By Richard Sharpe

August 2022

Ninety-five years ago this month Marvin Minsky, the pioneer of AI, was born in the USA.   In his 88-year long career he led, and some say occasionally misled, the development of AI leading him to the conclusion that “”somewhere down the line, some computers will become more intelligent than most people”.  It is possible that an AI machine could take over the resources of the world to help it solve problems.  But he felt it was hard to take this last possibility seriously.  He laid the foundation for a lot of AI on which others built.

Later Nick Bostrom in his 2014 book Superintelligence suggests that AI will create a superintelligence, one which will actually ramp up and become better than the intelligence of the most intelligent human beings, even when human beings act collectively.

What do those contributing to the Archives think of this prospect: the disconnection of AI and human intelligence?

Mandy Chessell CBE worked for 35 years at IBM as a software engineer, latterly on AI projects.

Chessell told the Archives, “The problem is that these are words that have a very loose definition. So, most good AI, is actually a combination of human expertise and machine expertise and there are things that we do well and there are things that machines do well. When we collectively work… we create a greater intelligence, and the way the Covid vaccine was created was an example of that network effect. We are already [acting] collectively, through our linking and IT creating a greater intelligence in-on the earth, compared to what we had 200 years ago.”

Professor Nick Jennings, from Imperial College, has built complex AI systems for operational use.

Professor Nick Jennings is the professor of AI at Imperial, among a string of other top academic appointments.  He told the Archives: “I think humans are amazing problem-solvers, and, really good at certain things that machines are simply not good at, and I am unconvinced will ever be really good at, or as good as humans at. But having said that, there are things that computers do that are way better than, than us as humans. And we’ve seen that throughout computing, you know, computers can add up better, faster, more accurately than people can… I think as AI progresses, some of the, some of the tasks that are better suited to computers and machines will be taken on by AI systems, but there will always be activities that are best done by humans.”

Among other AI projects, Professor Jennings helped build a disaster response system using AI and human interaction.  “So if you think about our disaster response application,,, one of the tasks that we needed to be able to do was to plan a schedule of which rescue teams would go to which bit of Nepal at what time, with what resources. And previously Rescue Global [the charity they were working with] did that as a hand calculation, they would sort of say, ‘Right, we’re going to send some over here, and we’re going to send some over there.’ Now that’s an optimisation problem. You have a set amount of resources, a set amount of times and priorities, and actually machines are very good at solving those sorts of things. And so we automated that. But actually, what we found was that when we had just automated the solution on its own, the human problem solvers would always want to be able to say, ‘Actually that’s nearly right, but I want to tweak this bit, and I want to send this team to this area, and this team to this area,’ for reasons that I understand, or have intuitions about that are not in the data or are not, the machine’s unaware of. And so they wanted to be able to tweak things. And so, that model of humans and machines working together as partners where some of the tasks are done by humans, some of the tasks are done by machines, and there’s a flexible interplay between that, has really shaped my view of AI. And as AI has become increasingly into the public consciousness, this mixed systems of humans being in the control, and being supported and augmented by smart computer systems is starting to gain prominence.”

Professor Jim McLaughlin OBE of Ulster University is an expert in biomedical IT using AI.

Professor Jim McLaughlin focuses on AI and medical systems: “[T]here is obviously a lot of emphasis now on the ethics of AI…AI is there to help, it’s an enabler, and it’s a tool. And I think very much like the role that even computing plays on an aircraft or a car, I think there is safety enhancement, performance enhancement, all of that, but absolute control and decision-making, and critical decision-making, may get prompts from the AI world but I don’t think AI should ever be to control it you know.  However, you know, clinicians have another role [than just taking blood etc].. [T]hat smart piece of actually knowing how to deal with situations and people and their emotions, their mentality etc., how they progress on. … I think it’s too dangerous to allow AI to be the absolute system that takes over from humans in every area that we need.”

Twenty one of the over 200 people who have contributed to the Archives have worked in some part of the field of AI.  The Archives has selected them so that you can judge for yourself by clicking through to our Artificial Intelligence theme page.