AI helps you iterate prompts
When you iterate prompts to AI it leaves a huge mess of changes. Human are notorious for not updating comments and the documentation. This makes life harder for the next person which perversely is often the same person. Iterative prompts with an AI can make the mess even worse. Each revised prompt gets further and further from the original. Here is a workflow that both improves the results and reduces your work now and the world of the next person.
When working with AI you give it prompts, it returns a result, you repeatedly give it refining prompts until it gives you what you want. At that point there is a mess of revisions. Try this:
- Write a Requirements document prompt and save a copy. If you use version control like Git commit add and commit this document as your first version. This is like writing the comments first.
- Use the Requirements document prompt with the AI. This is coding to match the comments.
- Review the results and interactively prompt small changes. Test driven development.
- When you have a stable result, ask the AI to generate a single prompt that to make the current result and translate it into markdown. This is the AI telling you how to talk to it. You & the AI updated the comments & documentation.
- Save the results and the new prompt Requirements document. Add and commit this to Git.
It is now easy to move forward, backwards, or even to a different AI “employee” to continue to ultimate success.
Learn how to program a robotic car in 7 weeks
Two years ago in “Why We Don’t Have Flying Cars, Yet” I explained why automation is the next big innovation for vehicles, not alternative energy.
Standford is offering a 7 week undergraduate class teaching how to program a self-driving car. Automation improves the under-satisfied outcome of cars but it is also technologically easier to make than low-cost long range batteries for an electric car.
The Predictive Innovation report the video was based on was first offered to GM but they turned it down. European and Asian car companies used the information and are now selling cars with automated driving features.
Volvo’s XC60 has a City Safety feature that automatically brakes to prevent crashes. It’s a pure gasoline car with lots of room and power. It is priced about the same as the Chevy Volt, although doesn’t receive any of the government incentives.
| 2011 US Car Sales | |
| Car | Units Sold |
| Chevy Volt | 7,671 |
| Volvo XC60 | 12,932 |
The Volvo XC60 with City Safe automatic braking sold 68% more cars in the USA than Chevy Volt. So not only was it easier to build, and thus more profitable, it sold more units. The automated car is more desirable to customers. One of the key points of the report was to offer incremental improvements with meaningful value to customers. That made sure the new features were high quality and low cost.
In addition to satisfying safety, automated cars are better for the environment than an electric car. Electric cars just shift the source of pollution from burning gasoline in the car to burning coal at a power plant. Automated cars use less energy.
First, replacing or repairing a car damaged in an accident uses more energy than the car ever will from driving it. And how can you count the cost of injuries and deaths?
Secondly, by reducing traffic congestion automated cars can save energy for all the cars on the road while reducing drive times and frustration.
Automation in vehicles is still a big innovation opportunity.
Predictive Innovation Training
Predictive Innovation: Core Skills Book
RoundSquareTriangle.com