unscrupulous_sw

Imagine the shitshow this will cause once law enforcement adopts this.

Currently eyewitness criminal sketches are still drawn by artist so they are naturally low fidelity.

That will change once you can generate a photo of a face (like https://thispersondoesnotexist.com/) based on your brain waves.

This will be disastrous on so many levels. The eyewitness might not have a good sample of a minority race. The GAN dataset itself might also only be trained on celebrity faces so it doesn't know how to generate anything else (e.g., a teen).

But it will be deceptively high resolution so police will rely on it.

If you have a generic face your life is fucked.

Are there any open sourced attempts at (ab)using all of these at once with some multicloud wrapper interface?

> Why can't you criticize your own model, though?

I do but lunchtime only lasted so long before it imploded

> Your assertion that hiring more women will lower standards is premised on a belief that men are inherently better. Isn't that embarrassingly obvious to deduce from what you say?

If you're making "embarrassingly obvious deductions" you might be pattern matching me into your favorite straw man and not really deducing anything at all.

I never said women are inferior to men and it's not possible to logically deduce that from what I said. The only premise I assumed is that there are currently more men than women at the top (e.g., pipeline problem). You can even have the average women be way superior to men and still have the problem of hiring more women leading to lower standards, if the few top women are all happily employed elsewhere. I am not even saying this reflects reality, just that your logic sucks.

Why does everything have to be a dog whistle to fit people into hate groups? Some assholes like me are merely technical correctness assholes.

Yes! This is the kind of response I would like to have from my coworkers. Attack my models, my assumptions, my arguments. Don't just silently disagree and then label me a sexist afterwards.

Your presumption that I have some agenda and it wasn't social ignorance is a little offensive but whatever. "Ignorance of facts" should not be used as an insult but an opportunity to educate. The social fact that I am guilty of not knowing is that this a commonly parroted argument instead of the first logical model you come up with.

I do think you represented the argument I had at the time very well. The difference is that I used competitive programming as my example for something more controlled (plenty of stats going all the way back to high school level and on the internet no one knows if you're a man/woman/dog). Your counterargument that it isn't predictive is still valid.

Regardless, my current stance is that the question of "whether it lowers the bar or not" is not a relevant question to ask.

The only point I was trying to make with my story is that you can evoke very passionate attacks just for talking about the topic.

Just to chime in, I also unknowingly got myself labeled sexist when the topic came up during a random lunch with "friends" (company provided lunch so it had random people from HR and recruiting).

Somehow or another the topic of whether going for 50/50 will "lower the bar". So I started thinking about the problem mathematically like any engineer would, with statistical distributions and estimates of the existing pipeline and all. The more I said I am leaning towards that it will lower the bar, the more uncomfortable people around me looked.

I was too socially ignorant to realize we aren't talking about math problem here and that they don't want anything said that might weaken their cause, good faith argument or not.

After that I got no more women in my interview pipeline and a "talk" from my manager.

Data laundering

This means crawling or using illegally obtained datasets then processing it with "machine learning" until you have enough plausible deniability to use it.

This could be used for bypassing copyrights. For example you can remix stock photos you don't want to pay for. You can crawl a competitor's dating network to build similar looking fake profiles. You can steal writings and automatically paraphrase it. You can steal algorithms by cloning their inputs/outputs. You can generate new porn by swapping faces and background.

An illegal dataset can also be used as a hidden input to improve your core product. For example you can buy up all stolen databases and logs and correlate the users. This can then be used for better ad targeting using data that isn't even available to google and facebook.

Looking at https://en.wikipedia.org/wiki/List_of_Chinese_spy_cases_in_t... there's at least two cases that are by professors (Xiaoxing Xi and Yi-Chi Shih).

But honestly the vast majority of the knowledge transfer is happening through more legitimate means. Even the famous Stanford machine learning professor Andrew Ng got poached by Baidu.

This is an anonymous account. The unscrupulous part of my user name is in reference to another thread where I used stealing and sharing interview questions as an example so I am not really trying to hide anything: https://news.ycombinator.com/item?id=20812755

But when I wrote that comment I actually did think that backing up emails would be universally agreed to be morally fine. It's just so tame in comparison to acts of the pirates who get idolized for building silicon valley.

The motivation for this question is genuine. I hate having a gap in my google calendar or not being able to recall the details of some project I was working on. But I did choose to move on so yea they are stupid fucking projects even though I might need to reference them later on when updating resumes and such.

> It is not uncommon for tech companies, which fiercely guard their intellectual property, to sue former employees or the firms they join after they leave.

Is it really that common?

I feel like most software engineers don't really take intellectual property all that seriously. Most will not think twice about backing up emails when leaving the company which might contain code diffs, design docs, etc.

Friends from different companies usually talk about internal system designs and technical problems pretty freely. One of the circles I hang out with even backup and share the interview question database from every company they've been at.

So while it's probably crossing a line, it's still not something I expect they will ever get sued for, let alone criminally charged. (This is in contrast with friends from the finance industry who are a lot more tightlipped about everything and you do hear stories about regular employees getting sued)