3 ms·

Here is my theory: these researchers were not naive and they knew the deal, they could work on their research as long as the topics were not too sensitive and a

Here is my theory: these researchers were not naive and they knew the deal, they could work on their research as long as the topics were not too sensitive and as long as the results did now harm Google's image too badly. So what likely happened (imho) was either a miscommunication, or they got results they did not expect. Either they did not realize which topics are too sensitive to touch, or they believed they'd get different results and did not. I cannot entertain the idea that they signed up with Google believing truly they could have full scholastic freedom. I am pretty sure it is even covered in elementary ethics classes that such a scenario is itself questionable.

jariel 6y ago
They probably felt that they had 'scholastic freedom' and they generally do. But if your job is 'AI Research' and you publish a piece about how much electricity Google uses to 'train models' ... well, you can expect scrutiny. In this case, the researcher was simply asked to hold the paper until more information was included about the fact that Google has actually improved certain paramaters quite a lot since the time the data was referenced in the report. This is extremely fair from Google. If the research was purely concerning some AI results, even how they could be unjust - fine. But as soon as your research specifically includes the sponsor, then it's perfectly fair for the sponsor to add more information. It's not censorship, in that case, it's frankly the opposite. The 'miscommunication' here is a kind of naive thinking by researchers who probably have never had normal jobs, who have never worn a 'company hat' and may not have any internalized understanding for communications. For many researchers - money comes from magical taxpayer funds, and the 'Uni Model' is the only model they understand and some might think that's how the world ought to be. If Google was giving people cancer due to the tar in their products, the researchers were discovering this - and Google was trying to suppress this, then you'd have a serious problem. This is not Marlborough/Cargill/Exxon, it's Google and it's an entirely different situation. The 'hammers' used by the social justice fronts from the 1960s just don't work anymore, because we have made progress and nothing is obvious.

jariel 6y ago

They probably felt that they had 'scholastic freedom' and they generally do. But if your job is 'AI Research' and you publish a piece about how much electricity Google uses to 'train models' ... well, you can expect scrutiny. In this case, the researcher was simply asked to hold the paper until more information was included about the fact that Google has actually improved certain paramaters quite a lot since the time the data was referenced in the report. This is extremely fair from Google. If the research was purely concerning some AI results, even how they could be unjust - fine. But as soon as your research specifically includes the sponsor, then it's perfectly fair for the sponsor to add more information. It's not censorship, in that case, it's frankly the opposite. The 'miscommunication' here is a kind of naive thinking by researchers who probably have never had normal jobs, who have never worn a 'company hat' and may not have any internalized understanding for communications. For many researchers - money comes from magical taxpayer funds, and the 'Uni Model' is the only model they understand and some might think that's how the world ought to be. If Google was giving people cancer due to the tar in their products, the researchers were discovering this - and Google was trying to suppress this, then you'd have a serious problem. This is not Marlborough/Cargill/Exxon, it's Google and it's an entirely different situation. The 'hammers' used by the social justice fronts from the 1960s just don't work anymore, because we have made progress and nothing is obvious.