How do you perceive the potential impact of generative AI in software engineering, considering Andrej Karpathy's statement that he writes 80% of his code with 80% accuracy using generative AI?
Sort By:
Oldest
Chief Enterprise Architect in Finance (non-banking)a year ago
A lot is still very unknown. But the way I see it is productivity gain and an assistant to create code while engineers can focus on critical thinking and problem solving and accelerate/uncover new customer problems that couldn’t be solved before. Yes it will eliminate some jobs but create new that we don’t even are aware of. And what will be the organization and engineering team composition of the future with gen AI. Lastly it democratizes and enables citizen developers. CIO in Telecommunicationa year ago
Well said. To add on, it will help solve problems that might not have been cost effective to even consider working on. And to double down on your critical thinking skills - that will increasingly be the differentiator. As AI tools mature and proliferate the differences will be in the creativity and intelligence of the problems they are given to solve and the information they are given to solve it. The adage "Garbage in : Garbage out" will still hold true. You can already see that happening now, and our AI tools are the most primitive they will ever be. This challenge will grow exponentially now.
CIO in Telecommunicationa year ago
Coding has been getting increasing easier for decades. This is the latest in that evolution, bringing the skill/ability to almost everyone.CTOa year ago
The impact I'm seeing in my own work is it speeds up the the authoring process by a lot. I wouldn't say 80% just yet but it definitely helps with writing code after I've written a test to help it along with what I want it to do. You still need to be able to read the code, check it's work, etc. but I also love how these tools help adhere to your style for better consistency and readability over time.The little talked about area that I like these tools for is creating low coupling between components in code. I think people had a tendency to over re-use / DRY their code bases up to optimize for writing code rather than reading it. AI tools make it a lot easier to replicate parts of code that function the same today but belong to features that will diverge over time. Saying it's like an advanced copy and paste is a bit reductive, but I do find that it reduces those moments when you kick yourself for coupling two features together and creating divergence with flags. I find this turns into a mess over time and am happy to see AI make it less tempting to do this.