Published on
In the process of putting this website, I've made pretty extensive use of AI. Initially I was very skeptical of AI. Now, I would say it's the first tool I try. But I am still wary of the code it produces.
While I don't have any hard numbers to support this, I think I would describe it as more useful than not. My reason for thinking this is that I imagine that if consistenly experienced it producing the incorrect output, then I would:
- Not use it. Or
- If I did, I would certainly not reach out to it first.
So my guess is that I've typically found it to be right and useful about 65 - 70% of the time. While this is a majority, this means that it still produces incorrect outputs about 30 - 35% of the time. So being wary working with it and its outputs is justified imo.
Much of the discussion about AI that you find online is not very good. On places like Reddit, you see people make dubious claims like "I haven't written a line of code in months. The AI does everything.". And if you express skepticism of AI, it's not uncommon for posters to reply with simple one word replies like "Cope." A lot of the posters seemingly spread FUD. And they make hyperbolic claims like "Use AI now or get left behind forever!" Although it's not clear why this is the case. And these people never ellaborate. Even when they're requested to do so. It's hard to know if people are talking genuinely about their first hand experience, if they're just shills, or in some cases if they're even people. Many of the posters that you interact with online may be AI bots.
It seems like a thorough analysis that a subject like this should be treated with is lacking online. As a result of this, I may put together a series of articles about my experience working with AI on this blog at some point in the future.