I know a lot of people don't like the direction of AI but people need to take there heads out the sand and stop pretending like it's never going to be good enough to replace you. All you're doing is hurting the people who are going to be effected by giving them false hope.
Comments
There are issues with AI which I don't think more training data will address. there are limits which current architectures cannot account for.
The outputs weren’t the goal, I wanted to see the capabilities. And it’s now around coding, where you juuust think it’s enough, but actual projects fall apart.
I’ll take some time to gather thoughts.
If programmers are automated with AI, then everything is going to be automated. Then all jobs would cease to exist with some time.
Even physical jobs, as all you need to ask the AI is to create an AI that can do so.
Overall it won't be replaceing software engineers at the moment. AI needs to automate problem solving to do so, and not predict what collection of words you want from the internet to replace engineers.
Something like Devin doesn't work yet, I agree. Chain of thought models can do some problem solving. I've tested that with parsing tasks. LLMs completely replace what a dev does.
It will get better, until when it will replace engeers? 10 years? 100 years? We were supposed to have fully self driving cars some years ago, but they were just promises.
There are fully self driving cars in SF, LA, and Wuhan. Expensive LIDAR and not cheap end2end vision. JEPA and world models will take 5 years. In the meanwhile, focus on making money.