Hacker News Reader

Comment by theturtletalks | Hacker News Reader

There is a real lag between people who can leverage AI vs those who can't. People who code using AI have a real advantage of using AI in different contexts because they understand how powerful it is.

It's very similar to when the internet got big and those who knew how to use it had a significant advantage over those who didn't. AI hasn't really diffused to mass audience yet, at least not to the point of using it to the extent coders do like making custom and even one-off apps, using CC or Codex instead of ChatGPT directly since giving it terminal access makes in 100x more powerful than the web interface.

View on HN

It's really important to define what "people who can leverage AI" means. I've found that using AI judiciously with close human supervision provides an advantage, but using the latest models with the latest harnesses can look like it's giving you a huge advantage superficially while actually costing you more time in the long run.

That's hard to define, but I measure it as how quickly people reach for AI for complex, tedious tasks. Let's say you're doing market research for a car to buy. A person who doesn't use AI, might just ask ChatGPT, but many websites block searches like that now. Someone who uses CC or Codex could ask it and it will use curl to get the live data and make a HTML you can browse.

Another is someone making a workout plan or a meal plan. Coders will ask AI to make a plan and add it to their calendar using GOG Cli. It's a fundamental shift in thinking and that hasn't diffused to the mass public yet. But just like the internet, it will get to them eventually. In the mean time, those who leverage AI like that will have an advantage.

If you're already trending towards 10x developer status in a niche area of programming, sure, AI will secretly make you less productive (in that niche) while making you feel like you're going faster, for a little while.

For everyone and every other situation, hard disagree.

We're currently wasting a lot of time cleaning up slop created by overusing AI. The loop is: developer uses AI to create a huge amount of code very quickly. But the developer didn't understand the underlying system very well so the specs generated by the AI contain a lot of faulty assumptions and the code was then built on those assumptions. So you have features that probably shouldn't have existed in the first place, but because they do exist people have provided feedback on them, and the developer then uses the AI to implement that feedback, but still doesn't understand the system so much of the code for the revisions is still wrong. So that individual developer saw a huge productivity increase, in the sense that he generated a vast amount of code really quickly, but the team as a whole saw a productivity decrease.

Ironic plot twist, AI is just exposing your coworker's weakness. They could slow down and understand before blasting away at code, or explicitly hang back and generate tech demos/PoCs for architectural choices while building understanding, but they didn't. The only reason things were ok before was because they were handicapped to a speed slower than their ability to figure things out.

Is it a weakness of the individual or is it a systemic issue where management sees how fast a single task can be "completed" with AI so they set goals that can only be achieved if you offload everything to AI and then fall into the trap I described above.

If management isn't giving people enough time to verify work, that's on them. If management is giving people enough time, but saying "go fast" and people interpret that as "YOLO" that's on the employee.

The only advantage in using generative AI is that the market is engaged in a bunch of FOMO over it, so it looks good on your resume. That's it. It makes you far worse at doing your actual job, though. I have to deal with the slop that AI enthusiasts put out at work all the time, and they are going to get fired as soon as AI isn't trendy because their work sucks ass.

Saying that AI is bad for code is like saying a Porsche is bad for driving, without considering that maybe the driver is the one that sucks. It sounds like your co-workers suck.

Reply on news.ycombinator.com