Fetch says every machine on the IoT needs a really good agent
Supercomputers can slaughter anyone at chess and Magnetic Resonance Imaging (MRI) scanners can see right through you – but they are useless at selling themselves. With Fetch.AI they can realise their full potential, says freelance technology writer, Nick Booth.
The technology industry is obsessed with fuelling fast growth, which must come at the expense of everything else, from forests to fossil fuels. The plan seems to be get big, get burning and get out.
Here’s a counter culture idea from a company in Cambridge, UK, comprising Deepmind experts, professors and more PhDs than University Challenge. Why don’t we make better use of what we’ve already got? There are magnificent but under-used machines everywhere, from the Metro to the Magnetic Resonance Imaging department, all being under-used.
Couldn’t we create surges in productivity by working them harder? That is the rationale of Fetch.AI, which wants to help machines everywhere by breathing ‘economic’ life into them. How?
Fetch aims to network all the machines into one great big Capacity Exchange that supplies capacity where it’s in demand, helping the machines to market their assets at the best possible price. This is achievable if we could create what economists define as perfect market conditions. To do so, each machine must act as an intelligent economic agent with the perfect information afforded them by the IoT.
Machines need an agent
However, machines lack the rationale to act on that information. Which isn’t surprising, as they were designed for one particular creative talent. To make the most of themselves, they need an agent. Step forward Mr 90 Percent, Al Gorithm.
Machines are ‘smart stupid’, with an unmatched talent for one thing, such as preventative medicine, creating energy or running transport infrastructures. But they’re useless at everything else. They can’t think outside the box, because nobody programmed them to. Yes, an MRI scanner can see right through you but they are remarkably naive at times.
Which is why they need an agent. Or a brain. This is exactly the problem that Cambridge-based Fetch.AI intends to help them with. It doesn’t want to be their agent, it intends to anoint them with the gift of Artificial Intelligence (AI). So, for example, an unemployed turbine might be able to sell its extra capacity to the highest bidder. Such as some local data centre that needs a source of green power and is willing, on this one occasion, to pay extra because there’s been a surge in online shopping and it needs to cool down its burning CPUs using green power.
Fetch’s method for empowering the world’s most intelligent machines is too complicated to describe here. Suffice it to say it assembled 50 of the finest intellects on the Cambridge technology scene and formed a multi-disciplinary team for everything from algorithms to aggregation.
Security is King
Security is a priority, says Fetch CTO, Toby Simpson. “We have experts in machine learning, cryptography, Artificial Intelligence, software engineering, you name it,” says Simpson. The industries that could most benefit from these autonomous economic agents (AEAs) are those with the most moving parts, says Simpson.
Four sectors spring to mind: transport, the supply chain, energy and the travel industry. When there are loads of moving parts, often presided over by dumb devices, there are massive management issues. Devices are too stupid to secure themselves, which makes the IoT inherently dangerous. And they cannot manage themselves into a team, because weren’t designed to. Management in these cases is very labour-intensive, like trying to keep multiple plates spinning. “In the end, what we are about is letting the plates spin themselves,” says Simpson.
The genius of this scheme would be that it would run itself. Are some machines better than others at striking deals? Can one man’s algorithm outwit an algorithm from a rival? You’ll have to ask Fetch.AI these questions.
Happiness comes not from relentless acquisition, but in learning to love what you already have. Who’d have thought we’d learn that from an agent? An automated agent, too. Maybe machines can teach us something valuable after all.
The author is freelance technology writer, Nick Booth