I probably don't have followers who are fans of the Global Disinformation Index, but in case I do, I have a question: these "adversarial narratives" the GDI fights in coordination with govt's & Big Tech, are they subaltern or hegemonic? disinformationindex.org/mission
True story: I was thinking about my grad school days when James C. Scott was a big deal & we were all "subaltern" this & "subaltern" that, & as historians we were trying to recover these subaltern narratives in the the texts & so on.
I was a student of apocalyptic literature, which is very much a family of subaltern narratives! So I was thinking about the GDI "disinfo == adversarial narratives" frame, & how it's just this naked bid for hegemonic narrative control -- totalizing narrative, we might say...
... and ALSO about how that same hegemonic narrative is presented as defending the weak, marginalized, colonized, subaltern. It's a total inversion of actual reality in a way that's just in-your-face & daring you to object. It's a wolf on TV saying, "speaking as a sheep..."
"Generative AI just remixes the past, & can't truly innovate." No, that is wrong. Here's why: jonstokes.com/p/yes-generati…
I put on my thinkfluencer hat to offer a simple, four-part definition of innovation as a kind of pyramid of qualities that I think generative AI will get to the top of:
Generative AI lets us turn electricity & transistors into novel + interesting symbol combinations at scale, thereby massive expanding the base of this pyramid:
As someone who has been following critiques of models since the models in question were financial models (mainly credit & risk scores), I am doomed to see the same take endlessly recycled again & again.
What's sad is that this is a real (if obvious) critique of credit scoring models, whether DL-based or not (even the least sophisticated credit risk model will reinforce existing inequities to some degree). But applied to generative AI this take is just wrong.
There's a totally uninterrogated metaphysics of creativity & innovation inherent in this tweet, & I strongly suspect it's paired with a very superficial & mostly wrong understanding of how generative AI actually works.
It stresses me out too, man. I feel like we face a choice between near-term chaos (b/c the genie is out & there are few or no guard rails) vs. a centralized power so overwhelmingly capable that it's effectively the end of human liberty. reuters.com/business/autos…
It also stresses me out because I want to type words about it & read about it, but there's too much of it & it's very hard to make sense of, so I gotta manage burnout. As I said in a prev thread, the pace of change is very reminiscent to me of "crisis mode" scenarios I've seen.
I keep thinking about this new MSFT model & how good it is. Here's what's stressful: this is this good & it hasn't even begun to be commercialized yet. If the ecosystem just stopped here & spent a year building products on it, it would be very disruptive.
.@elonmusk to really get this done right your AI lab needs a proper chaplain. I have an M.Div from @HarvardDivinity but resisted the woke mind virus there so am uniquely qualified to catechize your bots. My chaplain services are expensive but as they say, “buy once, cry once.”
.@rdnxyz gets it. The classic libertarian "public vs. private" bright line works against liberty in the age of private planetary scale networks, especially when those "private" network owners collaborate with the public sector to protect incumbents on both sides of the divide.
@rdnxyz What matters in the age of FAANGs is not public/private, but size, rent-seeking behavior, regulatory capture, and regulatory moats. Massive private networks & governments merge into one entity at these points. It's not even a symbiosis or partnership -- they're the same blob.
The question is: who will decentralize these centralized entities (against their will), & with what tools? In the previous decade, I'd have said antitrust -- set a leviathan to fight a leviathan. Now? I think we can win this fight with open-source + p2p networks + web3 + AI.