you can tell john has zero clue about gender issues though. Sub-zero, dude's about as uninformed in both directions as it is possible to be.

these guys are literally going out of their way to hate on any reporting about trans people as trans people, while insisting trump's lies cannot be called lies. "Impartiality" is the polar opposite in their mouths. For god's sake Prescott was getting his 'history' from History Reclaimed...!

You guys had one job, declared to the entire public you had one job, and then quite spectacularly you failed that one job. So it's less recriminations and more a description of reality.

'If both sides are criticising me, I must be doing something right,' I warble, as I run babytalk opinion on the BBC News website.

You cannot run the basic infrastructures of a functioning society- healthcare to education-on systems that cut out human empathy & negotiation. So we have to reassert over and over again that the problems aren’t glitches or errors—even big ones. The problems are fundamental, systematic, endemic.

arguing with a toaster about communism and modern history, to boot. we couldn't make it up.

fun question really, in practice there are major overlaps, social history tends to head towards lived experience and questions of agency and popular change etc. Econ tends to head towards explanations of economic change, standards of living, quantifying work and so on. Both valuable of course.

if it was me, thanks! I also did a paper to QMUL about this sort of thing, also on medium now. Could also have been @willpooley.bsky.social he's written about this too.

it probably wasn't be but I did write a thing a while ago about exactly this on medium, titled "no past, no future".

this one right here is so, so revealing Richard. Like the model's response invites the reader to believe that a. humans reason this way, b, those forms of reason are the same, and c. it's thus plausible for the ai to do so, when in fact it's just the ai doing what it always does. a money quote

it's why so much of the training is about the difficulties of moving from the specific to the general, and it is why historians are, as a group, often so productively cautious around the general and around generalisation.

The first lesson of historical x usually follows y is that over a considerable corpus often x doesn't follow y at all! complexity in historical data is emphatically not noise, it's one of our big c's for a reason!

"LLMs are now widely used in social science as stand-ins for humans—assuming they can produce realistic, human-like text"

isn't that a massive danger to the fundamentals of the entire discipline?

With that line up, 100% don't!

in terms of "explicitly required" all I meant was, I presume some folks out there (definitely not me to be clear) are designing assessments that require students to use an AI, so for those, obs allowed. I don't support that but am not naïve enough to think that genie goes back in the bottle.

These are the LLM bots designed to do a specific task(s), like write slop essays? I can see why blocking them from accessing a platform would not be possible, as they just spoof user behaviours, just tell students not to, if they do its cheating? like, that is emphatically cheating yes?

I have faith they *did* and now that memory is the captain.

gutted for Joe too, brutal loss of faith by Nick the ever-faithful there.

MMV but Isn't the simplest answer also probably the best, which is to not allow the LLM to be used unless it is explicitly and obviously required? Don't we do that with a bunch of other stuff already? I'm seeing this premise of generalised knowledge replacement everywhere, even in 'resist' pieces.