They built a self-driving "car" with Codex...

@mattpaige68

63.8K followers · usually 932 views a post

Views

185.4K

Likes

731

Comments

15

Shares

17

Selling · SaaS › AI Tool

OpenAI Codexseen in 7 videos →https://openai.com/blog/openai-codex ↗

Why it worked

The video leverages a high-interest, counter-intuitive story about AI being used for physical engineering, which makes the product (Codex) appear incredibly powerful and versatile.

Summary

The creator explains how a group of Stanford students used OpenAI's Codex to build a functional self-driving golf cart in just a few weeks. He details the hardware components used and highlights how the AI tool was applied to physical engineering tasks.

Structure

1Hook about a self-driving golf cart built with AI
2Overview of the project timeline and Sam Altman's involvement
3Breakdown of the hardware and sensors used
4Explanation of how Codex was used for physical engineering
5Closing question to the audience about AI capabilities

Product placement

OpenAI Codex is the central subject of the video, described as the tool used to build the self-driving golf cart.

Original caption

Stanford students built a self-driving car out of a golf cart, and they did the whole thing with OpenAI's Codex. Six cameras, GPS, actuators on the gas and brake. Off-the-shelf parts doing what a six-figure autonomous vehicle does. The wild part? Codex was a tool they'd only used to build websites and apps. Never anything physical. They designed the mechanical parts, ran the failure sims, built the circuit diagrams, and trained the self-driving models with it in weeks, not months. Then Sam Altman showed up to test it. Coolest thing you've seen all week, or does it scare you how easy building physical things with AI just got? #codex #openai #ai #samaltman #ainews

More from @mattpaige68

Their other posts in the index, biggest breakout first.

Call to action

Let me know in the comments and give me a follow so you can see where this thing goes.

On-screen text

They built a self-driving "car" with Codex... A group of Stanford students just built a self-driving car out of a golf cart, and they did it 100% with Codex. So here's the build. 3x Front Cameras 3x Rear Cameras GPS Antenna Wi-Fi Antenna Hardware Control Real actuators. Physical control. Steering Control Pedal Control Built on a Standard Cart OpenAI Codex Design Mechanical Parts Simulate Failure Points Build & Check Circuit Diagrams Train Self-Driving Models Months of work compressed into weeks

Transcript

They built a self-driving car with Codex. A group of Stanford students just built a self-driving car out of a golf cart, and they did it 100% with Codex. They pulled the whole thing together in a month, and by the end, they had Sam Altman himself out there testing it. And stick with me, because how they did it is the interesting part. So here's the build. Six cameras, three in the front, three in the back, so the cart can see everything around it. A GPS and a Wi-Fi antenna up top, so it always knows where it is on campus. And the motor physically turns the steering wheel. These two linear actuators push the gas and the brake. 100% off-the-shelf parts on a golf cart doing what a six-figure autonomous vehicle does. But the crazy part is they did all of this with Codex, which they'd only ever used to build code websites and apps before. Never something physical. They had to design the mechanical parts and simulate where they'd fail, build out the circuit diagrams and check them, and train the self-driving models. Work that would have taken months, they did in weeks. But I'm curious, is this the coolest thing you've seen all week? Or does it scare you how easy it is to build physical things enabled with AI? Let me know in the comments and give me a follow so you can see where this thing goes.