Irresistible charms
AI is insidiously worming its way into society, despite our conviction that we don't like it
It’s no secret that most Americans don’t really like AI. A recent Athena Insights poll found that 65% were concerned about it, and a June poll from Pew Research showed 63% think it’s moving too fast.
I don’t think these opinions are likely to change anytime soon. But I’m also not sure society’s stated dislike of AI is all that protective against widespread societal shifts. In other words, we might decry AI, but we’re also unable to resist its charms.
A few headlines from today got me thinking about this. The Guardian and BBC both reported on a study published by Cambridge University Press, which found that participants reading short stories rated AI work more highly than human work. Importantly, in one experiment, stories labeled as human-written got a ratings boost whether the label was correct or not, demonstrating there’s still a strong stated preference for human writing, but that preference did not carry over into quality ratings. (In case you share my initial confusion, the experiment used a design that allowed the researchers to separate the effect of the story’s actual authorship from the effect of the human/AI label.)
In a similar vein, Bloomberg covered a survey from the Collective Intelligence Project finding people trust chatbots more than politicians, and that around half of respondents would rather use the former to make “decisions relating to local services such as taxes, permits or licenses.”
I don’t think these findings, alone, are all that surprising or concerning. Chatbots are usually pretty good at distilling overwhelming or jargony information into plain, easy-to-follow instructions. And there are some caveats in the AI writing study worth mentioning: only six stories were sampled, participants were never directly asked which story they preferred, and some parts of the rating scale may have favored clarity (which AI excels at) over literary quality.
Even so, there’s something insidious about AI’s ability to cater to our preferences in ways that are often irresistible. Even as we tell ourselves and others we don’t like it, AI is increasingly being integrated into our societal fabric. Maintaining a low public opinion of the tech won’t help us resist these shifts; if anything, it might act as a false sense of security: sure, we’re going to adopt it to stay competitive and save some time, but don’t worry, we’re still “human-centered.”
The Bloomberg article was careful to distinguish AI-as-an-advisor from AI-as-a-decision-maker, stating that “while people might trust the information that chatbots spit out, only 3.7% of people are willing to let it make the actual choice.” I can’t help but ask: in a world where today’s AI models already out-persuade expert debaters, are we confident that choice is ours?
The analyses and opinions expressed on AI StopWatch reflect the views of the individual contributors and the sources they cover, and should not be taken as official positions of the Machine Intelligence Research Institute.