I have a friend who is currently doing an online AI business school program at MIT. This is the counter-curriculum I wanted him to read. I'm posting it here so he has no excuse not to consult it, and so it's on-hand in case I want to share it with others. I'll probably continue to update/add to it.
-
Resisting AI (2022) by Dan McQuillan
What if the current AI trend is not only kind of stupid, but also actually very dangerous in that it is inherently fascistic and will increase fascism? It's a book, but it's short and easy to read for an academic text. You can find it online for free, I imagine.
-
James Mickens USENIX Keynote (2018)
USENIX Security '18-Q: Why Do Keynote Speakers Keep Suggesting That Improving Security Is Possible?
James Mickens is smart. He's harsh. He's funny. He hates stupid people doing stupid things. Remember this is from 2018!
Bonus Mickens: this very funny column about how bad smartphones are.
-
Marcus on AI (ongoing)
Gary Marcus's substack He is a longtime AI researcher and gadfly. He's a believer in trying to achieve AGI, but thinks current hype and direction is wrong, so he does a very good job deconstructing hype. Reading some of his writing years ago was really helpful for demystifying all the singularity pipedream stuff I encountered when I was younger. He posts to his substack very frequently, and lots of industry stuff (currently a little bit too much coverage on Sam Altman), but it's still good.
-
Marcus on Deep Learning (2018)
Classic recent paper from Gary Marcus on how deep learning/LLMs are not a road to lead to AGI.
-
Ted Chiang in the New Yorker (2023)
He's a sci fi author best known for the story that Arrival (2016) is based on. He's also a technical writer. So he's a very good and clear writer.
These two short pieces were both published in 2023 and are excellent. LLMs like ChatGPT are more analogous to lossy compression than actual intelligence, and the way such systems are being developed will just serve to "sharpen the knife edge of capitalism".
-
Bonus, 2021 piece on why the Singularity will probably never come: Why Computers Won't Make Themselves Smarter (no paywall link)
-
Double bonus: There is a different blog post titled "Will AI Become the New McKinsey" from a cryptobro-turned-AIbro investor-bait site that is a completely unaware straight-faced that reads like it was actually generated by an LLM. Either incredibly depressing, or kind of a funny prank (sadly I think the former). Do not look into this guy, you will regret your choices.
-
Artifice and Intelligence (2022) in Tech Policy Press
Artifice and Intelligence digs into the problem of language, the problem of calling LLMs and such "artificial intelligence", including good recommendations for language to use.
-
The Elegiac Hindsight of Intelligent Machines (2023)
Chapter from a book by Baldur Bjarnason. I don't know this guy, but I enjoyed this essay. A critique kind of rooted in how businesses operate with regards to emerging tech like LLMs. I think his POV is how software development/the software industry should do things.
-
OpenAI Used Kenyan Workers on Less Than $2 Per Hour to Make ChatGPT Less Toxic (2023)
What it says on the headline. Important detail about the datasets this current models use that is swept under the rug. They don't work without massive amounts of outsourced labelling.
-
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? (2021)
I actually really dislike the "Stochastic Parrot" device/analogy itself, but this is still the classic critique paper, referenced in basically every other link here.
-
The Fallacy of AI Functionality (2022)
Many deployed "AI" products simply do not work well. A review of this fact, and some of the consequences.
-
ChatGPT, Galactica, and the Progress Trap (2002) (no paywall link)
Another critique pointing out the serious consequences when LLMs fail to deliver as promised, as they often do.
-
Someone smart organizes a reading group at Iffy Books in Philly. The resources section has a some useful links, but especially the readings and resources for a previous series on Resisting A.I. (the book at top of my list) are really good. Many of these links are there already, alongside many many more!
-
Superintelligence: The Idea That Eats Smart People (2016)
Keynote from Maciej Ceglowski. I don't back ths 100% but it's funny, and makes good points. Kind of a teardown of Bostrom ideas.