Can ChatGPT & an engineer electrify a 1958 Mercedes?

With a degree in mechanical engineering from Stanford, an early career at Ford and 10 years as CEO of Fictiv, can Dave Evans use ChatGPT to make his dream of electrifying a 1958 Mercedes a reality?

Dave Evans, Fictiv CEO, holding the Mercedes exhaust pipe. (Image: Fictiv)

Dave Evans, Fictiv CEO, holding the Mercedes exhaust pipe. (Image: Fictiv)
Hello, fellow car enthusiasts, engineers, and tech geeks! This year, I’ve taken on a huge engineering challenge: Turn my 1958 Mercedes 220s into a modern electric vehicle (EV). Watch, as armed with a toolbox, a sense of humor, and a wild vision, I set forth on my quest to build the electric vehicle of my dreams. The project’s parameters: I have to be able to work on it at night and on weekends and complete it in 2024. Oh, and my chief engineer? That’s ChatGPT 4.0. I’ll use it for research, coding, calculations, and more. Work that would take me months, ChatGPT cranks out in hours – but more on that later.

I’d like to invite you to come along for the ride. Think of my ChatGPT buddy and chief engineer as generative AI with an engineering degree. I’ll use Fictiv to source and build the custom parts and a few dedicated friends to help on weekends. Let’s call this project “Electra.”

For the past 10+ years, I’ve picked a build project to keep my mechanical engineering skills sharp, use my degree, and continue to work on projects I’m passionate about.

In the past, Fictiv teams have built a miniature electric car, an open-source motorcycle, and even a Fiat conversion. Every year I pick a new discipline that I know nothing about and then dive into it over the course of that calendar year.

This year, I wanted to apply my passion for climate tech and ChatGPT to convert a classic car into an EV. And so I took my love of automotive, where I started my career, and my love of engineering and asked, “What if I used ChatGPT for all of the hardcore engineering? What if we took an old car, something from the ’50s, and converted that to an electric vehicle? With ChatGPT as my chief engineer and myself as a mechanic, could I convert an old classic car to an EV and remove all its fossil fuels?”

The Car

Choosing which car to convert requires a lot of research. Many EV conversions are old Fiats, VW buses, or VW bugs. Sports car enthusiasts also convert Porsches and other cars built for speed and performance, but I didn’t see any groundbreaking, new classic car conversions to add to the community of VW conversions. I had an image of taking my wife out on date night in a town car or old chauffeur vehicle—a really cool classic car built for style more than speed. It all started with a simple question on my phone to ChatGPT.

 Initial ChatGPT query to identify the right vehicle to convert. (Image: Dave Evans/Fictiv)

Initial ChatGPT query to identify the right vehicle to convert. (Image: Dave Evans/Fictiv)

I spent a lot of time querying ChatGPT: “Give me a classic car, but with more headroom. I’m 6’5″, so I also need more legroom. Now I want more battery storage, but I still want a trunk.” Eventually, I ended up with the Ponton series, which is a 1950s series of classic cars. And I love the look, size, and shape of a late 1950s Mercedes 220S vehicle.

The 1958 Mercedes-Benz 220S before electrification. (Image: Dave Evans/Fictiv)

The 1958 Mercedes-Benz 220S before electrification. (Image: Dave Evans/Fictiv)

The Engineering

Here’s where it started to get real: Finding the old engine and developing the specs for the car and how to convert an ICE (internal combustion engine) into an EV. I needed ChatGPT to give me all the specs of the existing ICE engine, including the torque curves at different RPMs. Then it gave me all the gearbox ratios—everything that was standard on the car. Next, I wanted to know the options for EV equivalents, specifically the motor, and where to find those parts. ChatGPT showed me specs on everything from a Tesla Model 3 to a Nissan Leaf to other engines you can buy directly from the manufacturer.

Torque and required motor output comparison across gears, RPMs, and vehicles produced by ChatGPT.  Vehicle speed in different gears produced in Python by ChatGPT. (Image: Dave Evans/Fictiv)

Torque and required motor output comparison across gears, RPMs, and vehicles produced by ChatGPT. (Image: Dave Evans/Fictiv)
Vehicle speed in different gears produced in Python by ChatGPT.(Image: Dave Evans/Fictiv)

Vehicle speed in different gears produced in Python by ChatGPT.(Image: Dave Evans/Fictiv)

Another major consideration: Do you keep the original transmission and gearbox, or do you put in a new one? This was a big decision point as it would change the project’s scope. Consider this – A Tesla motor has an RPM range from zero to 18,000 versus the original ICE motor in the Mercedes that tops out at 5,000 or 6,000 RPMs before redlining. So if you put a Tesla motor in, you’d need at least a 10-to-1 ratio gearbox, which takes the motor spinning at 18,000 RPMs and steps it down so that the wheel speed is appropriate. So I’d have to do a new gearbox and drivetrain, which was out of scope for a one-year project. We had to find a motor with a similar RPM to the Mercedes, with the right torque curves, so that I wouldn’t blow up the gearbox, and could drive the hills of San Francisco, all while improving the performance compared to the original ICE engine.

From there, ChatGPT generated several Python scripts that compared four curves through the various gears for the original ICE engine versus several electric motors. I had it graph all of those torque curves against the engine RPMs to determine the wheel speed. This took some updating from me, but the template was remarkably accurate!

On a Sunday afternoon, ChatGPT wrote all of the Python scripts that would’ve taken me at least one month to write in MATLAB back as an engineering student and much longer than that as a rusty weekend engineer. As someone who’s not doing a lot of vehicle dynamics or engine performance calculations these days, this is not my wheelhouse. It’s safe to say it would’ve taken a long time to write the boilerplate, write the code in Python, pull the libraries I need, and graph it out, etc., to get the results ChatGPT generated in an afternoon. In one afternoon, we looked at five or six different electric motors and down-selected to the HyPer9 high voltage. Truly incredible!

What’s Next

I’m now at the stage where I need to do some hardcore engineering to solve four major challenges. And I’ve partnered with one of my super nerd friends, William Burke, CEO of Five Flute, to tackle these challenges together.

Here are the four engineering challenges we still need to tackle:

  • How to adapt the new electric drivetrain into the vehicle and transmission
  • How to reliably deliver torque from the electric motor through the stock transmission
  • How to design a waterproof battery box and mitigate safety concerns
  • How to mount this new power source in the vehicle while maintaining comparable vehicle dynamics to the original design

We plan to follow this post with some extra nerdy content, some live webinars over beers, and generate a lot of questions and feedback from the community to bring this project to life.

Stay tuned and check back here for more on the Electra project, her engineering problems, and potential solutions.