How AI is changing programming
Last week, we had a hackathon with our company's engineering team. The goal was to build something cool and valuable throughout the weekend. We had a few ideas, but we devised a Chrome extension that would rephrase your message according to the tone you selected or generate a response according to the messages you received. It also shows the mood of your message. We called it CommPilot.
We created a Chrome extension to inject an event handler into the Google Chat input field. The handler would send the message to our backend, which would then process and send it back to the extension. The extension would then replace the message in the input field with the processed one. The backend is running on AWS Lambda and using OpenAI SDK.
We completed the PoC in a couple of hours, and the most time-consuming part was dealing with the iframes in Google Chat. The algorithm was implemented in less than an hour and was just a prompt to OpenAI API. I can not imagine writing such an algorithm from scratch.
Yes, it is non-deterministic and sometimes hallucinating, but it is a great start. In a couple of days, you could produce a product that would become a competitor in the market. And it is just a PoC. Imagine what you could do if you had a couple of months.
We also use it to generate code. It still needs to be better. However, it is better than a mediocre developer.
Imagine having computers that will have LLMs embedded into CPU/GPUs. That would natively process the user input and produce the results without specific software.
With this speed, most of us will soon become technical product managers, and the rest will be replaced by AI.