Post
@stefan_hessbrueggen @joshisanonymous @Iris However, these are not new problems. I believe academia has struggled with ways to address the challenges posed by biased data sets and what we learn from algorithmic applications.
However, my background is not in academics, so I'm not very qualified to talk about how academia and science dealt with these issues--but I suspect the rigor of peer review and commitment to scientific principles/ethics were, at the least, moving in the right direction.
2/n
@stefan_hessbrueggen @joshisanonymous @Iris But on the technology side, that's a different question altogether...
the term "artificial intelligence" was coined by a technologist, straight out of Stanford, and within that context that happens to devalue slow thinking and ethical considerations.
Once "data science" was adopted into tech as a way to dilute responsibility ("the algorithm" is objective... math is objective... etc), then that played right into the philosophy that is now so prevalent in silicon valley.
3/n
@stefan_hessbrueggen @joshisanonymous @Iris The "results" of this kind of marriage are well documented, and as I've pointed out elsewhere, well represented in Cathy O'Neil's book "Weapons of Math Destruction" from 2016, way before the "AI" craze...
But the productization of algorithms isn't the only aspect that is troubling... as I'm sure you're aware, the underlying driving force of these current systems is data. And with the strength of capitalistic and monopolistic forces, both aspects of data acquisition are now completely devoid of any ethical guardrails.
If not the wholesale theft of intellectual property, but also everything gained from surveillance tech built into media platforms (such as social media, etc...)--then there's the problem of the data work itself, which has been chronicled and documented in many places.
(This here is perhaps a good place to start: https://dair-institute.org/projects/data-workers-inquiry/)
4/n
@stefan_hessbrueggen @joshisanonymous @Iris When you combine all those things together within the context of what tech leaders are ACTIVELY saying about the technology and what they hope to achieve with it (things like AGI, or massive worker automation, etc...), then it becomes hard NOT to see the fascist imprints in the entire system.
But, like I said, I'm likely not the best person to speak to the problems on the academic side, particularly pre-LLM--but there are a lot of articles, papers, and preprints that touch on this subject, and I'm sure others here can chime in as well.
5/end
@pythonbynight thanks for this expansive explanation that makes clear where you're coming from. I was asking, because, as you certainly are aware, there are completely benign and acceptable uses of machine learning that should not be thrown together with, say, whatever Sam Altman is thinking and doing. Or would you submit a bug report to Thunderbird, because the spam filter is based on a naive Bayes classifier? @joshisanonymous @Iris
Install bonfire.cafe
Get the full app experience
