So you want to get rid of motorists?

5 September 2017

Dan Hulme from FiveAI was the guest speaker at our September monthly meeting. Here he explains the start-up’s work towards, and vision for, a new mode of transport.

It’ll be a few years before you see our cars on Cambridge’s streets, but we already have them driving on test tracks.
Image as described adjacent

I guess it’s common in Cambridge offices for motorists to be in the minority, and mine is no exception. We all experience bad interactions with motorists every day, and we want an approach to road safety that doesn’t rely on people never making mistakes or misbehaving. At the same time, there are some journeys that bikes aren’t great for, and the other options aren’t much better. Buses don’t always go where you want when you want. Taxis are more convenient, but expensive and useless for longer journeys. Trains are great – unless you want to visit our Oxford office. Hire cars or car-share schemes work for longer journeys – but if you don’t keep in driving practice, or if you don’t have a licence at all, they might not be a good option.

Where my office is different is that we’re working to solve both these problems. FiveAI is a local start-up solving the challenge of making cars that drive themselves. One of the reasons we want a locally-built solution, rather than waiting for the US tech giants to solve the problem, is that their cars, built for US grid-layout cities with jay-walking laws and few cyclists, won’t work for our medieval towns, sharing the road with pedestrians, cyclists, horses, and the occasional cow. We also have a fundamentally different idea about car ownership.

Tesla wants the autonomous capability to be a luxury item: something that will make you buy their more expensive car. In contrast, we’re fed up with single-occupant cars taking up precious road space, and we don’t want to make this worse with zero-occupant cars.

Instead, we see autonomous vehicles as enabling a new kind of public transport, which is as cheap as a bus but as convenient as a taxi -and which can be used for longer journeys too.

That’s why our first scheme is a partnership with Transport for London and the Transport Research Lab. Named StreetWise, the scheme will be like a Dial-a-Ride service in an area of London with poor access to public transport.

I’m looking forward to sharing the road with our cars, if they enable more human drivers to switch to car-free living. When I talk to people about my work, they’re more cautious. They’re right to be concerned about the safety implications, and we’re being cautious too. People ask about policy: will an autonomous vehicle always stay out of cycle lanes, when will it overtake, whose safety will it prioritise, and suchlike. It’s too early to give definitive answers to these questions. This is a completely new area of regulation, insurance, and ethics. Advancing these is as big a challenge as the technology. Even so, we can already guarantee that there are whole classes of human errors and misbehaviours that won’t even be possible. Our cars won’t get drunk, tired, or angry. They won’t deliberately ram you as a ‘punishment’. They won’t start overtaking and then squash you into the kerb because they forgot you were there. They won’t overtake and turn left at the same time. They won’t be too lazy to indicate.

Our cars will treat all human life as equal, be that passengers in our vehicle, in other vehicles, or pedestrians or cyclists. They can evaluate complex risk more accurately and can act with super­human reflexes and the skills of an expert stunt driver to make things safer for everyone.

Achieving our vision will take a lot of expertise – but there are no experts in this field yet. That’s why we have offices in hotspots of science like Cambridge. If you have experience in machine learning or vision systems, 3D programming, or other relevant skills, and you want to help create a new, safer mode of transport that fills the gaps in traditional public transport, please do check the internship and permanent opportunities on our website at five.ai. Your fellow cyclists will thank you.

Dan Hulme