Sure, but as an individual researcher, the integrity of your scientific work is not compromised by the unethical goals of your employer or the unethical ways in which it raises funding. That's what I'm claiming. It's rather different from amateur vs. professional athletics, where both have the same goal of "run as fast as possible but within some nebulous concept of natural human ability": the goals are different between academia and industry.
Or put another way: Of the many deep moral questions raised by the Manhattan Project, not a single one was "Did they commit academic fraud and claim that an atomic bomb was scientifically possible when it wasn't?" They were employed to actually get the job done, not to act as if a job were getting done.
And at the end of the day, funding for university AI labs is largely driven by the existence of those same unethical industry goals - governments fund the work because it's good for the economy, students pay to take AI classes in their undergraduate degrees because it's an investment in their future careers, the industry donates everything from fellowships to entire buildings to academia, etc. I don't think you can cleanly wash your hands of industry's ethical concerns by staying in academia but working on the problems that industry finds interesting. If your goal is employment without any ethical concerns, you're going to have a very hard time.
(In the case of this particular post, the co-authorship of papers with industry makes it clear that the research directly benefits industry.)
And how is an individual researcher funded, if I may ask? We already have many homeless teachers living in their cars, so I don't think the power of public opinion will rebalance financial incentives to make an even more useless profession (in the eyes of the general public) attractive. Joe-Schmoe-the-janitor won't see financially supporting lone nerds as a first priority.
Huh, I thought I was pretty clear that I meant an individual researcher employed by industry to do their research ("for CS in general - and especially for AI - you'll be just fine in industry.... Go where the incentives are better aligned for you.") - was it not?
So, limiting research to projects small enough for a single individual, then? Because similar situations will and do arise in industrial research groups.
Ah, I see the confusion. By "individual researcher" I do not mean "independent researcher" - I mean the researcher, considering their motivations as an individual person. They can be part of a group of researchers, and they're generally working for some employer. After all, the "individual researcher" in academia is generally working with a group - e.g., the example given about the grad student who is "every professor's dream" only makes sense in the context of that relationship - and that produces the negative pressures described.
In particular, I mean that a researcher in academia, as a person (an "individual researcher"), is motivated by the demands of academia to get grants and fill their CV and is therefore incentivized to conduct dishonest science to make that happen, and a researcher in industry, as a person (an "individual researcher"), is not directly incentivized to conduct dishonest science - perhaps there's dishonesty in how their employer gained funding or what they do with the research, but that doesn't compromise the accuracy of their research, motivate them to game the peer review system, etc. The researcher as an individual has the choice about whether to be in academia or industry.
So, I don't think I follow how similar situations will arise in industrial research groups. (Though, as mentioned in my original comment, I'm probably missing something, because there were researchers from industry who coauthored these papers.) Even among a group of researchers in industry, the incentives should be to produce things of value to the employer, not to play the part of productive-looking researchers.
I'm specifically not claiming that independence will solve anything; the fundamental problem is funding, and (as you say) nobody is going to want to live out of their car to do good research. And you need access to facilities/tools of some sort to do your research; my claim is that industry can provide those at least as well as academia can, not that they are unneeded. Admit that you're constrained to work at a place that can fund your research and that no place exists that will pay you well and leave you to your own devices, and then find the place whose incentives to fund you are least likely to compromise your research integrity and most likely to reward you for actual good work. At the moment, at least in the society where I live, that happens to be industry.
Or put another way: Of the many deep moral questions raised by the Manhattan Project, not a single one was "Did they commit academic fraud and claim that an atomic bomb was scientifically possible when it wasn't?" They were employed to actually get the job done, not to act as if a job were getting done.
And at the end of the day, funding for university AI labs is largely driven by the existence of those same unethical industry goals - governments fund the work because it's good for the economy, students pay to take AI classes in their undergraduate degrees because it's an investment in their future careers, the industry donates everything from fellowships to entire buildings to academia, etc. I don't think you can cleanly wash your hands of industry's ethical concerns by staying in academia but working on the problems that industry finds interesting. If your goal is employment without any ethical concerns, you're going to have a very hard time.
(In the case of this particular post, the co-authorship of papers with industry makes it clear that the research directly benefits industry.)