating AI with hospitals and healthcare centers is gaining pace. The UK government has announced its intentions to put the UK ‘at the forefront of the use of AI and data in early diagnosis, innovation, prevention and treatment’ by 2030. While this may be ambitious given the current status of technological advancement in the NHS, hospitals are working with a wide range of companies to tackle immediate problems on the ground. Chatbots, DeepMind, and voice biometrics are all being used to alleviate unique problems in the sector, and some companies are taking a different approach to ensure that AI is used to its full extent.

Dr. Shafi Ahmed is a cancer specialist in practice for more than 20 years, and as such is in high demand from people who need his expertise. In a conversation with Steve Dann, who at the time was working in visual effects, Dr. Ahmed related that of the 300 students under his care he was only able to directly train two at a time because the operating theatre could not fit any more people. Dann suggested he filmed an operation using 360° cameras, which enabled Dr. Ahmed to show students exactly what to do without having 300 people breathing down his neck.

This led Steve Dann to found Medical Realities, and use his skills in CGI and virtual reality to train medical students without making more work for doctors. Following the success of the 360° operation, Dann created a virtual Dr. Ahmed to answer questions for him, which is now used in conjunction with teleconferencing to help Dr. Ahmed’s cancer patients feel more comfortable in follow-up appointments. Using this combination of AI, virtual reality and CGI, Dann is working with Leeds and Queen Mary University hospitals to create virtual surgeons that can train new doctors or assist in surgeries where one or more doctors cannot be physically present.

Bit by bit transformation

Bringing about an artificially intelligent healthcare landscape will be a significant challenge, due to the sheer amount of mission-critical work on doctors’ shoulders, outdated systems and handwritten records, and fragmentation between care facilities. But Rome wasn’t built in a day, and there is already significant progress being made that does not disrupt the daily work of doctors and nurses on the ground, while still offering improved care to patients.

The AI healthcare sector is ripe for development and investment, but while the data giants figure out how to transform the system as a whole, smaller-scale projects are making real changes. Piece by piece, patient by patient, AI is on its way to fixing healthcare once and for all.