I’ve dropped off my wife at her friend’s apartment in Long Island City, found some parking, and set about wandering the neighborhood. It had been about four years since we moved out to pick up life in the burbs - not terribly long in my mind, but I seriously did not recognize most of the neighborhood. When we had left, LIC felt a perpetual work in progress (everything was coming soon!), but now it felt so much more like it was nearing its end form, as evidenced by the plethora of strollers blocking the sidewalks.
I stopped over at PS1 (woohoo free for NYers) to take in some of the recent works but also to mull over two things I’ve been thinking about as I bake in the James Turrell room (the one with a big skylight. Super smart decision on my part).
One has been the reaction to ChatGPT’s model 5 release and how many people were devastated by the removal of the warmth/personality attributes in the language model’s responses. I think most folks had read the headlines or know people who interact with LLMs as if they are friends (ok fine, maybe just acquaintances) due to the tonality of the responses from the model. I must admit, I’m a little bit disappointed (though not surprised) about OpenAI’s decision to roll back some of that personality as default in the new model. We know that humans naturally crave connection with others, and if we can get connection without consequences, we (as a whole) will do that. The parallels to what is happening with these AI tools and what has happened with social media is obvious. I mean, OpenAI even copied the same “hey you might want to take a break” warning as a way to indicate responsible design. It’s not going to stop people from using it. Yes, while some pundits point out that it is totally possible to customize your model response to remove these kinds of emotional bait, the average user doesn’t know how to prompt in that manner. People see a chat box, and they just proceed like they’re used to - therefore, defaults matter. (Ultimately, I hope these products start to roll out a more deliberate onboarding process to actually educating people on how these powerful tools work, as opposed to pushing people in to the deep end.)
The other thing I’ve been thinking about, related to the first, was about Bowling Alone. Harkening back to my political science days, I wonder if we’re living out the extrapolations of Putnam’s observations today. Rates of loneliness has increased across the board, despite technology make it easier than ever to form connections; this is especially true for young men. Societies have gotten more isolated; not just in the United States, but globally - clashes between political and economic classes are more visible than ever. These macro trends seem to indicate that Putnam’s observations have continued to extend. And (related to the above) when connection is so easy in the digital world, do we have hope in reinvigorating the communities that are central to a well functioning society? I’d say yes…if we take both a longer time horizon and we zoom in at what’s happening on the ground level. I think Putnam’s original observation was focused on a certain permutation of community succeeding in order for society to flourish. But our society is now, on average, so much more educated and diverse that I think we’re just in the throes of figuring out how to bring together so many different communities and find the commonalities. It’s just difficult. But I think something we can do. Maybe it’s because my sample set of young(er) people is skewed - Gen Z folks are some of the most astute, empathetic humans I’ve ever met - I think we have the capacity for and the desire to connect with actual humans, even if it may be difficult or uncomfortable. We’re just figuring it all out right now, and it just feels bad.
I suppose I’m writing this for myself, as I continually worry about AI and its impact on me and the people around me long term. Gotta take a longer term view and trust that the dust always settles (unless you’re on Venus).