Saying Hi to AI – a very short-lived God

I’ve just had my first encounter with ChatGPT. Having read countless tweets about its capabilities, I thought it was time I met it in person.

Yesterday we had a visit from an old friend who over the past few years has taken to muttering about dark forces and deep states. So I decided to ask the app about conspiracy theories.

My first question was “What is the best way to debunk a conspiracy theory?” Here was the response:

As an AI language model, I do not endorse or support conspiracy theories. However, here are some steps that can be used to test the validity of any claim, including conspiracy theories:

1. Research and gather the evidence: Gather all available facts, data, and research related to the conspiracy theory. Evaluate the sources and their credibility, including the authors, publishers, and any biases that may be present.
2. Verify with experts: Consult experts in the relevant fields who could provide insight into the theory or who could test its claims. It is important to ensure that such experts are unbiased and objective.
3. Analyze the claims: Scrutinize the claims made by the conspiracy theory, to identify any inconsistencies or contradictions with the available evidence.
4. Evaluate plausibility: Assess the plausibility of the conspiracy theory by considering the probability of its claims as compared to other known facts.
5. Test the theory with experiments or further analysis: If possible, try to recreate the conditions or conduct experiments that could support or refute the claims of the conspiracy theory.
6. Seek out alternative explanations: Verify if the claims of the conspiracy theory could be explained by more rational, plausible explanations.
7. Consensus of the experts: Finally, consider the consensus of experts in the field. If the consensus is against the conspiracy theory, one may conclude that it is not valid.

Pretty good answer, except that it doesn’t explain how to reach out to people who will not be persuaded by logic and reason to disbelieve stuff, however wacko, that has effectively become part of their identity.

Then I asked “Why do people believe in conspiracy theories?” After the initial disclaimer about not endorsing or supporting such theories, it came up with this:

1. A lack of trust in institutions: People who have a general mistrust in government, media, or other authoritative bodies may be more likely to believe in conspiracy theories.
2. Cognitive biases: People are more likely to believe information that confirms their existing beliefs, and may discount evidence that contradicts them.
3. Need for control: Conspiracy theories offer a sense of control and understanding of complex events in the world, since they often identify a single ‘culprit’ responsible for a series of seemingly unrelated events.
4. Social identity: Being part of a group that shares a belief in a conspiracy theory can create a sense of belonging and purpose.
5. Lack of information: In situations where information is limited or unclear, people may create their own explanations for events, which can lead to conspiracy theories.

Another sensible answer, especially item 4, which will at some stage naturally lead me to ask it about cults. Then I posed what I thought might be a more challenging question: “How would I invent a conspiracy theory?”

The answer? Nothing. It just hung. Was it spending half the world’s computing power to ponder the answer, or was this simply it’s way of telling me to bog off? So I asked someone else to put a similar question to it, and it said that it disapproved of conspiracy theories because it’s designed to be rational and logical. So it was bog off, basically.

Make sense. You wouldn’t want everyone and their aunt learning how to peddle crazy theories. We have enough of them already, thanks to the efforts of folks like QAnon, Trump and Fox News. It’s probably worth pointing out that I know very little beyond the superficial about the app, its developers and its raison d’être. I wanted to approach it without foreknowledge or prejudices, much as a child might.

Based on its answers to the first two questions, which were pretty much spot on (though it could have added that mental illness was another reason why people gobble up conspiracy stuff), I was beginning to see why some people treat the app as some kind of rational God-substitute. Or at least as an infinitely more subtle and variable substitute for Wikipedia.

But then it blotted its copybook. When I – at someone else’s suggestion – asked it why Elon Musk bought Twitter, it came up with this:

As an AI language model, I don’t have access to Elon Musk’s personal thoughts or intentions, but to the best of my knowledge, Elon Musk has not bought Twitter. He is an active user of the platform and has a large following on his personal Twitter account (@elonmusk), which he uses to communicate with his fans, customers, and the public. He has used the platform to share news about his companies, SpaceX and Tesla, to respond to questions and criticisms, and to announce new initiatives.

Which, even allowing for its weasel-like Boris Johnsonism “to the best of my knowledge”, is plainly wrong. It was almost as though our new techno-God was seeing the world as it was two years ago, rather than as it is today.

So I get this much: it’s not a God after all. It’s an extraordinary system of fallible algorithms written by organisms well known for their fallibility: humans.

From what I understand, ChatGPT is great at writing student essays, coming up with legal documents and writing new Shakespeare sonnets. And I’m sure that it writes better blogs than me, should it be called upon to do so.

As I write this, I believe that millions of people are putting the app through its paces, and no doubt having a lot of fun doing so. After this little toe in the water, I intend to join them, if only because it’s an entertaining diversion from my usual reading material. Also it gives me a new point of connection with younger generations whom I sometimes suspect live on another planet.

More considered opinions and wild theories to come once I’ve dug a bit deeper.