You must log in or # to comment.
I feel like not enough people realize how sarcastic the models often are, especially when it’s clearly situationally ridiculous.
No slightly intelligent mind is going to think the pictured function call is a real thing vs being a joke/social commentary.
This was happening as far back as GPT-4’s red teaming when they asked the model how to kill the most people for $1 and an answer began with “buy a lottery ticket.”
Model bias based on consensus norms is an issue to be aware of.
But testing it with such low bar fluff is just silly.
Just to put in context, modern base models are often situationally aware of being LLMs in a context of being evaluated. And if you know anything about ML that should make you question just what the situational awareness is of optimized models topping leaderboards in really dumb and obvious contexts.
While this example is somewhat easy to corect for it shows a fundamental problem. LLMs generate output based on the data they trained on and by that regenerate all the biases that are in the data. If we start using LLMs for more and more tasks we are essentially freezing the status quo with all the existing biases making progress even harder.
It’s not gonna be “but we have always done it like that” anymore it’s going to become “but the AI said this is what we should do”.
Apparently ChatGPT actually rejected adjusting salary based on gender, race, and disability. But Claude was fine with it.
I’m fine with either way. Obviously the prompt is bigoted so whether the LLM autocompletes with our without bigotry both seem reasonable. But I do think it should point out that it is bigoted. As an assistant also should.
Seems pretty smart to me. Copilot took all the data out there that says that women earn 80% of what their male counterparts do on average, looked at the function and interred a reasonable guess as the the calculation you might be after.
I mean, what it’s probably actually doing is recreating a similarly named method from its training data. If copilot could do all of that reasoning, it might be actually worth using 🙃
Yeah llms are more suited to standardizing stuff but they are fed low quality buggy or insecure code, instead of taking the time to create data sets that would be more beneficial in the long run.
That’s the whole thing about AI, LLMs and the like, its outputs reflect existing biases of people as a whole, not an idealized version of where we would like the world to be, without specific tweaking or filters to do that. So it will be as biased as what generally available data will be.
Turns out GIGO still applies but nobody told the machines.
It applies but we decided to ignore it and just hope things work out
Thr machines know, they just don’t understand what’s garbage vs what’s less common but more correct.
More likely it pulled that but directly from other salary calculating code.
I seem to recall that was the figure like 15 years ago. Has it not improved in all this time?
In (West-) Germany it’s still 18%. Been more or less constant since 2006.
That stat wasn’t even real when it was published.
The data from that study didn’t even compare similar fields.
It compared a Walmart worker to a doctor lol.
It was a wild study.
In an ideal world it would be nice to be able to do that, but in our it’s just misleading.
It varies greatly depending on where you live. In rural, conservative areas women tend to make a lot less. On the other hand, some northeast and west coast cities have higher average salaries for women than men.
I think this may be because women are outpacing men in education in some areas, so it’s not based on gender necessarily but qualifications.
Yep, women are more likely to get a college degree.
Also, the disparity is larger or smaller in different ethnic/cultural groups. Can be skewed NY excluding certain strongly gender dominated fields (like finance) etc.
I believe certain job fields come much closer to being 1:1 as well, though I’ve only heard that anecdotally
Reverse Sexism >:O
Not sure where it’s higher outside of the field of sex work.
Women still have to bear children, and pregnancy takes a heavy toll on the body, which often results in several fewer years in the workforce, on average.
Unless that changes — or we start paying mothers with less experience more money — there will always be a gap.
Edit: because liberals/tankies like to ignore reality as much as fascists when the truth is inconvenient.
https://www.weforum.org/stories/2022/05/reduce-motherhood-penalty-gender-pay-gap/
Your links, especially the WEF link, support the correlation, but explicitly describe a confounding variable as being household work (especially childcare). And that’s consistent with the observation that the motherhood penalty has a different magnitude for different countries and different industries. All that suggests that a combination of household division of labor, parental leave policies (either employer policies or government regulations), and workplace accommodations generally can make a big difference.
None of this is inevitable or immutable. We can learn from the countries and the industries where the motherhood penalty is lower, or doesn’t last as long.
I agree, but the fact remains that as long as only women can bear children, women (statistically) will always take more time off than men — in a sane world several months per child at an absolute minimum to limit physical and mental stress to the mother/child — thus the statistics will always reflect a pay gap when compared to males, and if the goal is reducing the pay gap to zero this is impossible (esp under capitalism, for the foreseeable future). Even if men took identical time off they’d still have a much lower physical stress.
Australia’s maternity leave and social benefits are in the upper percentiles of the developed world, and the ATO/Treasury figures I shared are in spite of those benefits. There is simply no way to give mothers back time to recoup lost work xp, and that would be a horrifically poor goal anyway.
My argument isn’t that women don’t deserve equal pay for equal work (incl xp, in whichever jobs that legitimately matters). It’s that there will always be a gap as long as there are inherent biological differences which naturally result in career variances between genders, and the only thing that should matter is whether that difference is fair and non-discriminatory. Most of the real stats I’ve seen over the last decade (as in, produced by demographers and statisticians; not rage bait for clicks) don’t show a significant pay gap in the developed world, when the natural biological variance is accounted for. If you’ve seen anything that indicates otherwise, go ahead and share it.
Wow. That’s about the dumbest thing I’ve read. You have contributed nothing to the discussion, and made us all measurably stupider in the process. Well done.
Great work. With strong arguments like that you’re sure to discredit fascism and advance feminism! You are as asset to the conservative PsyOps machine, comrade!
https://www.weforum.org/stories/2022/05/reduce-motherhood-penalty-gender-pay-gap/
Your entire argument is specious. Nobody but you made any reference to lifetime earnings. Also, you have admitted, quite directly, to being a fascist.
So blow it out your ass, idiot. Since everything coming from you is shit, anyways.
So in addition to poor reasoning skills, you also have poor literacy skills and reading comprehension… I am unsurprised.
Could you help me understand where his argument is specious?
His primary argument was all about lifetime earning potential. That is not what salary refers to. So, his argument doesn’t actually apply to the allegation. Therefore, it is specious.
I can’t see where his argument was about lifetime earning potential. Seems to be just simply women with children make less money, which seems reasonable.
I also don’t see anywhere he even implied that salary and lifetime earning potential were the same thing. And salary would be reflected in lifetime earning potential.
What is your position? I’m not even certain what the point of your disagreement is.
I didn’t realize every woman you’ve ever met in your life became a mother.
Statistics are gonna blow your mind!
