A dabble with an online artificial intelligence platform proves a dispiriting experience for AM's regular columnist Professor Jim Saker, emeritus professor of Loughborough University's business school and president of the Institute of the Motor Industry.
A few weeks back I was involved in a long debate over what the impact of AI was going to be for the retail automotive sector in the UK. The people involved were industry leaders and were attempting to look into a crystal ball of uncertainty to map out a path going forward. In essence what would AI do for us and alternatively what will it do to us?
When I returned home, I decided the obvious solution was to ask ChatGP itself what impact it thought that it would have on the industry. I dutifully typed in the question and within seconds the programme produced 10 areas that it believed AI would have an impact in retail automotive.
The response was quite underwhelming, there were 10 areas identified that could be clustered into four groups.
One group was to do with the enhancement of customer experience based around utilising better data, the use of chat bots and virtual assistants enhancing virtual showrooms using augmented reality. Most of these are in operation now or at least in development.
A second cluster was around predictive inventory management, price optimisation and supply chain alignment; again, these are all areas that are in operation with existing software.
The final two areas that the AI programme said it would impact were slightly more interesting and showed some areas of real potential development. One was based around predictive maintenance utilising real time data from sensors, this is a capability that is already available and is linked to the issue of connectivity with the car but has the potential to be developed much further to minimise breakdowns and improve vehicle reliability.
It was also suggested that this was inextricably linked to autonomous driving which again is likely to be enhanced by AI but probably restricted by other external factors such as insurance liability and the ethical base upon which any system is based.
The final area that ChatGP said that AI would impact was on fraud detection and security and also the enhanced need for data security and privacy. This presents quite an interesting challenge with AI saying that what it produces needs to be protected from technologies such as itself.
In fact, the expectation is that we will have to protect it from itself. It can help to identify fraud but will need people in dealerships to impose better security measures to keep the data private and meet legal guidelines.
Having spent some time looking at my printout I felt somewhat disappointed by what the programme was suggesting.
There seemed to be numerous areas that it had missed. I personally would have thought that technical training would be enhanced by AI. It could also have a significant role to play in problem solving and implementing the diagnosis of a fault on a vehicle.
My list started to go on and on.
Perhaps the interesting but slightly scary thing is that after this article goes online, in the spirit of AI, it will have picked up my disappointment and will do better the next time it is asked.
Dale Reed - 15/12/2023 16:40
I'd be curious which version of ChatGPT (not ChatGP, unless there's a new variant that can give you a prescription at the same time, not a bad idea thinking about it!), version 4 is considerably better (at least, some of the time) than 3.5. The problem with ChatGPT is you rarely get the answer you want on the first try, and have to spend some time finagling to get the right prompt which leads to the answer you needed. It's why a new role of 'prompt engineer' is fast becoming a decent career option with a 6 figure salary; knowing the right prompt to use to get what you want from the product is a definite skill to learn. Ultimately at the moment this kind of AI can't invent or create anything new, but it can create realistic responses from its model and data. Unfortunately that data is the flawed data that exists on the internet already, it just does a decent job of amalgamating it into a reasonable response for you, albeit one that often needs serious fact checking. Given how early on we are in it's progression though, you have to admit for what is effectively an algorithm, its responses are pretty realistic, and will only get more convincing as it grows. Exciting and also worrying times!