Home Creators Posts Import Register Favorites Logout

Content

NOTE: If you want to participate in the conclusion of the book club, which is a live collaborative writing session (more details later but tldr we all get on the same google doc at the same time and write stuff together) fill out this when2meet that I'll use to schedule it. I'll announce the date/time on Saturday June 6 so fill it out before then!

~~~~~

So the thing I was most looking forward to getting out of this chapter was potential avenues for affecting the direction of AI, and I think the main examples we got were:

  • Advocacy against the problematic use of AI in your workplace/field (e.g. Karla Ortiz w/art)

  • Participate in local democratic processes (e.g. San Diegans vs. city surveillance system)

  • Push back on reactive AI policy (e.g. the students at River Dell High School)

I definitely appreciated these specific examples---I think it's useful to see specific types of collective action in different settings to get a better sense of what is possible, and hearing success stories is always nice and motivating.

I guess I was a little less moved by the stuff on regulation. It was interesting to learn about the different structures of horizontal/vertical regulation in different countries approaches to dealing with AI, but something about the whole analysis felt off to me. I think it hit the hardest when reading the last line of Kai's hypothetical future:

Big Tech companies have gotten so rich off of AI that they can easily mold public perception. Academic research and tech journalism are both completely dependent on industry funding, and companies heavily lobby for regulation that keeps new entrants out in the name of safety.

So this is supposed to be the theorized worst case scenario future where we're reactive and fearful towards AI but... it doesn't feel that far off from where things are now? Big Tech companies already have a lot of influence of public perception (I can think of plenty of successful proposition in recent California elections that were totally driven by tech interests but presented as 'for the people' that many people bought into). Idk so much about tech journalism but between DOGE and the Genesis Mission, academic research is rapidly becoming even more depending on industry funding than it was. And last year Open AI was one of the biggest donors to Trump's super PAC, and Trump's willingness to support AI companies has already been pretty clear.

Basically, I feel like the author's conception of the potential for regulation is based in an imagined version of the US government that isn't yet compromised by incentives from the tech industry, which is...not super grounded in reality? That's not to say regulation is a hopeless method for improving things, but I guess I just expected them to actually contend with the barriers that exist to productive regulation instead of just imagining a future where it existed.

Speaking of imagining a future where we've somehow figured out how to manage everything, did anyone else find Maya's imagined future scenario strange? I thought it was weird how many of the differences from Kai's scenario didn't really have anything to do with AI (like getting rid of legacy admissions? I'm all for it but...what does that have to do with technology?) My most charitable reading of this section is that they're trying to say that the version of the future where we figure out how to mitigate the risks posed by AI requires addressing the underling problems that AI exacerbates, which is probably true, but...idk I just finished this chapter wanting more.

That said, I do think I got a lot out of this book, and I think the technical knowledge that it focused on will be well complemented by the more critical analysis I'm expecting from Karen Hao's book next. Anyways, curious to hear yalls thoughts and v excited to discuss the book as a whole in our writing session next week!

Comments

Jacques Vincelette

Fanon might say “Ça finit en queue de poisson…” anticlimactic fish tailed open ending. Or as Carsie Blanton puts it “The future is you. And the future is me. And all of our friends. And maybe even some of our enemies.”

Justin Lee

I agree with you Dr. Fatima, that the author has a rather optimistic view of enactment and enforcement of governmental AI regulation. It may have to do with the political imagination of the author because there is mention of regulatory approaches of other countries like China or EU members and some acknowledgement of limits like regulatory capture by big industry. However, I think this approach focuses mainly on governmental regulation and misses other ways to create pressure and hold Ai companies accountable, from actions of informal affinity groups or other watchdog advocacy organizations which create a more sustained concentration of popular power that can help within the arena of government and beyond it. The youtuber Andrewism would call it "acts of confrontation, noncooperation, and prefiguration" that lead to social change.

Justin Lee

I also agree I wasn't necessarily thrilled with Maya's world which is painted as the preferred future with AI use embedded in our lives. There are clear steps forward compared to where we are now, but maybe I wanted something even more imaginative, radical and liberatory (that goes beyond capitalistic ways of life) but that may not be the realistic future that the author was going for. I think the reason the legacy admission ban was brought up was to show how in the "ideal" future folks learn to live with randomness and uncertainty, so adopting lotteries proves there was in fact a mindset shift. So yeah as you said, it means addressing the underlying problems that AI worsens, including mindset and attitudes towards randomness

AlexTheTutor

As an educator with experiences, I feel like I can write a whole book about education and AI… all these "solutions" I've applied in my own life with great success. I’ll try to keep this brief: 1) AI is destroying our language, LLMs change and will continue to change how people write, how people think, and how and what people say (source: Hiromu Yakura) to the point that it will reduce language into the bland LLM style; everyone who consumes AI will speak and think the same (source: just my intuition). My solution: when age appropriate, have any kid learn to the art of rhetoric, learn how to write, so we can learn how to think. 2) AI is already being actively used in many schools in the US, and so kids accept it as the authority telling them to use AI, so there’s nothing wrong with using AI to write (but ask any recent high school graduate to identify a noun or verb or preposition in a random sentence, and chances are they won’t know and say they weren’t really taught that, or even how to locate states in their own country). Solution: no AI in school, go back to basics 3) Some countries have thankfully banned cell phones in schools, no country is perfect, every country has issues to deal with, but this is an essential first step, what does it have to do with AI? The idea is this: just give kids a break from technology and have them actively converse with teachers for a few hours each day without being interrupted with a notification, it really makes a huge difference, kids can actually socialize and mature, and they don’t develop the habit of always using AI/social media/tech to do everything. Solution: no cell phones in school, no cell phone for anyone under 16. 4) Also, “fun” fact, but in South Korea, middle school boys can take a bunch of photos of their classmates with their phones, access an messaging app, find a certain chat group, upload photos of classmates, pay a little money, and in return get AI generated nude photos of classmates, so… that’s something… not fun 5) I’ve mentioned this to a bunch of superintendents of school systems and teachers and etc, that if you don’t have a plan for AI, AI will have a plan for you, I also mention that take any random teen girl, that 1 out of 3 of those girls when they become adults will have AI generated porn of them (my intuition), and same reaction from parents, teachers, and etc: shrugs, disbelief, dismissive replies, based on my intuition, not enough people will take this seriously until really bad things start to happen. Solution: be proactive. 6) I think of doctor Ignaz Semmelweis who in 1861 wrote that washing hands before delivering babies is good for the mother and baby, and not washing hands somehow kills the mother, but that after his publication the entire medical community banished him as an outcast, it turned out that Semmelweis was of course correct (germs kill, but no one knew about germs), and so when I mention AI to other stakeholders, I definitely do not make any friends, I destroy any chance to network, but AI (if used improperly) is society’s greatest germ, a germ without any antidote. 7) Also not fun: some school systems happily brag about using AI for surveilling students, so not only would such school systems be pro-AI, they would argue that they NEED AI to do proper surveillance, that’s creepy, immoral, and wrong, but what are the chances that all the parents would take their kids out of a school like that versus of just going along with whatever the school does without a second thought… 8) The legal/legislative system evolves far too slowly to adapt legal frameworks in response to rapid changes in AI (dAI/dt > dL/dt). This will create even more pressure on the social contract (people’s expectations on what society should do), and this could either lead to general apathy (meh, nothing we can do) or a growth of extralegal activities, or both, because we collectively don’t have a perfect conception of the law, because lobbying is a significant force, because of lack of time and resources, any legal solutions will be imperfect, will be litigated in the courts, and create a lot of backlash, further creating resentments for the legal system. Solution: exactly what we are doing, be informed, stay up-to-date, think before doing (in other words, before investing in AI or using AI, think about how we should use AI and for what) 9) (an idea I had 9 months ago): either the promises of AI will happen or they will not happen, if the promises happen, then AI will replace workers, and people will lose their jobs, but if the promises don’t happen, then AI companies will either go bankrupted or need to lay off their workers, in either case, the stock market value of AI companies will go down, resulting in further layoffs, so whether the promises of AI will or will not happen, future layoffs are already baked in. Solution: actually teach basic finance and budgeting to kids and adults, using paper and pencil, so kids can actively write out their plans, erase some plans, edit plans, and create by hand a revised financial plan with budget and etc, what does this solution have to do with AI? Basically, the more things you can do with your hands without AI, the better (learn how to do before blindly using AI for everything without thinking), I know this is not a universal global fix to every problem, but this is something that anyone at any time can do with their kids or friends or etc

Laura Robinson

While I think this book was an engaged and thoughtful analysis of the state of AI in early 2024, the wide scale push of AI has sort of made issues come to light that perhaps couldn't be reasonably anticipated. There seems to be an underlying assumption that generative AI will continue to improve in a way that will have large impacts on labour in a variety of careers long-term. But it now seems that tech is running up against limitations inherent to the nature of LLMs and the nature of work. It seems that at least in tech a lot of the most effective resistance has been just asking whether productivity gains are actually being seen, or is the gain from being able to generate code quickly is then lost by having to repeatedly reprompt to fix issues from that generated code and having more critical issues be missed before reaching production. Also there was some assumption that costs would scale down like a lot of computational technologies have, but it seems that largely hasn't happened and in fact improvements in models have required more and more intensive compute. The code generation tool Sora was taken down and an announced partnership with Disney was ended not because of labour organising on the part of video producers or filmmakers but because the videos were very expensive to make and did not get enough of a following for it to make economic sense. Which is not to say that organising is a bad thing, but it may be the case that it's better to do it on the infrastructure side. There seems to be a good amount of grassroots organising around resisting the building of data centres. And as journalist Ed Zitron has pointed out a lot of data centres are financed with relatively high interest debt and so even small delays in project rollout can have a big impact on bottom lines.

phil e the theyby

I did kind of get the vibe that the authors stated that generative AI COULD get better, and in this chapter they seemed pretty optimistic that it WOULD get better in ways that really haven't borne out in reality, and that they also weren't very convincing in making me think that such growth was possible. That said, it's possible that I was misinterpreting the kinds of tasks they expect generative AI would be useful for.

phil e the theyby

I definitely came away from this chapter feeling like we're a lot closer to Kai's world than Maya's. That said, maybe it's just the corner of the internet I'm in, but I kind of feel like a lot of young people...aren't really buying the AI hype? One of the things that feels the most hopeful is seeing graduation speakers who talk about how AI is the future just being booed to hell and back by the students they're speaking to, and THAT gives me meaningful hope. I believe young people are a lot more capable of critical thought than we give them credit for, although they could definitely benefit from support, understanding, and guidance from older generations in that process. I honestly feel like a lot of this shit would wash over if we lived in an earlier era of global capitalism. The obvious truth is that even generative AI doesn't do what it claims to do, and that will ultimately hurt the bottom line of companies that rely on it over time, which will lead them to use and rely on this less (I think the idea that the technology's failures will lead people to stop using it is less true of predictive AI, because I believe the point of predictive AI is more for social control and optics than for actual accuracy). But the kinds of actions corporations and the American government are taking now are very different than earlier times; I feel like those actions truly only make sense if the people making those decisions anticipate a collapse of our existing system. The focus is SO intensely on burning everything to the ground for extremely short-term profit that I'm not even sure that AI will have time to prove its failures before the economic system just collapses under the absurd policy decisions. A note I made at this point when they were speaking about Maya's world: "Since AI mainly automates tasks, not jobs, companies have changed their processes to account for the need for regular retraining, upskilling, and changes to job responsibilities" (288-289) I wrote, "This seems antithetical to the 'go fast and burn everything' ethos of modern capitalism,' and I stand by that. Overall, I really enjoyed this book and thought it was super valuable. The last chapter didn't bother me much, just because the authors' work isn't really on global policy change and full analysis of all of our systems, and ultimately I found the book as a whole incredibly valuable. I'm a lot more able to understand different types of AI and their inherent limitations, and I find it much easier to talk about it with other people than I did before. And I'm saving this for the end because it was more of a funny observation than a big-picture analysis: "But back to Kai's world. Generative AI will likely enable new genres of entertainment apps that are potentially more addictive than those that exist today. For example, it is possible that a user would be able to ask for a video of, say, a 'battle between a T-Rex and a stormtrooper set in a red alien planet' and the app would instantly generate it, perhaps even in 3D" (282) The weird thing is that, on a psychological level, I think this is...probably better than what we have now? To me, the biggest problem with social media algorithms is the complete and utter passivity it inspires, leading us to be dragged along by algorithmic rabbit holes that tap into the most primitive parts of our brains and utterly divorce us from our humanity. Even using AI to generate images/video of something that *you personally come up with*, while simplistic, kind of feels like it's somewhat better just by virtue of it requiring SOME degree of creative thought. That said, I remember reading someone's account of falling into a rabbit hole of using generative AI to make sexually explicit material (maybe it was on here?) and how it completely fucked up their brain, so maybe the simple reality of being able to generate anything ultimately leading to increasingly intense stimuli (whether in terms of sexuality, violence, taboo, etc.) actually is worse; I'm not really sure.

Justin Lee

I find your comments on AI in this era of capitalism intriguing. Im am agnostic as to whether predictive or generative AI "works" per se. The author believes predictive AI structurally doesn't do well but is optimistic about generative AI. So AI can "work" for companies who like cost-cutting and consolidating control, but AI wont work because AI leads to flawed decision-making (including choices of employment or school admissions) and counterproductive effects, going against what many folks consider acceptable (like using AI to eliminate the unique human touch in art or intellectual work). I tend to think of AI as another way to lift the bottom line of companies. I dont know if overreliance AI alone will lead to companies failing, but it's possible. And I agree with you the current system might break even before we see AI's full capabilities I can't say companies operate on fundamentally different principles, it is the logic of capitalism (which I find illogical lmao) to maximize profit no matter what, but the current environment of neoliberal deregulation and austerity has emboldened them, allowing them to go to even further extremes. I agree it seems as if they are operating like they want to destabilize and destroy the current system, and I argue they can. Why? Because they have government that can bail them out with trillions and mouthpiece journalists who will explain away their failures to avoid blame. called the business cycles of capitalism, where the biggest players remain unscathed and the average person gets the brunt of the damage.

Justin Lee

Your comments on the AI generated images/video made me think too. As you said, a big problem now is passive consumption, especially of brainrot, low quality media. I think of Ipad kids, scrolling on youtube shorts, spending their days consuming short form content, and I compare it to my childhood where I would play on my nintendo DS but I also spend time playing pretend with toys and stuffed animals, being creative with storylines and imaginative worlds with my siblings. Other things aside (including moral concerns and to what extent creative thought is required), the generative AI that requires user engagement could be preferred to what we have now. Although of course, it could even better to unplug and try other ways to create. Another case study of the technocracy video where technological solution could help but there are other alternatives that could be more effective?

Rob

a bit late but is AI empire reading going to happen like late Jun or will it be July pick ?

AJ Biffl

My top 3 least likely elements of Maya's world (in no particular order): - getting rid of legacy admissions - increased funding for enforcement, research, and investigative journalism - reforms to enable research to "be effective on the timescale at which new technologies tend to be developed and adopted"