Something true enough (on accident)
I have experimented casually with a small variety of Machine Learning-"Generative AI” tools, and work with folks who have tried many of them.
I have been uninterested in having robots write for me, but I have leaned on them for graphics and images. I am also mindful of many of the historical biases baked into these systems (because these systems know only what has been, and have no ability to imagine what can be).
My most advanced usage was in the development of an event poster, which I composited from several robot-generated outputs. (One can assume unpaid humans were also involved, without their knowledge or permission.)
More people (and companies) continue to lean on these tools without understanding what they are or what they’re good for. Google results are continuing to be polluted with hallucinations. This week both Ana Navarro-Cárdenas and Esquire Magazine cited historical presidential pardons of relatives that never happened, but which they had “learned about” from ChatGPT.
Reminder: ChatGPT and other LLM robots do not “know” things; they spit out sequences of words that are probabilistically similar to words people (or other robots) might have written in the past, in response to questions that the robots deem similar—even if those earlier answers were false, fictitious, or about wholly different topics. Do not turn to ChatGPT for facts or medical advice.
It would be the equivalent of asking some guy named Randy you met last week waiting in line at the auto parts store. I mean, it’s not impossible he might come up with something true enough (on accident), or that he might randomly happen to know a fact or two of relevance (that he himself picked up from a woman he went on two dates with). But the odds? Not that great.
(Sorry, Randy. You seem nice.)
One @#$%ing thing
Tonight I cancelled my subscription to ChatGPT I’d forgotten about (required for access to Dall•E image generation). Either until I need it again, or maybe never. Sam Altman doesn’t need my money, and I can afford to hire an artist on commission for next year’s dance.
All the @#$%ing things
Night 28: Donated money to three orgs
Night 27: Addressed a hazardous tile floor
Night 26: Picked up trash with the Trash Falcons
Night 25: Learned more about Pete Hegseth than I wanted to
Night 24: Canceled recurring subscriptions I no longer need
Night 23: Dwelt in gratitude
Night 22: Picked up pie from a favorite local business
Night 21: Downsized my clothes closet
Night 20: Increased my monthly contribution to the ACLU
Night 19: Deleted a blog from two decades ago
Night 18: Researched nonprofit board opportunities
Night 17: Contributed to Trans Lifeline
Night 16: Spent time together with loved ones
Night 15: Bought from a not-for-profit online store
Night 14: Refined an icon and wordmark
Night 13: Contributed to the LGBTQ+ Victory Fund
Night 12: Contributed to The Guardian
Night 11: Read, reflected, and rested
Night 10: Sent money to support vaccinations in Nigeria
Night 9: Sent money to a friend in need
Night 8: Gave gifts and spoke words of appreciation aloud
Night 7: Contributed to a California-focused nonprofit newsroom
Night 6: Made homemade donuts for my team
Night 5: Opted into a paid Buttondown tier
Night 4: Reviewed my local election results
Night 3: Deactivated my X account
Night 2: Contributed to my local nonprofit newsroom
Night 1: Started by starting
Words, sorts, thinks, and actions by Chris Ereneta, from Oakland, California. Thanks for reading! Thoughtful feedback and questions are welcomed at that.often@gmail.com