First, I don’t claim to be an expert, only an informed individual. I read widely from different sources and I work in academia where it seems AI is all anyone talks about lately. I have also read the book Resisting AI: An Anti-fascist Approach to Artificial Intelligence by Dan McQuillan, and highly recommend it.

AI, artificial intelligence, is a misnomer. There is nothing intelligent about it; it does not actually think or analyze anything. AI works on statistical probability algorithms. In other words, a dataset is created with all sorts of information and when you ask a question, the algorithm uses your prompt to look for patterns that create probable matches in the dataset in order to return an answer. This is why AIs “hallucinate“, or as I prefer to name it, make stuff up. Because AI has no intelligence, no brain, to say it hallucinates obscures what’s really going on.

AI has been around long before ChatGPT came on the scene. If you use Amazon’s Alexa or Apple’s Siri, that’s AI. Facial recognition software is AI and “self-driving“ cars are AI. ChatGPT is just another kind of AI. It’s built on large language models. That means it’s fed lots and lots of words from different sources such as Project Gutenberg, your blog if you have one, copyrighted books for which OpenAI and a couple other AI outfits are being sued by authors . If it’s on the internet, it’s pretty safe to say it’s been fed into an LLM dataset.

How does it work? Say you’re in high school and you go to ChatGPT and ask it to write you a five paragraph essay between 500 and 800 words on the meaning of the scarlet A in Nathaniel Hawthorne’s Scarlet Letter, the AI searches its data looking for Hawthorne and Scarlet Letter and assembles a five paragraph essay from what it finds. It doesn’t do any analysis whatsoever, it merely looks at the data it’s been fed and uses statistical probability to assemble an essay from what has already been written about the Scarlet Letter. This is great if you’re a student, but while the grade on the assignment might be a pass, the student has failed to actually learn anything. Studies also show that high-achieving students who who use AI do worse on exams than under-achieving students who use AI.

In the United States, AI can pass the bar exam, the test aspiring lawyers must take in order to practice law. Yet there are a number of instances in which lazy attorneys have used AI to write their legal briefs and then been sanctioned by a judge because the AI cited to cases that didn’t exist.

But AI doesn’t just make up things, it is also biased just like the people who write the algorithms and the data it is fed. And this is where it is most dangerous since a good many people consider it to be a neutral technology.

Say you are a Fortune 500 company and want to make sure the people you higher and promote are going to be the most successful at their jobs. Currently, you’re trying to figure out who to hire for CFO (Chief Financial Officer). So you get some help from AI by feeding into it the resumés of past CFOs who have been supremely successful in the job. Then you feed in the resumés of all the potential people for the position, asking who will be most successful. Since the data you fed in is historically going to belong to well-off white men who attended elite universities and had privilege and access to people and jobs that helped them along their career path to stellar CFO, whose resumés from the batch of candidates do you think will be selected as the best choice?

Will it be the Black woman who attended mostly public schools, worked her way through undergrad and managed to get a graduate school scholarship to a top, but not Ivy League, school? Maybe the man whose family immigrated to the United States from Guatemala when he was a small boy? Will it be the person from Vietnam who attended top-notch schools in Asia before moving to the United States on a work visa? Or will it be the white man who came from a connected, well-off family and attended Harvard as a Legacy and got important jobs at big companies because his family knew people? Companies that use AI to aid in their hiring practices, quickly discover the diversity of their employees, especially at the top levels where they aren’t exactly diverse to begin with, do not increase.

When it comes to facial recognition, AI is terrible at recognizing black faces. There are numbers of instances where black men have been wrongfully arrested and held for hours, even days, because they were mis-identified by facial recognition. Of course, the police are never held accountable because it was the AI, not them who made the mistake.

And that is another danger of AI. Most people have no idea how it works but accept the results as true. When the results turn out to be wrong, who is responsible? The person who accepted the results? The person who made the query? The people who created the data to train the AI? The company who wrote the algorithm? What about those Cruise robotaxis in San Francisco that are no longer allowed to operate because they ran over people? If there had been a driver, they would be in jail. But since there was no driver, the company is under investigation and who knows if any of the victims or their families will receive compensation. Maybe the company will eventually be made to pay some kind of fine for withholding safety information but no one is going to jail.

Dan McQuillan’s book, Resisting AI, argues:

AI is a political technology in its material existence and in its effects. The concrete operations of AI are completely entangled with the social matrix around them, and…the consequences are politically reactionary. The net effect of applied AI…is to amplify existing inequalities and injustices, deepening existing divisions on the way to full-on algorithmic authoritarianism.

Resisting AI, page 3

AI is about quantification and optimization, a neoliberal capitalist’s dream, and it creates a kind of “abstract utilitarianism” as McQuillan calls it, a “calculative ordering” that structures the world in a particular way. Anything that can’t be turned into math gets left out, thus perpetuating discrimination, inequality, and all the other divisions that continue to grow and grow.

Then there are all the things behind AI that nobody talks about because we are too distracted making hilarious pictures of ourselves “in the style of” doing outrageous things we could never actually do.

AI doesn’t magically know what Van Gogh’s painting style is, or what a Corvette car from 1965 looks like. There are thousands and thousands of people hard at work identifying and labeling all the data so the AI can use it. These people are generally in the Global South in countries like Venezuela and Kenya, poorly paid, and forced to work long hours. They get no benefits like sick leave or health care or vacation. They do things like draw boxes around objects in pictures and label them to train self-driving cars. Or they transcribe voice recordings from Siri and Alexa. They tag images from social media and perform all sorts of other tasks including wade through traumatic data that affects their mental health. All so social media companies like Facebook can moderate content, a student can get away with not having to think, and someone can post a picture of themselves as an elf in a scene from Lord of the Rings. If you use AI, you are supporting extractive labor practices.

There is an environmental toll as well. AI requires lots of servers, and servers require lots of electricity. They also require a whole lot of water because servers get hot and if they overheat they go kablooey. The amount of water needed to keep all the servers cool is astounding. And in the United States, places like Northern California, Phoenix, Arizona, and Los Angeles are among the top ten data center locations. Places that are already short of water, not to mention in the case of Phoenix, extremely hot, requiring even more robust cooling systems. There are some who claim AI will help solve climate change, they are wrong.

If these things aren’t enough to get a person to stop and think before using any kind of AI, consider what relying on AI might do to you. If you think the internet has shortened your attention span and Google has made you a little less smart, if you can’t navigate your way around your own city without GPS and you don’t even know how to read a map, what’s going to happen when you start asking AI to write your emails? Or when you let it tell you who you should vote for in an election because you just don’t have the time to do the research yourself? What are you going to do when AI takes your job and instead of using AI to help you work better, you now find yourself as a gig worker labeling data so AI can work better?

AI can be used for good things. In medicine AI is being used to help doctors find cancer and diagnose complicated illnesses. AI is great at sorting through anything where there is a gigantic dataset in physics or other mathematical sciences that would require a human years and years to sort through. An AI can be useful, but not all on its own. AI is not Artificial General Intelligence (AGI) and likely never will be. AI doesn’t think and still requires humans to both feed it and look with critical discernment at the information it produces.

Congress heard testimony in 2023 from various tech people, including the ones who unleashed ChatGPT on the world. Many of them talked about how dangerous AI is (while their companies are making millions off of it). But what they perceive as dangerous—a future with AGI in which the computers decide they are in charge and take over the world—is not the real danger. That is a tech bro fantasyland nightmare. The real danger is in the here and now, to our society, to the environment, to our ability to think critically.

I’d like to say I have never used AI, but since it is being embedded in more and more apps and websites and databases, I can’t say that. I can say I have never willingly used AI, and I refuse to use ChatGPT or DALL E, or anything of that sort. So I’m sorry, you will never see a blog post written by AI here, nor will I tell you how great AI was in helping me plan my garden. When the electric grid goes down and fossil energy is gone, I’d like to still be able to think and write and make decisions and find my way around on my bike or walking without my phone telling me to turn right in .25 miles. I’m not worried about thinking robots taking over the world, I’m worried about unthinking humans.

Reading

• Book: The Comfort of Crows by Margaret Renkl. This is a lovely book. There is a short essay for every week of the year with sharp observations of what goes on in her backyard. And each essay is accompanied by gorgeous art created by her husband.
• Article: Global heating will pass 1.5C threshold this year, top ex-Nasa scientist says. In 2023 we exceeded 1.5C for a day or two, this prediction is that we will permanently arrive there on our way to an even hotter planet in 2024. So much for keeping warming below 1.5C.

Listening

• Podcast: The Great Simplification: Jane Muncke: Perils of Plastic Packaging. Dang this was scary. If you’ve followed me for awhile you know James and I went through a period of moving away from buying food packaged in plastic. There are still a few food things we buy that come in plastic, like tofu, because we can’t buy it any other way. This episode made me think about all the other plastics in our food and home. After our initial purge, we’ve been resting on our laurels, but it’s time to continue to de-plastic our lives. You may have heard recently that there is an alarming amount of nanoplastics in bottled water. When I mentioned this to someone I know who was about to buy bottled water, he shrugged and said, we all have to die of something. It left me rather speechless. While it is impossible to create a plastic-free bubble, especially since the bubble is made of plastic, I prefer to keep as much plastic out of my body as I can, thanks.

Watching

• Series: First Kill (2022). We’ve only watched two episodes I think and the series lasted only one season, but it’s not bad. Two high school girls, one a vampire and the other a monster hunter, fall in love, and so spins out a tale of star-crossed lovers.

Quote

Apocalyptic stories always get the apocalypse wrong. The tragedy is not the failed world’s barren ugliness. The tragedy is its clinging beauty even as it fails. Until the very last cricket falls silent, the beauty-besotted will find a reason to love the world.

Margaret Renkl, “Who Will Mourn Them When They Are Gone?” in The Comfort of Crows

James’s Kitchen Wizardry

It was a hot chocolate sort of day today. The base recipe we use calls for cashews, but because there are global worker safety issues with cashews, we use hazelnuts instead James also added a dash of mint extract. We enjoyed our cups after coming indoors from 0F/-18C, James from checking on the chickens and tidying the coop, me from shoveling the little bit of snow from the front porch and public sidewalk. And if you are wondering how I shoveled snow, as of Friday I am no longer wearing a sling. My fractured clavicle is healed. More on that next time.