Silicon Valley is a strange place and allegedly full of transhumanists (basically cyberpunk for real people). As far as I understand it, this means humans are flawed beings and building machine gods to transcend our own deplorable meat existence is a good thing. Perhaps I am exaggerating here, but the rather strange statements emanating from AI-related rich people with massive platforms are so resoundingly odd that I am less inclined than usual to bother with subtlety and nuance. These days, the United States has become the source of excellent screwball comedy, and that is just the political and business news. Hard to compete even for SNL, I imagine. It does mean that my rather serious podcast fare is far more amusing than it has any right to be. And then, inured as I now feel from all the drama, real life and legacy media suddenly wakes up to the fact that all of this might mean something for academics, too. This follows on from some posts earlier in the year about AI and social science. While the excitement and anger online appear to have settled down, it might just have reached the wider (academic) world now. My summaries of the earlier AI anxiety: Also: this morning History in Organization has 799 subscribers! I’m a sucker for round numbers, let’s see if it will happen ;-) Should we all be worried?Well, not all of us. There are real differences among disciplines, approaches, and generations, and, of course, a lot of opinions. The Atlantic, currently a purveyor of excellent middle-class anxiety-inducing academia rage bait like “The end of reading is here”, or the similarly influential “The college essay is dead: Nobody is prepared for how AI will transform academia”, has been at it again: “AI-generated articles are flooding academic journals”. As far as the end times go, I imagine the wider world feels generally ok about this. It only briefly interrupted my online shopping for wind-up FM radios, battery-powered torches and water purification tablets for our emergency preparedness kit here in the UK. What I found most curious about it is that the professional debate has been all over this already: the economists modelled it; the organization scientists analysed it; most major journals in our field have been aggressively recruiting new editors to deal with volumes that have been rising for years, but have suddenly gone on a new exponential. Last week, I heard about the BBSRC (a UK research council) debacle that led to the knee-jerk institution of demand management. Very, very messy apparently. The Atlantic piece summarised the wider problem well: “Reviewing a growing number of submissions with a critical eye now takes time that minimally compensated or unpaid journal editors, including me, do not have.” Ditto unremunerated reviewers of papers and grant applications. Don’t even get me started on academic publishers licensing our work (if not already OA) for dear money to the AI companies. Another time… But what really stumped me was this claim by the author, apparently a medical ethics and health policy expert:
Let’s look for a moment at what is happening at RetractionWatch and Nature, which recently reported the largest retraction event in one day: 1,400 papers by IEEE, or the Institute of Electrical and Electronics Engineers, for links to paper mills. Meanwhile, maths is reeling from recent breakthroughs in which AI solved tricky maths problems most of us have never heard of and the practical applicability of which nobody has yet bothered to explain to me (in fairness, it is unclear how hard I would listen to it). Instead, emotions run high from the end of maths to questioning whether it is all smoke and mirrors. I genuinely have no opinion on any of that. I just thought it is worthwhile highlighting that none of this appears at all connected to the humanities and social sciences. I mean, “even in the hard sciences”? I think what the author of The Atlantic article meant to say was: exclusively in the sciences that think of themselves as “hard”. All valuable knowledge is hardIn German, we have a humorous and nonsensical putdown for those going all Nietzsche on us: “Nur die Harten kommen in den Garten.” (Only the hard ones get to go to the garden.) It means that posing as a hard one comes with only questionable benefits. In academia, that is not quite true, though. Being in the hard sciences, or the next best thing, a quantitative scholar anywhere less hard, provides you with an unshakeable sense of status and superiority. Which is presumably why it has been the holy grail of transhumanist endeavours — though shockingly delivered via a lot of language. But such language is sublimated into weights in large language models, therefore presumably restoring the primacy of numbers. Mathematical, logical understanding is valued highly, remunerated handsomely, and at one time identified by the ESRC (another UK research council) as a priority area for developing more capability … you get the gist. Mathematical-logical intelligence is often equated with intelligence in general. When I did my one session of teacher training as a doctoral student at the University of Liverpool in the early noughties, we were introduced to Howard Gardner’s model of multiple intelligences. I found it eye-opening at the time. If you don’t know about it, you should definitely learn about it. He proposed a range of different intelligences, highlighting that academic education favours certain ones over others, but that this is a convention, not a “real” order of value. In 2008, I was at Harvard and joined the faculty procession when Drew Faust became the first female president of Harvard. My friend Anette started talking to another academic, and it turned out to be Howard Gardner. I was very starstruck. Silicon Valley clearly favours the hard stuff, too. So the capability of current AI models is demonstrated by solving hard maths problems. Not by writing the new War and Peace. Clear answersI was actually really good at maths in school. Straight-A student, you would say here. Which meant that I paid no attention in class, other than noting down the homework, doing it all the day before the exam and passing with flying colours. (Really annoyed my schoolmates.) What I took from this experience was a resounding disdain for any scholarly disciplines that offered clean and definitive answers in response to applying rote formulae. It all just felt like mobile phone games or crossword puzzles. Only years later did I understand why that was, as I studied for the compulsory full teacher training certificate. I found a piece in the educational literature that maths was taught all wrong: you are given a formula to apply, but never taught its derivation. Maths in school took all the discovery out of mathematical problems. So I have empathy for maths scholars watching AI deliver proofs for seemingly intractable problems. If there were nothing left to be discovered in an archive or through a new interpretation or theoretical development, I would also feel bereft. But what we discover is different. Because there is no external proof, no right or wrong. Which is why AI is so good at it. Things can be confirmed and built on. Cybersecurity either fails or works. Regressions are correct or not. Data either checks out or it doesn’t. None of this really happens in the humanities. And the social sciences can be (relatively) neatly split between the positivist, quantitative, empirical fields and those that are not. And that is the main reason why conceptual arguments are not easier to mimic than experimental results. We are witnessing a very specific type of intelligence is being disrupted, and it is the one that has been widely held as superior. But it was only ever one type of intelligence, and some rebalancing is not going to threaten every type of scholarly work. Just some of them. So, what’s your view on this whole debate?
You're currently a free subscriber to History in Organizations. For the full experience, upgrade your subscription.
|
No comments:
Post a Comment