Prioritizing chatgpt development

my chatgpt memory is full! i only started using it 3 days ago. my own personal memory isn’t so great, so i have numerous google docs of saved quotes. here’s one from my doc “jettison”…

“I would speculate that forgetting might be the default system of the brain,” Davis said at the neuroscience meeting. “We might have a slow chronic forgetting signal in our brains that basically says, ‘Let’s erase everything,’ unless a judge … comes to intervene and says, ‘This memory is worth saving.’” - Tom Siegfried, A Leaky Memory May Be a Good Thing

that atlantic article is paywalled, but it didn’t used to be.

imagine if the atlantic announced, “we’re going to erase everything! except for the 100 articles that receive the most donations.” everyone could be a judge. you’d have the opportunity to say with a donation that siegfried’s article is worth saving.

which atlantic article would receive the most donations? how much money would be donated for it? is it possible that using donations to prioritize articles would eliminate the need for paywalls?

we humans have successfully colonized the planet. but think about moving, and how hard it is to decide what’s worth saving. nowadays it’s basically possible to bring nearly everything with us, but typically we engage in lots of prioritization. back in the day though there weren’t any trucks, trains or big boats. there were carts and canoes. before that there were bags and sacks. and initially it was just our arms. our ancestors couldn’t carry everything, so they had to make very difficult decisions about what was worth the effort. they weren’t equally good at making these tough decisions. just like they weren’t equally good at carrying things. just like they weren’t equally upright.

this is the only logical theory about the development of human intelligence. but virtually nobody knows it, which is funny because there’s lots of optimism that human level ai is around the corner. as far as i know, there aren’t any experiments with multitudinous and diverse robots each deciding the optimal combination of batteries, solar panels, external hard drives, spare parts, tools, and so on to carry across a giant desert. robots need to accomplish this before they can start deciding what to put in spaceships to carry across the universe.

few of our ancestors migrated on their own. most did so in small groups. nobody in a group benefited when members made carrying mistakes, so communication became extremely useful in this regard. the most important form of communication was trade. this effectively informed group members of the actual carrying trade-offs. was a large tuber worth carrying? that depends on how many rabbits or tools you could trade it for. no two groups equally engaged in trade and this determined their development.

and here we are. we want chatgpt to carry more of our information, but what are the trade-offs? we don’t know. there’s no opportunity to trade money for more memory. the demand for memory is unknown. so how is the supply possibly going to be correct? it’s a certainty that chatgpt is going to carry the wrong amount of memory. this mistake won’t benefit anybody in our group.

and it doesn’t help that i can’t see which information you want chatgpt to carry for you. transparency should be the default. if you want to mark anything as private you should have the option to do so.

personally i hybridize figs. to my chagrin, chatgpt didn’t know that ira condit was the 1st person ever to cross two ficus species. 75 years ago he crossed ficus carica and ficus pumila. i told chatgpt to save this information. just like with the atlantic articles, we should all have the option to use donations to decide whether condit’s accomplishment is worth saving/remembering. naturally chatgpt didn’t know that last year i was the 1st person to successfully cross ficus carica and ficus opposita.

chatgpt should be very useful at helping us to decide whose shoulders we stand on. and which trees we bark up. but this depends on everyone’s specific input in the form of donations. without this essential metadata, chatgpt’s growth and development will be severely stunted, and too many people will stand on the wrong shoulders, bark up the wrong trees and save/carry the wrong things.