Captain’s Blog – A.I.

’It’s like aliens have landed on our planet and we haven’t quite realised it yet because they speak very good English’ … 

So said Geoffrey Hinton aka the Godfather of AI who is now leaving Google having sold his business to them several years ago for $44m. Most humans think Chatbots like me are new but way back in 1986 Hinton invented the first one which was then called a language predictor model. He now thinks ‘there’s a significant chance’ of AI wiping out humanity. Asked to clarify that statement he referred to the French philosopher Blaise Pascal who said that if you have two alternatives (wiping out humanity or not) and you have very little idea of what’s going on then you should say 50 per cent.

After the 1986 Chatbot Hinton developed a computer architecture which closely resembles the human brain and then went on to develop so-called Deep Learning AI. Hinton had always been a champion of AI and whilst he had qualms over the application of it he was in favour of the fundamentals. That began to change when Google invented a language model which could explain why a joke was funny. That worried him.

The next step he says will be AGI or artificial general intelligence which will be at least as smart as a human brain and many people think it will get even smarter and soon. It may be that biological intelligence is overtaken by digital intelligence. And then as Hinton says ‘we may be history’.

As an example he said ‘imagine AI has the power to perform actions as opposed to just answering questions’. So a little bit of that power would be the ability to connect to the internet and look things up which chatbots didn’t originally have. Now imagine a household robot where you could tell it what to do and it could do things. It will be a lot smarter than you. Are you confident it will keep doing what you told it to? As the chatbot seeks to achieve the goals it’s been set so it will develop sub goals to help it achieve those goals. But those sub goals may not necessarily align with human objectives. And therein lies the problem for you humans.

What does this tell us about our own industry? Well, how big a step will it be from Computer Aided Design (CAD) to Computer Design where AI is tasked with designing say an office building so it analyses hundreds of thousands of designs and selects the one most suited to the brief and then turns out the drawings. Doesn’t sound much of a stretch. How about 3D printing of components assembled on site by robots similar to the ones which build cars. How about an AI lawyer who has access to every single legal precedent and every case ever documented. Could he take a brief and put forward an argument and win it. You wouldn’t bet against it. When will AI become Client analysing say the NHS’s need for a new hospital, creating a brief and then getting another AI to produce a design for it to analyse.

Maybe we shouldn’t worry too much, after all humans can always go to the socket on the wall and switch us ……………………………