@david_chisnall From friends who are in the tech industry, it seems that using LLMs to actually figure out what the LLMs want to see improves your success of being seen by a human. And that's all kinds of wrong.

@david_chisnall pro hack for people of a certain age: remove as much as you can from your experience to get the job. I started getting to actual humans when my experience did not show my age. I simply write further experience available on request. I kept last 10 years or so so I can show that I have 10 yrs with any technology that has been out for 6 months :-)

@david_chisnall
Though this looks ridiculous, LMs usually reflect the pattern in training data.

That is, human tend to discriminate people based on genders, races, ages and more.
To fix this, either model releasers should process their data or we human should stop discrimination (unfortunately, I think the latter is harder).

@david_chisnall I was curious one time and decided to pass my resume through one of those things. It made up all sorts of skills for me. Some that didn't even exist. Some that were absolute nonsense.

@david_chisnall Go get a lawyer now, these companies are going to get sued into oblivion and you're going to want to get your cut of the pie early

@david_chisnall I agree with you in principle, but we do this all the time with or without AI. If you take a persons car insurance profile and swap the gender you will get two different rates. Same with prison sentencing. Nobody cares about bias, and the cases that do make it through often involve some privileged status of the victim or political buy in by third parties. Presuming your partner is an affluent white woman and anti-AI sentiment is sufficiently high, yeah she might have a case. Everyone else is out of luck

@david_chisnall what if the LLM just wanted to make the company more human by hiring the more emotional gender?

@david_chisnall Feels like a good example of machines trained on the past producing even more of the past.

@david_chisnall this has been known for so long. There is a paper that was published by Timnit Gebru et al in 2021 https://dl.acm.org/doi/10.1145/3442188.3445922 That got her fired from Google. They’ve known it so long that it’s just malice at this point and we’re just waiting for the bellwether

@david_chisnall sadly they really won't. Bias in recruiting is super rampant already, and lawsuits get no where. Personal experience of lawsuits, and bias. Personally and professionally.

But yeah, big suck. See also black sounding name vs white sounding name.

@david_chisnall She--excuse me, HE can lift that point avg higher by subtly indicating they're white, cis, het and not fond of queers and BIPOC! Heil Shitler!

ugh. Actually, be curious if that was true. Maybe include an intern stint at the Heritage Foundation?