Is AI/ML being hijacked by SJWs?
I noticed in the past year diversity and other woke concepts have been injected into AI/ML work across the industry, despite not being an ML researcher it certainly interests me for it's ramifications in systems and software design. It seems these SJWs want the high salaries sitting on "committees" around AI and ethics and want to skew them to their own agenda. Ethics was always an issue in AI but the SJWs seem to have claimed this spot at many FAANG companies as the ultimate authority. They seem to claim "bias" in AI that seems to only be remedied by injecting their own biases into AI models. These sort of people that want politicians censored and search results skewed getting in control of machine learning would be a disaster for the industry, I fear anyone raising issue with this will get dismissed by HR.
What is sjw
No, these are people raising important concerns that before we rush headlong into integrating powerful AI/ML into everything that we have systems in place to ensure models don’t cause/reinforce societal harm.
Social media already causes societal harm, but we can't get enough of it.
Sir, social media is coming to grips in real-time with the serious harm it has caused society. The potential for damage with ai/ml is easily orders magnitude greater. We should be overly cautious
you are right 100%
You air quote “bias” but several arXiv articles found bias in ML models. Mostly stemming from flawed training datasets. Obviously bad training data will cause cascade of problems in the model. People who typically write posts like yours are in a position where bias isn’t really an issue for them. So automatically if anyone brings up the topic, they become an “SJW” or “woke”. This is exactly why diversity is important, because a minority would take this a lot more seriously.
Why do you assume I'm not a minority?
Not surprised. You were the one that gaslighted a black woman when she complained that her coworker called her the n word. Racist moron
Go woke go broke.
Tbh I find it hard to believe a field where somebody is smart enough to be training/developing new AI models is flooded with SJWs but I’m sure there’s some that get through. There should be ethicists but I hope the field doesn’t become diluted with people incapable of actually training models who just look at data and find bias because that is much less complicated and a place where SJWs could find their niche
Idk if SJWs run the space, but I do know that addressing these issues is good engineering and business. I don't think ethicists are the right people for the job, since model bias isn't inherently social bias, but they're better than no one. Models are only as good as the data, and the data is always bad. What qualifies as bad is up to the person making the model and the related stakeholders. Some changes are essential to the product's integrity, others are not. Now we get to the real question, whether SJW-type bias correction is essential. After all, this isn't quite like an ice cream parlor serving less flavorful ice cream to black people: no one is making a point to sabotage the ice cream. However, if you had a way to check if your shop was selling ice cream that somehow lost flavor when people of a different gender or race ate it, wouldn't you want to know and fix that? That's one hell of a customer impacting bug! More so, it's one hell of a legal liability if you're talking about protected classes. Fixing the bug is good engineering and good business. So should SJWs lead us to fix the bug? I think they've got a bit of tunnel vision, but are better than nothing. Not enough model debugging takes place in general, regardless of whether model bias is on socially relevant issues. Cross-validation metrics don't tell the whole story, which is why we run pilots. Checking for biases can be the difference between solving a problem and failed pilot. It's great that someone's pushing for this added pre-pilot step, even if they have an agenda. Of course, if you can do better and get buy-in, step up!
SJW ruining tech? Cry me a River there has always ethicists in computer science even from my college sophomore classes just making sure we make accountable decision and address important concerns. Also name calling and labelling doesn't help your argument. You can present your argument without stooping to that level.
If companies allow this - this is good for capital