[dekhn]

When I worked at Google I had the chance to interact with many of the leadership including the founders and CEOs, as well as many SVPs. I have to say, both Larry and Sundar have the personality and motivational skills of limp noodles. When I met Eric Schmidt, he immediately quizzed me about every detail about my product and product plans and further goals and was encouraging, like an actual leader. It was a wonderful discussion and I left the room feeling like I had actually been listened to by a leader who wanted the company to grow in new directions (like cloud and machine learning for health research).

I also got to meet people like Urs Hoezle and Luiz Barroso who are responsible for Google's technical position in computing today; they were both also great leaders who really deserve to have their own sub-company to run (TI and Core).

I was truly lucky to be able to have coffee with Jeff Dean and Sanjay Ghemawat almost every morning for a year, which led to some great research collaborations. I was extremely disappointed to see Jeff defend Megan Kacholia's firing of Timnit (yes, I read the paper) and call it a resignation. That pretty much killed the reputation of Google Research's leader.

[rejectedandsad]

….how did you manage to get into this position? Were you a senior principal engineer?

[dekhn]

I have a long history. I started out as a Biophysics PhD doing MD simulations and applying machine learning to biology (search for 'David Konerding' in Google Scholar). Since Google is very much like academia, my research training prepared me to be successful inside. Getting inside was the hardest part!

When I joined google, the only way I could get hired was as a test engineer on an SRE team, which I then converted into an SRE via mission control and wrote docs and shared them internally with folks until the principal engineers read them, and then they gave me infinite resources (Google Exacycle) to do everything I described above. I used that to get promoted to Staff SWE, and used that to launch a product (Google Cloud Genomics) and do some interesting machine learning for drug discovery (BTW, at this point mny career was effectively complete- I had set out to do everything I wanted, and was interested in what to do next).

The above happened because (beyond a wide range of boosts provided by parents and country) I have an intense drive, wanted to work at Google more than anything, and exploited the internal structure of the company to maximize my power. I kept networking to meet more and more people, and by meeting those people I got more access and support. I helped build up a team- Google Accelerated Sciences- which does basically what I thought Google should be doing all along.

unfortunately, at that point Google politics and personalities intervened and I was kicked out of the cool kid's club.

[Aperocky]

> I have an intense drive, wanted to work at Google more than anything, and exploited the internal structure of the company to maximize my power

> at that point Google politics and personalities intervened and I was kicked out of the cool kid's club.

It sounds like a power struggle gone the other way from reading this. And it certainly doesn't seem like either of the sides are more noble than the other.

But props for you for doing what you loved best, if only one day any of these can be decoupled from politics.

[alisonkisk]

There's a difference from politicking into a club and politicking others out of a club.

[rejectedandsad]

Wow, very impressive! Turns out I've followed your work for a while and never knew your HN username. Makes me wish I had the innate ability to be able to achieve even a fraction of this or have a scrap of the prestige - instead, I'm stuck on a lower rung for good (I blame heritability of intelligence via my parents).

[dekhn]

Huh, OK, I didn't know anybody followed my work! Note that I am not particularly intelligent- I always struggled in school, and had a ton of imposter syndrome. I tried and failed to become a successful scientist and instead pivoted to what I'm actually good at (scientific computing). Most of my drive came from fear of not being able to make enough money to live in the bay area, or from being thought unintelligent.

Never underestimate the power of imposter syndrome!

[selimthegrim]

Where would you go now if you wanted to do such things? I have people in UW CS tell me a physics PhD in CMT is no proof I can learn how to analyze RNASeq data quickly enough for them and they want a postdoc with the experience already to teach DL to (I would have thought the other way around would be easier)

[timy2shoes]

The other way around is definitely easier. RNA-seq analysis is a mostly solved problem with DESeq2 (or edgeR/limma). The tutorials are very detailed. The most difficult part is experimental design, which you probably know already.

Deep learning, on the other hand, is so fraught with pitfalls and traps. Even if you can code up a model successfully, it's very easy to trick yourself that you're doing very well (see a previous discussion at https://news.ycombinator.com/item?id=27376839). In my opinion, most of the work should be spent on making sure that you're not tricking yourself.

[dekhn]

literally the first 5 years of training in any large scale data analysis should be "how not to trick yourself into thinking you found something significant that generalizes"

[selimthegrim]

So then why I am getting static from these PIs? Do they have some R01 report deadline a year away breathing down their necks?

[dekhn]

There's a reason I invented this saying: "The very best machine learning models in biology are distilled from postdoc tears".

[gone35]

> I helped build up a team- Google Accelerated Sciences- which does basically what I thought Google should be doing all along.

MD?

[shadowfox]

Probably "Molecular Dynamics".

[dekhn]

Yes. also known as "the computational job which will consume all CPU forever, while not answering useful scientific questions".

[1024core]

> I was extremely disappointed to see Jeff defend Megan Kacholia's firing of Timnit

Jeff had 2 options: pick a safe leader, or a corrosive person. Timnit has shown incredible tone-deafness and arrogance; just view her Twitter exchange with Yann LeCun. She was clearly out of her league, and doubling down every day with newer and newer antics. She needed to go.

[dekhn]

I didn't say that Timnit shouldn't be fired.

[thu2111]

Are you objecting to Google's refusal to let her "unresign" then?

Btw plenty of us ex Googlers think Jeff Dean's reputation was enhanced still further by that move. Gebru was acting in crazy ways, she really should have been fired much earlier, but taking the opportunity when she so helpfully presented it seems like a no brainer. Sad that the rest of Google lacks that approach.

[dekhn]

I don't think Google should unresign her. I think it should admit it fired her (technically speaking), and pay her a bunch of money (to make up for the bad handling of the firing).

I agree, she should have been fired earlier. However, she was just promoted, which meant she must have had several quarters of excellent perf.

I think one thing that would be truly excellent, but won't happen, would be to have Megan Kacholia defend herself by explaining her actions to the wider Google community. In particular:

Why was Megan, a person with effectively no research experience, a VP in Google Research and making decisions about research papers being withdrawn? Her lack of experience in the area seems to have led to an exacerbation of tensions.

Why did Megan convert an offer to have a discussion about a resignation into an interpretation that Timnit resigned voluntarily, coupled with an accelerated departure (immediate, with termination of all Google services)?

Why was Megan pushing so hard to have the paper withdrawn, given that the paper wouldn't be that impactful on Google's reputation, made some useful (if obvious and a bit overstated) points, and wasn't being published in a prominent venue?

To Jeff and Megan: to what extent did Jeff support Megan's decision to fire Timnit over a refusal to withdraw a paper, or did his support come after learning Megan had fired Timnit? To what extent did Timnit's previous tweets calling out her employees and mentioning confidential Google Research activities play a role in her firing? Was her discouraging email really a reasonable justification for accelerating her termination?

My guess about all of this is that Megan and Jeff decided to fire timnit when timnit posted several negative tweets attacking Jeff and google Research, and used the paper and the email (and vague offer to resign) to justify the firing and they didn't think through the implications of firing a twitterati like Timnit in a roughshod manner. This seems most consistent with all the evidence I've seen.

[1024core]

Given the statement: "A || B" (where A is "take this course of action" and B is "I will quit"), if A is false (i.e., !A is true) then logically B has to be true.

The moment an employee says "Do this or I quit", and the company refuses to "do this", then the company is correct in assuming that the "I quit" part will apply.

[thu2111]

You didn't understand my question at all.

Gebru said "do this or I resign". The company accepted her resignation. She then attempted to "unresign" by claiming her statement she would resign wasn't an actual resignation. Clearly you support this kind of meritless word games from her.

As for Megan, why is a non engineer running YouTube? Why did someone at lightweight as Gebru end up being paid as a researcher at all given that her research had no real validity? Why was she allowed to behave in such toxic ways for so long? Google is neck deep in identity politics, that's why.

[actuator]

> Jeff defend Megan Kacholia's firing of Timnit (yes, I read the paper) and call it a resignation.

From what has been put outside, don't you think Timnit's mail to employees as a manager was out of line for someone in her role.

Even to me who doesn't have intimate knowledge of the whole thing, that didn't look appropriate.

[dekhn]

Yes, Timnit definitely pushed the boundaries of what is considered acceptable discourse at work. But, speaking both as an ex-manager, and an engineer who worked there (and resigned twice), it definitely wasn't a resignation.

To resign at Google you tell your manager you're resigning and then fill out a form (that's the process that makes it an official resignation). What Megan did to Timnit was immediate termination, combined with an advanced exit date, which only happens if you're truly and deeply violating Google rules or your country's laws.

When I brought up the Timnit firing situation my VP literally said: "It takes me a year and a half to fire a bad employee, I don't know how Megan did it so fast".

[Judgmentality]

> It takes me a year and a half to fire a bad employee

What? I could understand 6 months, but how the hell do you add a year to that?

[moandcompany]

COVID-19 and consciousness about complications from WFH, new stresses or demands on employees in their family/home lives also ended up resetting the clock on a lot of PIPs or plans to set those PIPs in place ...

Also, managers/directors/etc often have to overcome the issue of reversing momentum on past praise or ratings given to these individuals.

In the Google perf (performance review) process, many individuals that probably should have gotten PIPs and/or counseled out also got high performance ratings in past reviews (for other reasons) from the very same managers and directors.

Generally to initiate a PIP, there needs to be a rating to justify the action (i.e. "Needs Improvement" - lowest rating). Giving a rating two levels or more above or below the last rating also requires justification in the rating process.

Google also cancelled/deferred its mid-year perf cycle during 2020 due to WFH/COVID challenges, which delayed opportunities for managers to give such ratings or feedback through the formal process. (for context: Google traditionally does "perf" twice per year, which are the formal opportunities for employees to receive performance ratings as well as nominations for promotion); this may or may not be correlated with interesting product launches or changes you might see as an end user during the year).

[Judgmentality]

This just sounds like excuses for incompetence. This could easily be done in 6 months at any functioning org. Yes, I am calling Google dysfunctional.

I really don't care. I've never wanted to work there anyway.

[neaanopri]

"Performance Improvement Plan"

[Judgmentality]

I've seen lots of those before, but never anything that could remotely approach 18 months (I typically see them last only 1 month). I cannot imagine the frustrating bureaucracy that is Google.

[izacus]

Wait, you're seriously claiming that a SV corporation (where, as I understand it, most people have 2 weeks of notice) can't fire a person sooner than in a year and a half?

Even to my socialist EU mind this sounds untrue and pretty crazy.

[devonbleak]

The US is insanely litigious and most large companies are very aware of the cost of unlawful termination litigation. That means to actually fire somebody you need cause, and that means having an airtight paper trail which takes time to generate OR having the employee do something that directly violates something you've explicitly told them in writing they can be terminated for. You're not generally dealing with stupid people, and if they're obeying the letter of the law/policy even while defiling the spirit of it you really can't do much to terminate them.

There are other things you can do to make it clear they need to move on but it's usually way easier and more efficient to just directly negotiate a severance and have them resign.

[dekhn]

I didn't say that. It's a complicated situation, right? If an employee threatens to kill another, they can be terminated immediately (with security escort).

i'm talking about employees who are in good standing, like Timnit was at the time. She had just been previously promoted. A person like this, even if they are annoying many coworkers through emails or papers, can't just be terminated because the manager doesn't like them (huge liability risk). Instead, Google (or IBM, or whomever) wants a paper record showing that somebody is unable to do their job, is put on a performance improvement program, cannot improve performance. At that point, Google can terminate the employee and if there is a lawsuit or mediation, Google has the paperwork required.

[londons_explore]

I think there are complexities with firing some people... If a manager just walks round firing people with little evidence, all other employees will live in fear of being fired. That isn't good for morale, productivity or creativity.

Instead, the problems of the to-be-fired employee needs to be abundantly clear to everyone nearby, so that when that person is fired, it doesn't have deep social impacts.

Combine that with desiring to fire someone when a replacement is trained up (often many months), and at the end of a big project (sometimes a year), and not just as management is reshuffling... And suddenly a firing takes 1.5 years.

[romanhn]

A friend of mine, a non-technical manager at Google, similarly struggled to fire an obvious underperformer for over a year, so I totally believe it.

[Laremere]

My read on the situation is:

1. Google wanted her gone, ranging from good reasons (she was too abrasive) to possibly suspect (in this instance she was being abrasive about a paper with ethical questions on Google's practices being stopped from publication with no explanation given)

2. Her ultimatum email can reasonably be construed as a resignation. It roughly said "Do this or I'm quiting", and Google responded with roughly "We're not doing that, thanks for telling us you quit, we accept".

3. This does not follow the typical resignation process used at Google, but that doesn't mean it isn't a valid resignation. It's unreasonable to assume a lawyer didn't look this over before they went ahead with it. The lack of Timnit suing Google for wrongful termination (from what I've seen) agrees with this.

4. Googlers were angry about this situation, because they disagreed with leadership's actions.

5. The leadership's response is legally bound to stick strictly to saying she resigned. This only inflamed #4 more.

6. They were ethically bound to not disclose all details of a situation involving an employee (where as Timnit could paint whatever story she wanted). Even if Timnit gave a full go-head, there were others involved and doxing is a real threat when names are exposed.

7. They were bound by business interests (at its root legal and ethical obligation to the shareholders) to not expose all of the details of the paper, the objection to the paper, and the processes involved.

So was Google in the right here? The situation obviously wasn't handled well. There were clear problems with Timnit, and she did give an ultimatum. However there are reasonable concerns related to "our ethics person gave an ultimatum and we called them on it" - but without the details it's hard to form a nuanced opinion.

Other takes welcome.

[dekhn]

I mostly agree with your logic chain with the exception of (3) in that it was a valid resignation (it wasn't). I also wouldn't read too much into Timnit not suing Google (she does not seem to be treating this situation strategically).

I don't know if a lawyer didn't look at the firing, but Megan has a tight relationship with HR, and if she told HR to resignate the employee because they were damaging to the company, it would be done immediately as a special-case override. I'm 100% certain Megan has had to terminate employees for cause in the past (think: ads engineer who threatens to steal money from google) so I think she has an expedited path.

[choppaface]

If Timnit sued, what could she recover? Money? To do what with?

Timnit’s goal is likely closer to that of effecting reform. A lawsuit isn’t necessarily the best tool for that unless it’s so big it establishes new precedent. Even the Andy Rubin / breach of fiduciary duty lawsuit was largely ineffective.. While Timnit has a strong case here, it’s not really as strong as the evidence of the General Counsel of Google openly engaging in and protecting the sexual misconduct of himself and others.

[GVIrish]

I think the email can be inappropriate and Dean's response can be problematic at the same time.

[andyxor]

Timnit was a bully, the definition of 'toxic employee', so glad Google had the balls to let her go, PR hit be damned, it means they still have true leaders

[dekhn]

my complaint is merely that they continue to insist she resigned. She didn't- she was terminated without cause.

I personally think Timnit shouldn't have been hired in the first place, but if they were going to fire her, they needed to follow the path, which takes about 1-2 years, of establishing that she was not a good employee for Google.

[mchusma]

Thank you for following up with some nuance. I did read her email, and do disagree with your stance that she did not quit, and presumably Googles legal team did as well. But I appreciate your thoughts.

[ksd482]

Why shouldn't Timnit have been hired according to you?

[dekhn]

According to her, she was recruited by Jeff at a conference specifically to work on improving Google's machine learning equity. As part of hiring somebody, I don't just meet them at a conference and read their papers. I interact with their prior employment network and read their social media. I think if Jeff had paid any attention to her tweets before she was hired, he would have thought twice about bringing her on.

[Apocryphon]

I wonder if anyone's compared Timnit's firing to Apple's firing of Antonio García Martínez. It sounds like both are politically opposite versions of the same mishire situation.

[systemvoltage]

I agree, even if she was in the right - she was and is still a bully. Just go on Twitter and examine her feed which gets updated literally every 5 mins everyday for 10 hours a day. Toxic, fuming and deeply disrespectful to others. She spews so much hatred on Twitter, it is unimaginable why anyone would hire such a personality. She wanted to become martyr and she indeed has.

Here is more discussion: https://news.ycombinator.com/item?id=25292386

[raclage]

I haven't really followed this since the initial blowup. Did more information come out about her conduct there?

[choppaface]

Do you think Jeff Dean defended the firing because he really understood the circumstances and the consequences of taking the position, or do you think he was naive? Was Jeff involved in the firing of people like Michael Church?

[spoonjim]

Understanding the circumstances and consequences of taking a position, and not being naive, is almost definitionally the job of a leader. Technical wizards who are clueless n00bs should be individual contributors, not the head of Google Brain.

[choppaface]

And I actually mean the question seriously and thoughtfully, although the downvote barrage wants to take it as flame bait. Jeff’s calibration towards firings is unquestionably different than the rest of industry, but how different? And how much training / preparation did he really have? Important questions for any org head. The thing here is a lot of current and past Googlers seem to agree Jeff messed up. I’d like to understand why through the lens of evaluating what bar we set for the people-managing skills for engineering leaders industry-wide.

[spoonjim]

In 2020 if you can't figure out that a white man firing a black woman for something related to race and AI is going to be a big problem for you as the head of AI for perhaps the world's most potent AI juggernaut facing increasing calls for regulatory scrutiny you really have no fucking clue what you are doing. Overall the entire situation needed to be handled far more carefully even if the outcomes were the same. He thought that because he was the boss he could do whatever the hell he wanted to, and the fact of the matter is that there are forces more powerful than him.

[nostromo]

> Larry and Sundar have the personality and motivational skills of limp noodles

And yet they were still more motivating than Sundar, who is so flat he sounds like he's bored to death, even when he's talking about super cool tech like AI and quantum computing.

Page wasn't polished, but he could energize teams about building the future in a way Sundar hasn't been able to.

[dekhn]

I think you misread- Sergey has lots of personality and motivational skill (he's a brilliant natural leader with an amazing sense of humor). The main problem with Sergey (told to me by several SVPs) is that he's so smart (but not wise) , 10 minutes into your well-prepared presentation he'll ask "well, why didn't you do it this other way? It's faster and simpler?". This will cause your presentation to halt, you have to waste a bunch of time epxlaning why that is a cool academic idea that won't actually work in the real world, and then everybody on the team is amotivated.