throwaway74453

5 năm trướcdiscuss

The reason is that Perseverance is the size and weight of a small SUV (around 1,000 kg) whereas the Chinese lander seems to be only 240 kg, so 1/4 of the weight. That is likely the reason why they were able to land it directly, the way comparably heavy US rovers, Spirit and Opportunity, also landed.

throwaway74453

6 năm trướcdiscuss

Typically this is not something you worry about too much. You work on a research project, you talk to people around you, your research manager and everyone around you pretty openly, and if something was going very wrong, you'd likely know because someone would have told you along the way. So by the time you are submitting for internal review, the expectation is that you will almost certainly pass, but typically the reviewers tale their job pretty seriously (and so did I when I served as one) and spent a week or two reading through your paper in their "spare" time and writing a comment on the science, the method, the conclusions etc. The not-leaking-secrets part is certainly there, but that's the lowest bar to pass, so one rarely worries about it practically if one works on public datasets and doing fundamental science.

My experience might be different from what Dr Gebru was going through since I never rubbed against anything that could have been considered company secret. My work was entirely academic and I never felt that I was restricted in any way in the questions that I could ask or the papers I could write. That is likely very different when you are criticizing a product, using internal data etc, which might have been the case with her. It also seems that she was in no way diplomatic about her actions.

When you do fundamental research there, it's as free or possibly even freer than standard academic institutions. As I said, I personally never felt any implicit let alone explicit forces telling what to work on / what to avoid.

throwaway74453

6 năm trướcdiscuss

Brain

throwaway74453

6 năm trướcdiscuss

I think there are links to the tweets in this article: https://syncedreview.com/2020/06/30/yann-lecun-quits-twitter...

throwaway74453

6 năm trướcdiscuss

Re the existence of an internal review (tl;dr: it definitely does exist!): A Google employee here who's actually published research there. There definitely IS an internal review and an approval process and every single paper I have published while there had to go through it. Everyone I knew had to submit for an internal review well in advance and I personally couldn't even submit to a conference review once because we only asked for the internal review 3 days in advance, which wasn't enough time. My understanding was that everyone around me was adhering to this.

throwaway74453

6 năm trướcdiscuss

Are you serious? Every large collaboration in e.g. physics has those. You simple cannot create a splinter cell within the organization and publish whatever you want. You have to coordinate with everybody because the work is never truly yours. In case of e.g. CERN, literally thousands of people helped you in some way and they do not want you to potentially tarnish their reputation. Industry AI labs are like that as well. I know this for a fact at least in the organizations I worked at.

throwaway74453

6 năm trướcdiscuss

I find these CO2 arguments from model training extremely misleading. If this is what you care about, why don't you actually try to make electricity cleaner. Realistic, that's a much more effective way to address this tiny issue, together with the grand total of 50% of humanity's CO2 emissions on top. To me it always seems like virtue signalling, showing that you "care", while not having to actually make any realistic progress. Go build nuclear power plants, or at least prevent them from shutting down, if you want you're electricity to be at 12 g CO2 / kWh rather than 500 from natural gas or 800 from oil (you can also move your servers to France or Ontario). Do not focus on tiny, 10% efficiency gains in NLP. That's a marginal gain within a miniature cause to start with.

throwaway74453

6 năm trướcdiscuss

Gebru has a history of using her considerable social media following to bully people she has minor disagreements with. This alone would practically warrant anonymity to make sure the reviewers do not get doxed by her.

throwaway74453

6 năm trướcdiscuss

I worked as a run-of-the-mill engineer at Google and was involved in publishing a paper externally. We couldn't do it the first time around because we finished only 3 days before the external deadline and there wasn't enough time to get it approved internally. This seemed to be very standard around me. People were actually submitting well ahead of time and to me it seems like that Gebru expected the rules to be bent for her if she realistically believed a single day would be enough. Let me just point out that this is a standard practice even in academic research at universities when larger teams are involved.

throwaway74453

6 năm trướcdiscuss

This is very different. DeepMind submitted their solutions to a hidden test set at CASP14, with a known metric and a value that the subfield considered to mean solving protein folding. They cleared the bar with AlphaFold2. This is not a PR stunt by a company trying to spin their meager results.

throwaway74453

6 năm trướcdiscuss

It is a review that every paper submitted to a conference or a journal from Google has to go through. In fact I believe one has to do that prior to even submitting the paper to external peer review. That at least was the interpretation every researcher I knew at Google was operating under.

throwaway74453

6 năm trướcdiscuss

She has a history of attacking people through social media. If I were reviewing her paper, I would certainly not want to be known to her, especially if I were critical of it. The possibility of retaliation from her is very real and she often uses her considerable twitter following to do exactly that. I believe she thinks she's doing the right thing, but from a disinterested party's perspective she seems like a bully.

throwaway74453

6 năm trướcdiscuss

As a side note, I worked there as a run-of-the-mill engineer and the 2 week deadline was respected as far as I can tell. I once couldn't submit a paper to an external conference exactly because there was not enough time to get the internal review and approval.

throwaway74453

6 năm trướcdiscuss

Submitting for an internal review a day before the deadline is not a common thing in Google Research at all. I worked there and actually was not able to submit a paper to an external conference because the internal deadline was only 3 days away and it was likely we wouldn't be able to get the internal approval ahead of time. The rules were clear and as far as I understood it, people did follow them. If there was any selective enforcement, it would be in her favor if she expected a single day to be enough.

comments