Tonight I’m thinking about how to talk to machines.
Of which—machine talking, I mean—I realize I now spend most of my waking hours doing. On the more obvious side, I converse with language models: this past week included a lot of playing around with DeepSeek, using Claude for some reasoning experiments, & prompting GPT many many times every day. But I also talk to my machines in a variety of other ways that add up to my (probably) >10-hour daily device screen time: I tap, ask, share, touch. My hands spend more time resting on the metal of my laptop than they do anywhere else.
All these actions interest me. playing around? conversing? tapping? I’m cognizant that machines aren’t just conduits by which I interact with other humans; no, I spend much of my time talking to the guts of machines and models themselves, & considering what they have to say.
Some thoughts:
In the UI world, designers have hold a keen eye towards optimizing interactivity — tapping, swiping, etc. — to improve the user experience. The physicality of these gestures interest me. I wonder if they also improve the machine experience as well?
Machines learn. In AI today we discussed genetic algorithms, which introduce randomness and diversity into problem-solving. We looked at crossovers and mutations in chess board configurations, which really looked like the meiosis diagrams I remember from AP bio.
there’s a lot to learn from machines; we could benefit from spending more time thinking about, engaging with, experimenting on, and critically examining them.
And what spaces could be opened up if we integrate machine lingo into thinking of humans with machine lingo? Machine learning is a field decidedly driven by growth mindset and a can-do spirit — gradient descent! optimization! For all my qualms about techno-accelerationist communities, I am inspired by the optimism that pervades them. I sometimes wonder if we’ve grown more confident in machines' capacity to learn than in our own: what happens if we reroute this faith—this belief in incremental progress, in the power of iteration—back to humans? If we saw ourselves as capable of continuous learning, fine-tuning, adapting in real time?
Also, related line of work that interests me is imaginative, forward-thinking computing. Software / the Internet / machine learning models are our creations. How can we knead the digital into shapes that delight? How can we infuse joy, playfulness, and community into our use of technology?
I suspect that attaining such joy hinges on holding respect for machines: a willingness to spend the proper time peering at their organs and Elphaba-ing them into more beautiful forms. I am particularly inspired by the likes of Spencer Chang and Andy Matuschak in this area — their work in shaping tech with intention and care.
(note of caution to not take my anthropifications too literally! there are lots of complex questions in this area & my late-night brain has tripped a little over them)
(Inspired by Laerdon’s post a bit ago)