Empire of AI: "Nerve Center" and "Dreams of Modernity" (Patreon)
Content
"After analyzing one thousand years of technology history, the authors conclude that technologies are not inevitable. The ability to advance them is driven by a collective belief that they are worth advancing. The irony is that for this very reason, new technologies rarely default to bringing widespread prosperity...Those who successfully rally for technology's creation are those who have the power and resources to do the rallying. As they turn their ideas into reality, the vision they impose---of what technology is and whom it can benefit---is thus the vision of a narrow elite, imbued with all their blind spots and self-serving philosophies. Only through cataclysmic shifts in society or powerful organized resistance can a technology transform from enriching the few to lifting the many."
The aforementioned authors here are Daron Acemoglu and Simon Johnson, who wrote "Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity". next book club pick, anyone?
~~~
something I tend to think about AI is that deep learning itself is a really impressive and (in some cases, not for literally everything ever) useful development in software, but the biggest tech companies are just deploying it in the worst ways. I still think that's true to an extent, but I think these chapters somewhat took the air out of that feeling for me. e.g., yes there are definitely useful applications in fields like climate change and healthcare, but companies like OpenAI aren't just not doing that, their dominance as AGI-motivated above all else is actively disincentivizing it:
"the main challenge of working in [climate change and healthcare] was not technical limitations. It was quite the opposite: persuading talented scientists to focus on problems that necessitated rather simple machine learning solutions, instead of the latest cutting-edge techniques that satisfied their ambitions and looked better on a research resume."
this seems like a clear example of what AI could be, but Altman et al are forcing it not to be. booo
another example of this: for anyone who read AI Snake Oil, did "Dreams of Modernity" also make you think of the part about AI research being increasingly dedicated to reaching higher and higher benchmarks at the expense of catching up the understanding of how AI systems work the way they do? basically the authors attributed a lot of the issues emergent in modern AI as a result of these systems being built not on "alchemy" rather than any kind of scrutable logic. they had made reference to that being a result of the influence of industry, but I think did a lot to flesh the nature/extent of that influence for me:
"[Early on, ] Companies quickly came to dominate over governments and foundations as the biggest funders of AI research and were soon setting the research agenda based on advancements that could also produce short term-profitability... [Today,] The major AI powers are no longer setting the agenda so much as bending an entire discipline to their will."
obviously it's bad that this is happening insofar as it's producing the exploitative manner in which the big AI companies are operating, but it also just makes me sad to see the space for curiosity and creativity shrink (at all, let alone in service of such bs).
it reminds me of the vibe of my old grad department---it was an astrophysics department in the bay area, and like this chapter mentioned, we had a bunch of professors who had either come from tech or were simultaneously working in tech, and most students would also go on to work in tech. this relationship had a significant influence on the culture/values of the department in a way that, to me, felt like it was crushing a lot of actual intellectual curiosity; everything had to be some machine learning model using whatever massive data set, and you did better science by having a better ml model and a bigger data set---it felt like such a novelty to see research that relied on something like a clever way of anticipating some kind of novel signal based on astronomy phenomenology and basic physics, instead of just bootstrapping your way to some statistically significant correlations.
on that note, not to simp for AI bros or anything, but using mannequin challenge videos as data for processing 3D space was kinda brilliant