2026-06-18
• no tags • 661 words
Having bemoaned my lack of inspiration for things to write, somebody has kindly sent a question to answer via e-mail, so I actually have some material to put on this blog! Yay.
The question is this: Which one is better for programming assistance, a human without any experience (you need to teach this guy first and pay him for sure) or an artificial intelligence such as ChatGPT etc?
I think the main point to consider here is that being maximalist on a single index is rarely what we as humans want. Or put another way, the word better here is used in a general way, as if there is a single overarching measure through which we can rank things as better or worse. In reality, we have to ask: is AI better or worse on given indexes that we care about, and then ask, which are the indexes we care most about? Generally, there are a lot of things we care about, even if only a few come to mind when you think about it.
So to make this more concrete I'll talk about two indexes but of course it's nonexhaustive.
If producing code at an increased speed and reduced cost is the index we want to improve, I think it's probably fair to say AI is more useful here. It's pretty good at generating code. Of course this is cost to the market so once externalities are factored in, and especially once VC money and hype dries up, the story may well be different. However at present it does seem to do a good job of it, although human knowledge is still really needed to separate the wheat out from all the chaff it also produces.
However, we may also want to allow for the human environment to flourish, which involves, for instance, interpersonal relationships. AI may be a more efficient programmer, but you cannot go out for lunch with it, you cannot make friends with it, you cannot date it or marry it (well, I believe Musk wants to change that, but anyhow ...)
As a result of this disparity of benefits, if we opt for one over the other, we have to make trade offs. We could be more efficient programmers, or we could have a healthier social situation. Which do we prefer? In fact, we could very well start using AI just thinking of the faster code, and not realise until much later than we've traded away something else we valued more, but simply didn't realise consciously we valued.
I personally think that of these two listed the human element is more important, so I would be inclined towards the human, even if they are less efficient. Other people may think otherwise. If you are invested in AI financially, you'll likely see the prospect of profit as more important than whether or not you can have a satisfying friendship with it.
Of course, these are only two metrics of a long list. You could add deskilling, wealth distribution, mental health, the whole AI psychosis thing, code quality (it may be good at getting the job done, but does it get it done well?), so on and so on ... And, if you share Cory Doctorow's position that code is a liability not an asset, well, then even the benefit of AI becomes a tax.
All this aside, thoughts on things have a strong emotional component, too. Personally, I never was particularly fond of AI (maybe even just aesthetically), and the more I encounter it, the less I like it. I'm considering (like I have with some of the other less pleasant offerings of the internet, like social media, or pornography) taking steps to remove and block it altogether, although of the three it may well be the most difficult to ban completely, as it doesn't make itself known as clearly (I can block chatgpt.com, but can I block all the websites which use ChatGPT in their authorship? Do I have to simply be actively choosier, and more judicious in my selection of content to engage with? Read more indieweb and do less google-searching?). I think it's better to try improve the human environment, and not give in to a convenience that only truly benefits the proprietors of the software.