We need to talk about AI

Guest Post by Alex Berenson

More and more people say artificial intelligence will transform the world – in years, not generations. I truly do not know what to think; I’m interested in your predictions (and experiences with AI).

I have about six stories I’m trying to write — about the cancellation of the mRNA bird flu vaccine contract, about Jake Tapper’s absurd efforts to absolve the media for its role in hiding Joe Biden’s decline, about the ridiculous pushback that Robert Kennedy has faced for the modest changes to the guidelines around Covid jabs, about even more terrible behavior by Pfizer, about the long-term structural changes that have made Big Pharma even more problematic as an industry — but I’m finding the debate over how AI may impact our future too important not to discuss.

Comments like this increasingly fill X:

—

I wish I knew whether the doomsayers (are they doomsayers? I can’t even tell, arguably they’re utopians, predicting a future of wealth and leisure) were right.

As far as I can tell, the abilities of these large language models — which we commonly call AI — seem for now to be mostly in the arenas of test-taking and software coding.

They can answer questions well, though they regularly “hallucinate,” but cannot ask original ones. They can mirror the emotional states of the people who use them – making them effective in the kind of fake therapy that merely encourages patients to keep doing what they already are – but not challenge. They can write fluently, but without originality: here is what I am going to tell you, here’s what I’m telling you, here’s my summation, the classic form that middling students are taught to take.

But maybe I am underestimating what they can do? And maybe “for now” is the key phrase on their limits? After all, the models are still relatively new, and no one can doubt their test-taking, information-gathering, and coding abilities have exploded in the last five years.

I want to hear your opinions, to learn from you, especially if you use the engines, doubly especially if you have used them for more than trivial work. Is AI hype, or is it going to remake society? And are the existential risks it may pose real?