AATISH MISHRA
Aatish ponders how AI could disrupt the automotive journalism space
AI can regurgitate specs and give you recycled opinions that it has probably been trained on, but it hasn’t experienced these cars for itself
AM I GOING TO LOSE MY JOB IN A COUPLE of years?
If you’ve been online in the last couple of weeks, there’s been no escaping the discourse around AI. GPT4, ChatGPT and other language models have proved to be incredibly adept at performing a number of tasks, be it having conversations, writing, editing or more complex tasks such as problem solving, giving feedback on ideas, coding or explaining the nuances of language — you name it, AI can probably do it. One of the many posts I saw online was about the industries and labour markets most likely to be disrupted by these GPT (Generator Pretrained Transformer) models. It featured a study stating that in the context of the US workforce, “80 per cent will have at least 10 per cent of their work tasks affected by the introduction of GPs, while around 19 per cent of workers may see at least 50 per cent of their tasks impacted.” Content writers and copy editors sit close to the top of that list.
I’ve seen the power these models have up close, and we’re already leveraging them today. Until not too long ago, the photo and video teams provided visual content that the editorial team added a caption to and posted it on evo India’s social media platforms. Today, the photo and video teams can function independently — putting in a prompt into these models and generating a grammatically-correct, catchy caption for their visual content in a matter of seconds. The models are adept at giving headlines, précis-ing content, cleaning up bad grammar and stunted sentences and so much more. Feed it pointers and it’ll draw up a whole story for you. As a test, I asked it to help me pick between a Porsche 911 Carrera S and a Huracan, and the sort of feedback it gave me was surprisingly close to what any automotive journalist would give someone confused between two cars.
However, it isn’t all doom and gloom just yet. For starters, AI can regurgitate specs and give you recycled opinions that it has probably been trained on but it hasn’t experienced these cars for itself. You still need to get inside a car and drive it, to understand it properly and give a proper verdict on it. No chat model can tell you how soul stirring revving out a Porsche flat-six to 9000rpm is. Or how brutal the acceleration from one of the new-age EVs feels. Or what pulling 2 lateral Gs in a road-racer will do to your innards. They can’t tell you about the awe you feel when you drive through Ladakh or the north east. They can’t capture the roller coaster of emotions that a motorsport weekend is. Sure, they can imagine it — using the million pieces of information they’ve been trained with. But they will never be able to feel it, understand it, process it and communicate it.
Secondly, AI doesn’t have a personal opinion. It might share different perspectives but it doesn’t have its own opinion. Fundamentally, our jobs come down to giving our opinions, based on the vast array of experiences we’ve had with cars. When I drive a new car, the only way I can give an appropriate verdict on it is if I’ve driven its rivals, and have a firm understanding of their strengths and weaknesses. I can describe the feelings I felt on a road trip, but those feelings are a culmination of my world view — a world view that has been developed from my very existence. Thoughts and feelings that make us human. AI today isn’t sentient yet and certainly cannot share what it feels. That’s where the role of the automotive journalist will remain alive for a long time to come.
This could change — I can see a future where a machine can give an opinion on how a car would behave based on certain parameters. If we can programme simulators to mimic real cars, it shouldn’t be impossible to reverse engineer them and do the opposite — create a rig that will drive the car and understand how it performs, and possibly even “feels”. Since this machine will be able to simulate the exact same conditions for every car, it will also be highly accurate with its findings. Manufacturers already do something similar in their development processes, before they set their technical specs in stone. It will just be a matter of extrapolating data and communicating it in a language our readers are familiar with. With the way technology is progressing, almost nothing seems impossible if there’s a will to create it.
GPT models might not take away our jobs but it will certainly change how we work. It will be a handy tool to make us more efficient, and possibly more eloquent. However, I see another AI threat looming on the horizon, one that is far more real than GPT models. Autonomous cars. Once cars start driving themselves, what do I do as an automotive journalist? ⌧