My Policy on AI

Another day, another Slashdot story about AI, this one headlined "AI-Generated 'Slop' Threatens Internet Ecosystem, Researchers Warn" So I guess it might be time to chip in my £0.02 worth, just for the sake of having something I can refer to. Incidentally, you can tell that this blog isn't AI slop - if I was using AI to generate posts, there'd be an awful lot more of them!

The simple version of my AI policy is this: I am not willing, in my professional life as a software developer to work with technologies that I would actively avoid in my personal life. For a while, that list has included blockchain and anything crypto(currency)-related (see Web3 is Going Just Great and Paul Krugman's post today - crypto is a scam). Now, I've added AI, and generative AI in particular.

In my humble opinion, almost all AI is morally and ethically abhorrent. I give the following reasons:

  1. Generative AI is, by design, a bullshit generator. Its sole purpose is to generate content which looks plausibly like it might have been created by a human being. Since it cleaves to no particular moral creed, it feels under no compunction to ensure that the content it generates is truthful, accurate or honest. It just produces whatever it thinks will satisfy its user. Unfortunately (or fortunately, depending on one's point of view), this enables it to evade our fraudulence-detection mechanisms which have hitherto been predicated on the assumption that generating a convincing fake image, or fake document, or whatever, is hard. Generating a pamphlet arguing convincingly that climate change is a liberal fraud, or an image showing a politician in flagrante delicto with someone they shouldn't be is now a simple exercise in finding the right prompt. And you can bet that every wannabe ratfucker on the planet is salivating at the prospect.
  2. Running AI models takes large amounts of energy. Given that many of our current methods for producing energy exacerbate climate change by releasing carbon dioxide, this is not great. To be fair, this objection would be ameliorated somewhat if we could replace our energy generation with carbon-neutral methods, but we are not there yet. Therefore every time Google or whoever adds some shitty AI-powered search result, it is helping to kill the planet just that little bit more.
  3. Training AI models involves vast amounts of data, and it seems pretty clear that a lot of this has been stolen. To make this doubly wrong, it is also clear that many of the people pushing GenAI are doing so in the hope that they can replace the work of genuine artists with AI-generated facsimiles. The aims of these people (as I see it), ultimately, are twofold: a) you don't have to pay people for selling AI-generated work and b) you have control over the work - no danger of any "unhelpful" moral messages from bleeding-hearted artists creeping in!
  4. The other thing that the GenAI boosters are keen on is destroying jobs. Why would you pay a graphic designer, when you can just feed some prompts to an AI image generator? Why would you pay a copywriter, when an LLM can churn out some passable text? Apologists like to point out that job-destroying technologies have come along before, and given rise to new jobs to replace the old, and therefore we shouldn't worry. Unfortunately, this inductive argument fails - just because a thing has happened before is no guarantee that it will continue to happen. After all, I have successfully not died on every day of my life prior to this one - therefore I will never die...? Gradual replacement of old jobs with new ones we can cope with. The sudden and swift decimation of jobs across the board - perhaps this will be more of a problem.

For these reasons I avoid all AI in my personal life. I have an iPhone, but I took great pleasure in denying Siri access to all of my apps, and then switching it off altogether. When Google introduced AI summaries, I switched to DuckDuckGo (who also have AI summaries, but you can disable them). If you want to ensure I will not buy your product, advertise it as having AI features (if you're lucky, you may be able to sell me a 10 foot bargepole, with which I will proceed not to touch your product).

Finally, if you want a short and pithy summarisation of my views, suitable for printing on a T-shirt, it would be this:

Man makes God in his image, and the people building GenAI (Altman, Zuckerberg, Musk et al) have built one in theirs: utterly devoid of morality.

Home