Published on [Permalink]
Reading time: 2 minutes
Posted in:

ChatGPT as a chatbot is quite good, actually

One very good use case for ChatGPT is… actually chatting. Who knew?

I was waiting for someone at a restaurant and had 10 minutes by myself — a rarity these days. Instead of scrolling through one social network timeline or the other, I opened up the ChatGPT iPhone app, and learned that the only Leonardo painting in the Western hemisphere was right at my doorstep. Also learned that there are only 20 paintings by Leonardo and more than 300 by Rembrandt, and a few other tidbits.

Are any of them true? Well, trust but verify as they say, but why should I trust it any less than a random X account? Or worse yet, a non-random threadboy — now there’s a useful word I saw just recently — who posts unsourced graphs and opinions-as-facts.

Now sure, X and other social networks have upsides, like bonding with fellow humans over topics of joint interest. But these are for the most part shallow connections, empty calories for our socialite stomachs. A more stable and sustainable equilibrium for most people, certainly for me, could be real-life interactions for socializing and algorithms for tickling the mind without any pretense that we would be buddies. Social networks try to be both but are not great at either, the same way pickup trucks try to be both a car and a truck but are mostly gas-guzzling parking spot-hogging behemoths, and at the same time the most popular vehicles in the US. So, a perfect analogy.

From this perspective, ChatGPT’s forgetfulness is an excellent feature. It remembering prior conversations would bring it a step closer towards parasocializing, making it even worse than human social networks. I have no doubt that the feature is coming any month now, if it’s not already here. If and when it comes will the the point when X/Threads/Bluesky et al should sound the alarm — or introduce friendly algorithms of their own.

✍️ Reply by email

✴️ Also on