Congress notices AI's environmental impact, "no" to robot patrols and AI tutors, and some good reads
5 robot-related stories for this week
Congress looks at the AI explosion’s environmental costs
Looks as if 2024 will be remembered as the year the public grokked that AI-in-everything is going to have a real impact on the environment -- directly in carbon emissions and water usage; indirectly in the second-order impacts of making a whole lot of chips, from mining to manufacturing to transportation.
What impact, though? How much carbon from electricity demand, and how much water used for cooling data centers? It's a more complex question than trying to reckon the effect of digital life in total, because AI -- unlike playing Fortnite or mining Bitcoin -- could also contribute to reducing environmental damage. It could improve models, find efficiency improvements in materials, reduce fuel-wasting transportation snags, and so on.
I'll have more to say on this shortly but this week I just want to mention a new development in Washington: Some Democratic lawmakers have introduced a measure, the "Artificial Intelligence Environmental Impacts Act of 2024" (full text here) which would require the EPA to report on AI impacts and set some (voluntary) standards for reporting same.
Annals of Robot Disappointment: Subway Patrolbot Is Off the Beat
New York City's experiment with a patrol robot for the subway has ended. It wasn't a success.
The big camera-on-wheels patrolbot freaked some people out. More importantly, it didn't add much to the capabilities of the police.
Some aspects that leapt out at me, because they frequently come up in robot trials:
Instead of reducing human labor, the robot created more, according to this New York Times article . Human cops often shadowed it, and one cop was glad it was mothballed because he didn't want to be responsible for it.
Part of the appeal of the tech wasn't the use of the robot, but the show of the robot. Yes, one transit official put it, there are surveillance cameras all over the subway, but crooks often can't see them. The shiny robot on the platform "is a reminder technology is watching folks."
Condescending coverage is dumb. Stories with a tone of ha-ha, the robot doesn't work ignore the fact that these are early, early days for non-factory robots.
"We'll Have Flying Cars Before We Have AI Tutors"
Satya Nitta worked for years at IBM to create an AI tutor before ChatGPT and other Large Language Models popped into everyone's consciousness in 2022. But even those generative AIs haven't changed his conclusion: It can't be done.
Here, he explains why he thinks there is no substitute for a human being when it comes to teaching. "It is a deeply human process that AI is hopelessly incapable of meeting in a meaningful way," says Nitta, whose current company instead focuses on creating AI assistance for teachers.
Eldercare Robots that Honor the Humanity of the People They Serve
Some discussions of robot eldercare ask us to believe that limited and clunky robots (some of which were originally designed for stores or factories) are going to be adapted easily to situations that are both extremely subtle and extremely stressful for humans. It's not very convincing.
But robots can be built and deployed in a way that honors the humanity of the people they serve. That's the point of this terrific piece by Kat McGowan about the work of Selma Šabanović, Rens Brankaert and other roboticists who don't think of elderly patients as passive recipients of technological magic.
The piece is a little harsh about robots that do seem to work in some contexts, like Paro the robot seal and the ElliQ "companion" for solitaries. But it's a great account of what robots-for-the-elderly really involves.
AI tool of the week
Here's one that's being used to measure and combat de facto apartheid in South Africa.
Next week: News about some exciting improvements to this blog.
You may enjoy the great book Technology as Symptom and Dream by Robert Romanyshyn, which argues that when western science culture banished magic and spirits it just keeps returning in new forms, including via new technologies such as radio, TV, movies... Now we'd have to add the web/net
Thanks for sharing about the state of AI "tutoring". It all reminds me of the view 20 years ago that the internet would revolutionize teaching and give students the tools to effectively educate themselves. Historian David Noble wrote a damning critique book called "Digital Diploma Mills" - which seems so relevant again.