manfredo

98 ms·

811.
by manfredo 7y ago
Go around and ask your co-workers, "would you you consider a push by management to have fewer meeting a political act?". I sampled my co-workers and none of them considered it political. This works both ways: what data do you have
812.
by manfredo 7y ago
> if you are not having sex, _are you being sexual? No. > Or, if you are not discussing politics, _are you being political?_ No.
813.
by manfredo 7y ago
At this point we are diverging almost entirely from questions about whether AI is biased, and more about whether AI is cost effective. The allegations of bias in AI are frequently about inter-grouo differences in things like job advertiseme
814.
by manfredo 7y ago
No, I have repeatedly and explicitly stated that protected classes like race should not be used as inputs into any sort of model. The notion that I have said that we should use race as an input to a model is completely false. What I am sayi
815.
by manfredo 7y ago
Yes there are and he cited them in his memo. And the studies have been replicated in studies that draw participants from as many as 51 countries and find similar results across different cultures [1]. If you think his presentation of this d
816.
by manfredo 7y ago
Again, the vast majority of "firearms" but not the vast majority of "arms". People aren't buying F-35s at their local gun store.
817.
by manfredo 7y ago
The article explains that this program was terminated in 2017: > This operation, disclosed in an August 2013 Wall Street Journal story,[2] was officially ended in August 2017,[3] and the FDIC settled multiple lawsuits by promising to Con
818.
by manfredo 7y ago
I don't think you're understanding the point I'm making. I agree, we should make our models that are as accurate as possible. But this view is not universal. There are some that point out that prioritizing accuracy may create
819.
by manfredo 7y ago
The field of AI is not new to me. Quite the opposite, I find that many people in this thread are making fundamental misunderstandings about what AI and machine learning are about. Hence the commenters like the one above which say AI is limi
820.
by manfredo 7y ago
Nearly all countries have rates of women in tech in a band between 10 and 30%. No country has a majority female tech workforce. The notion that there is a "wild variation" in the rates of women's representation in tech is sim
821.
by manfredo 7y ago
Which loops back around to my original point, which is that the notion that we should alter or influence the algorithm because its output does not match our worldview or politics is not the removal of bias it is the deliberate injection of
822.
by manfredo 7y ago
> Could a room full of software developers write a bunch of blog posts about random topics? Sure. But, as much as it may seem otherwise from the outside, that's not what a newsroom is doing. In fact, that's not even what the tr
823.
by manfredo 7y ago
That's a proportional measurement. A small difference in an already small value can result in large proportional differences even when the absolute difference is small. This is sort of like how small towns might see their murder rate s
824.
by manfredo 7y ago
No, you can test and gauge the quality of that data. At this point, you're not really even making a point about whether or not we should introduce biases in our models to make them match our worldview - you're saying the entire en
825.
by manfredo 7y ago
I'm not sure I follow. I state that there is stigma attached to acknowledging differences between groups. Calling people racist for doing so seems to reinforce my claim that it is stigmatized, no?
826.
by manfredo 7y ago
Can you elaborate on how European imperialism is related to whether or not we should take lead exposure into account during parole hearings in the United States?
827.
by manfredo 7y ago
> Yes, law enforcement should use this model to determine where to assign police resources to minimize crime. And if an activist engineer deliberately biased the model to avoid indicating disparities in crime, then we will have sabotage
828.
by manfredo 7y ago
Yes, African Americans are disadvantaged as compared to whites. But how is that relevant to me previous comment? When a judge is evaluating whether or not to grant parole, is he or she looking for a causal factor that will directly cause th
829.
by manfredo 7y ago
Great then go ahead and include additional features. The problem is, recidivism rates in real life are not equal so as the model increases in accuracy it will inevitably produce disparities in its results.
830.
by manfredo 7y ago
Right. And when you force a model to avoid disparities when disparities really do exist, the outcome is inaccurate data and the consequence is that decisions are made based on inaccurate data.
831.
by manfredo 7y ago
This is their latest data: https://diversity.google/annual-report/ All roles: Asian, 43.9% Black, 4.8% Latin, 6.8% Native American, 1.1% White 48.5% Women: 33.2% Men: 66.8% Tech Roles: Asian, 51.8% Black, 2.8% Latin, 5
832.
by manfredo 7y ago
If truth is irrelevant then just return random output and call it a day. The reality is, truth is relevant and sometimes the truth is inconvenient. Tech workers may want to build an AI that measures risk of recidivism that produces unifor
833.
by manfredo 7y ago
> In order to be discriminatory, they would have to be broad, ignorant statements based on personal anecdote and prejudice that he pulled out of his ass right before adding 'on average' at the end of the phrase; since they are
834.
by manfredo 7y ago
> I've come to believe that "bias" is the primary reason these AI's perform so poorly. They are basically "biased towards the known stuff" in layman's terms. At least in my field, we need AI's that
835.
by manfredo 7y ago
The narrative is what I have observed and experienced working in the SF Bay area, and it's what the overwhelming majority of my colleagues report as well. You can also see it playing out in this thread, where people are not-so-subtly s
836.
by manfredo 7y ago
> The problem is that these systems are fundamentally unable to distinguish causation from correlation. Which is admittedly a hard prob even for humans, but at least we have some capacity to tease these out. I'm not sure why causat
837.
by manfredo 7y ago
The problem is, some groups do commit crimes at higher rates than others. If we are engineering the model to produce outputs with no disparities when disparities do exist in the real world, then our model is going to be biased. For instance
838.
by manfredo 7y ago
It does not vary wildly between countries. The overwhelming majority of countries fall into the range of 10-30% of tech roles filled by women and the remaining 70-90% by men. I don't think that a wide variety of countries falling into
839.
by manfredo 7y ago
Sure, I suppose I should have qualified that statement to refer to left leaning Americans (which are the primary demographic in US tech companies). Among these circles it is not socially acceptable to acknowledge disparities between groups,
840.
by manfredo 7y ago
And that "something" we want the AI to do is usually grounded in the real world in some way, be it shopping patterns, or crime. And the real world has disparities. For example, crime rates are not equal between men and women. If w

More ›