I once tuned in to NPR when they were talking about artificial intelligence, and they were talking about how the seminal figures in the field (e.g. McCarthy) were white men. I reflected that if I had to pick the least interesting possible topic on AI, it would probably be how white the AI researchers were in the 1950s.
> The Dartmouth conference has become an origin myth... Of course, the origin myth served to empower these men to tell their own story. And it's a story full of erasure... We hear nothing in that origin myth about the relationship that AI has to industrialization or to capitalism or to these colonial legacies of reserving reason for only certain kinds of people and certain kinds of thinking.
(later, same show):
> White men wanted to call themselves universal and produce themselves in the machine.
It’s just odd that they feel the need to explain to their audience most professors, and especially mathematics and computer science professors, in the 1950s were white men. Or that a lot of the funds for the research came from industry or the military.
> Or that a lot of the funds for the research came from industry or the military.
I think that's interesting and newsworthy. Maybe because we know it already it seems obvious, but a younger generation might not understand how deeply enmeshed the military-industrial complex was (and to some extent still is) in academia.
It's something you are supposed to learn in history class and there are many, many topics that you should learn from history at some point in your life. A news publication is not a replacement for history class and is probably one of the worst places to get a complete view of history.
To be clear, I have no problem with history podcasts (especially episodes about computing history), I just think they chose a poor lens for the subject matter. The military-industrial complex's influence on computing? Incredibly germane. But, and this is just how I heard it:
> You know, the most disturbing part of the history of AI for me comes from the fact that these men who were working in artificial intelligence looked at those massive, noisy, hot mainframe computers and saw themselves in it. They looked at them and identified a deep affinity that there was something fundamentally shared between their minds and these machines.
> White men wanted to call themselves universal and produce themselves in the machine.
> I think underneath all of that arrogance and hubris is a real lack of faith in people.
> And what I have always found so shocking about the Turing test is that it reduces intelligence to telling a convincing lie, to putting on the performance of being something that you're not.
> ...And in effect, replace God with science?
To me, it felt as if the piece was dripping with contempt for people that actually started the work on the basis that they were nerds with the wrong identity.
Yeah, I wasn't really endorsing the program, just pointing out that "history doesn't belong in news" is a weird critique (not yours) of a history program.
Highschool history class taught me that America won the Vietnam War, because the teacher was a Nam vet and refused to believe otherwise. You can't count on schools to teach anything more than the biases of the teachers. Some of them are great but many are not.
To every extent, I would counter. It's incredibly rare that a person is currently involved in STEM research at a university and some sort of US Military grant isn't providing at least some amount of funding to it/them.
My university for one allowed students to view comprehensive data about grants provided to research groups (after much internal campaigning), but absolutely refused allowing that data to be accessed by the general public. The reason was obvious when you looked at the data: 70%+ of the bucks came directly from various militaries.
> The Dartmouth conference has become an origin myth... Of course, the origin myth served to empower these men to tell their own story. And it's a story full of erasure... We hear nothing in that origin myth about the relationship that AI has to industrialization or to capitalism or to these colonial legacies of reserving reason for only certain kinds of people and certain kinds of thinking.
It would be nice if instead of harping on this point they actually told what they think the missing story is.
If they think that AI is missing certain kinds of thinking - by all means tell me what they are concretely. Like i don't know if i buy that lack of diversity in earlier AI research meant that the AI research only allowed certain kind of "thinking" (AI doesn't really match anybody's thinking regardless of race), but i would be interested in a well researched argument that it did and what those other kinds of thinking are.
I think the problem fundamentally is that these stories tend to be platitudes (lack of diversity = bad) but don't actually go deep into what that concretely means. Ultimately i want something that makes me think about the topic in new ways; you need to dig beyond the surface to do that.
>Ultimately i want something that makes me think about the topic in new ways; you need to dig beyond the surface to do that.
You, a person that already knows about this, will never get that from a short segment on a public radio show intended for a general audience that knows next to nothing about it.
It is intended to make that audience think about it differently, because they have not even bothered to think critically about this at a surface level.
The fact that these models (along with every structure in any society) embeds the biases of those that designed them seems to continue to elude so many commenters here though, so it seems they really need to keep hearing this too.
It’s NPR, all their listeners think about these topics obsessively.
Also, the models embeds the biases of all the people who create content on the Internet. The designers biases come in to l look at with the prompt hacks put in place later to prevent the AI from expressing bad thoughts.
None of these data sets are based on "the Internet". They are a specific subset of the internet, and the training and reinforcements are in no way neutral (because that's not a thing).
They embed biases from the training data, which is taken from the internet at large. The models themselves aren't inherently biased. They're just trying to generate the next token or scene. And these models aren't from the 50s, or made by researchers in the 1950s. The models have guardrails added to try and prevent bias (and other deemed harmful content) being generated.
None of these data sets are based on "the Internet". They are a specific subset of the internet, and the training, reinforcements, and guardrails are in no way neutral (because that's not a thing).
>We hear nothing in that origin myth about the relationship that AI has to industrialization or to capitalism or to these colonial legacies of reserving reason for only certain kinds of people and certain kinds of thinking.
It must be exhausting thinking like this all the time.
Studies consistently show that liberals are unhappier and more mentally ill than conservatives. There was even a big Gallup poll about it this month. One of the floated reasons is that liberals tend to think negative thoughts, as you say.
I think this is the transcript: https://www.npr.org/transcripts/1161883646.
> The Dartmouth conference has become an origin myth... Of course, the origin myth served to empower these men to tell their own story. And it's a story full of erasure... We hear nothing in that origin myth about the relationship that AI has to industrialization or to capitalism or to these colonial legacies of reserving reason for only certain kinds of people and certain kinds of thinking.
(later, same show):
> White men wanted to call themselves universal and produce themselves in the machine.
I mean, seriously?