Yup, exactly. This is also the stance taken in the book Lingua Franca:

> Ethics in AI should not be so convoluted as to require a separate profession; ethics should be distributed across an organization, flowing through and between individuals so that conversations of design naturally arrive at ethical questions. More crucially, everyone should have the tools to speak up and suggest improvements, so that no one person or set of people hold the keys to building ethical technology.

https://linguafranca.polytopal.ai/handbook/agency

If we look at biology and medicine, there's both. The NIH requires everyone funded on their training grants to be given instruction in basic bioethics, and at the same time there is a profession of bioethics that is involved in grant review and oversight committees because a lot of cases aren't simple and covered by the basics. Bioethics enters the picture from regulation rather the way independent financial auditing does.

Also even with CS you have this. To use security as the comparable system, everyone involved in an application should be engaged and aware of security as a component and goal. Anyone should feel empowered to raise security concerns. You still have security experts who are more knowledgeable than the average developer who build (or collaborate on) security critical subcomponents and who audit/red-team the system to ensure security.

The same can be true of ethics.

That's a great resource (that links to even more great resources) - thanks for posting it.

I disagree that professional ethicists aren't needed in an organisation like Google though. Just as security should be lived and breathed by every developer, you still need some experts to consult when things get gnarly, light the way when you need many years of study to understand the concepts, and spot things that laypeople would miss.

The large language models Timnit was studying are a great example of this; people like her were (and are) needed to help the engineers understand the breadth and scope of the issues. Everyone in the org having their hearts in the right place doesn't replace expertise.

I think the article is taking the stance that ethics should be a concern at the point of development, and those that create these large language models and those that use them should be accountable. Right now, those groups are not accountable, so instead a separate group of people is asked to take responsibility. This is where AI Ethics as a field has ended up, and in my opinion, stalled.

I don't think these two agree? The go suggests keeping a separation between who's doing the researching and who's checking its ethics. Like, the ethics researchers should not be accountable to people who are incentivized to ignore them.

Instead, the quote suggests that the separation is unnecessary, and that elevating some people to specifically look at ethics is bad since it's everyone's job

> You are basically paying people to shit on your own products. Makes no sense.

It's good to be critical of your own products to make them better. But you have to actually be willing to do that for it all to work.

It’s a great resource to pay people to do that internally.

It’s a complete disaster to hire people who see their job description as shitting on the company publicly, with no constructive proposals for improvement.

One of the criticisms in that paper was that training AI used a lot of electricity. That's technically true, but what's the proposal there? The power efficiency of hardware is already being increased.

> what's the proposal there?

That's what Google was afraid of: the alternative is to decide that the cost of the model isn't worth the effort (externalities < benefit) and to abandon it. Which might be the correct decision, but one that loses Google a lot of potential money.

The paper didn't attempt to examine if the benefits of the model outweighed the costs. They calculated the total energy used and called it a day.

Correct. It left it up to the model creators to potentially evaluate those tradeoffs for future models.

I find the conclusions ludicrous. Should we also abandon particle accelerators because they use too much electrity for ill-defined benefits?

Really the claim is toxic academia style knife fighting for tenure at best.

I bet all the big tech companies spend way more electricity on things other than training big models.

"You use too much electricity" is tantamount to "tech is a flawed business model". Arguably solvable with increased green energy sources but that's more a job for society in general than just tech companies.

How often is the electricity use considered a constraint in algorithm design for AI models?

Moore's law isn't the only way to gain efficiency, and even if the available compute goes up, doesn't mean the total power usage will go down

Do less AI training, so less AI could be a valid alternative.

Part of the issue in these events is that the paper that was written by Google's own ethics researchers simply repeats estimates of AI energy consumption from outside academic papers, instead of examining them. The reality is that Google's ML infrastructure is dramatically more energy efficient than the rest of the industry. A person with access to these facts who chooses to ignore them isn't really doing their best as a researcher, they're just grinding an ax.

There are other examples against the meme of AI using "a lot of electricity", for example a recent paper from Google where they show that a learned computational fluid dynamics technique uses only 2% the energy of standard techniques.

https://arxiv.org/pdf/2102.01010.pdf

Not only are ML techniques finding ways to do physics with less energy, the amount of waste by non-ML computing absolutely swamps ML computing. I don't care if training a model uses the amount of CO2 that a translatlanic flight uses, because those flights are flying constantly, some of them not filled to the max. The whole criticism about energy in the paper seemed to miss major details that I'd expect top researchers to get right.

BTW, I ran a large scale computing platform at Google for several years. It made absolutely enormous, stunningly huge amounts of CPU available to scientists and we used it to make several important breakthroughs. However, after a few years, I finally was shown what the power cost implications were, and I shut down the project because the costs were not justified by the scientific benefits. Compared to what I was doing, ML is relatively skimpy on CPU.

I'm not sure that your scientific facilities cumulatively spent as much CPU time as DeepMind spent playing StarCraft. If we want to talk about energy and ethics in machine learning I guess that's where I would start.

Yes, my calculations used far more time than DM, with the caveat that my jobs ran on CPUs while most of their training is on accelerators.

I believe we reported that Exacycle provided 700K high speed Xeon cores for over a year, and the actual number is far higher (I can't share it because you'd be able to make a reasonable estimate on how many computers Google has).

> I'm not sure that your scientific facilities cumulatively spent as much CPU time as DeepMind spent playing StarCraft.

You gotta pay if you wanna play. Humans are the result of a much longer and more expensive evolutionary process. It's only normal to have to pay a little to do the same for AI.

The proposal is not to continue the trend of simply adding titanic amounts of parameters to models and develop more efficient ones.

The criticism would need to be constructive to be useful, or at least specific enough to act on. Also, you can't go embarrassing the parent company, especially one as image-conscious as google.

Any kind of public criticism is embarrassing from a PR perspective.

Either you accept that, or you don't.

"Ethics researches" that can't publish public criticism , with minimal oversight restricted to preventing IP leaks, are a pointless exercise in fake responsibility posturing.

Agree to disagree. One of the main “enforcement arms” of ethics research is to say the unsaid things out loud. Saying them introduces legal liability onto the company in future lawsuits.

I.e. “Google knew that X would result in Y and they still released X; they should pay for the damages of Y.”

The PR side of ethics research barely matters in comparison.

> Saying them introduces legal liability onto the company in future lawsuits.

My first thought was ExxonMobil when I read this comment.

> In July 1977, a senior scientist of Exxon James Black reported to the company's executives that there was a general scientific agreement at that time that the burning of fossil fuels was the most likely manner in which mankind was influencing global climate change.

> According to the Union of Concerned Scientists, "The funding of academic research activity has provided the corporation legitimacy, while it actively funds ideological and advocacy organizations to conduct a disinformation campaign."

Nothing has happened to Exxon Mobile. None of the executives are in prison as far as I know. You'd think the company would be bankrupt by now...

https://en.wikipedia.org/wiki/ExxonMobil_climate_change_cont...

Humanity has a fair history of introducing products that are later shown to have large externalized costs: tetraethyl lead, PVC, etc.

I interpret your point to be that publicly investigating the harm of future products---by the only people who can do so---is a bad idea. What is the alternative? Regulation? (Gasp!)

Perhaps we should start the disinformation campaigns before the products are developed. "If you are arrested on AI-provided evidence, you are guilty. Period."

Which is why it’s idiotic to hire people whose entire job is to create bad publicity for the company.

Few (none?) companies would be happy to pay someone to be publicly and loudly critical of their products. Especially if it wasn't constructive criticism that they could act on.

Completely agree with independent auditing. If automotive safety had been deregulated, the body count would be orders of magnitude higher. This chart says it all:

https://en.wikipedia.org/wiki/Motor_vehicle_fatality_rate_in...

The regulation of automotive safety in the 60's/70's had a huge impact.

> It makes no sense for them to hire their critics and then think they can stay independent. You are basically paying people to shit on your own products. Makes no sense.

Shit on your own products?

So if I'm a good engineer and fix mistakes, I'm "shitting on my company's product?"

There's this belief in this post that un-examined AI is some shiny golden award, and ethics researchers are tarnishing it. It's the other way around.

Does that not depend on your perspective? There might not be an alignment between what Google’s executives and shareholders want out of the company, and what an ethics researcher might consider ethical behavior.

> Does that not depend on your perspective?

Well, it depends on your breadth of understanding of the issue. If you have a limited scope, it doesn't seem like a big problem. And it also assumes that we all agree on certain basic human rights. If we don't then yes, it is a matter of perspective.

Read "Weapons of Math Destruction". It discusses how existing algorithms discriminate in the following case-studies:

- courtroom sentencing (and recidivism prediction)

- mortgage and loan rate determination

- educator performance

- job applications that use 3rd party screening tools

The book looks at specific examples of where these black-box algorithms are deployed that have ruined people's lives with no accountability. The ethics concerns that are being raised were needed a decade ago, or more, but it is only getting worse with the ad-hoc deployment of this un-baked technology to almost every industry.

I have to stress again that these black-box algorithms are ALREADY IN USE and cannot be subpoenaed by courts because it is considered intellectual property.

So yes, it is urgent to crack open this technology because it is all too easily being sold without any investigation, accountability, or thought to the consequences. "Move fast and break things" doesn't work if it ruins peoples lives by landing them in jail, or pushing them into poverty.

The book demonstrates this is not a what-if strawman, but reality.

Thanks, I definitely understand the human rights implications and the importance. I mostly expect a large corporation to make decisions based on profits over ethics though, unless the former depends on the latter (through pressure from the public/PR, legislation and so on). From that perspective, I could see Google regarding an employee remarking negatively on the ethical aspects of an otherwise popular product as ”shitting on it”.

"Weapons of Math Destruction" was our textbook for CS 6603 at Georgia Tech. It was a fascinating/terrifying look into some of the systems that seem to silently impact the lives of millions in the U.S. Cannot recommend it enough.

> it is urgent to crack open this technology

Just as easy as putting the baby back in. It would be more constructive to work on ways to detect and reduce harm.

Bullshit. It is literally being made right now. Legislation can help solve this pretty quick. No audit? No sell.

To follow your analogy, it is more like, "wear a damn condom and take a sex-ed class before having an unwanted baby."

I don't understand how your conclusion follows from the graph. That data would appear to indicate the opposite: that regulation had no effect.

The graph shows that vehicle safety improved steadily throughout the entire 20th century, without any sharp discontinuities. The key line is deaths per billion VMT. That is the figure which is invariant to how much driving people do, and which thus accurately measures how safe driving as an activity is. You appear to be looking at "deaths per million people" but that figure is confounded by big changes driven by the economy, which is why the 1970s energy crisis is highlighted on the graph. All you're seeing there is economic change, not change driven by regulation.

If regulation would have actually made a difference then it'd show up in the red line, but it doesn't. Conclusion: the regulation is useless. Private companies were doing fine at improving safety and always have been. And as regulations have costs, that in turn suggests they should be scrapped (zero impact + non-zero costs = repeal).

- You are basically paying people to shit on your own products. Makes no sense.

They aren't really researching ethics. This is what i think is going on there. recycling an old comment.

{ AI Ethical research can never be an accurate representation of the majority of humanity - or even the majority of its users! If it can it is not sustainable for any long period of time.

What about far more stable principles, such as murder and racism, you ask?

They are prone to being overplayed or downplayed, are state executions murder, or justice? What if the victim/ hangman happens to be black? Why should it matter? Just ignore some issues? That's misleading by omission.

It would be better to just admit "yes, we at Giant Tech know our ethics are bs, but we had to put something down or our machines won't work. Maybe we are not ready for advanced ai. Maybe there's a limit on what programmers can do. Yea. We know. turns out computers DO have limits. We'll have to find other ways to make money."

"But why should we say that!" cry all the executives when this speech is proposed to them.

"because it's true, and if we don't act now the company is screwed. And our clients will also get screwed" is the answer of the timid executive who first suggested this.

"how will the truth help us?" respond all the execs in unison.

"if we can keep up the lie for long enough, we shall all long be millionaires and retired before it implodes! We shall long be out of danger! who cares if some people lose money?"

"Yes, but don't you feel bad for all the shareholders? And how can we possibly fool people for decades to come that our ai isn't bs?" Responds the poor executive weakly.

"By creating a fake team and telling everyone they are ethics researchers" they say. "Really they are just pawns to help us earn more money. Fish get eaten by bigger fish you know?"

"can you just help me change a few lines on my press statement?" Asks the first executive.}

It may also be that Google didn't realize the extent to which they were inviting some of the more organizationally destructive and divisive elements of modern politically correct culture into their organization: https://jakeseliger.com/2020/12/03/dissent-insiders-and-outs.... It may be that we're going to see a trend towards "Leave politics and religion at the door."

To be honest pretty much any bigger professional organization already operates under "Leave politics and religion at the door." umbrella. It's pretty much the only way how you can have an organization filled with mixed political viewpoints function in the same direction.

For some reason however, Google and their employees kinda fail to understand this basic tenement of work culture. They're not a scrappy 10 people startup anymore which can be both a workplace and a political party fully agreeing on political issues.

> It may be that we're going to see a trend towards "Leave politics and religion at the door."

Don't get my hopes up.

-It may be that we're going to see a trend towards "Leave politics and religion at the door."

what are you left with? misinformation is far more prolific than truth. it needs to be held in check, somehow. ignoring it will make it waorse

The bigger the company, the stronger the motive to control politics and religion (for profit etc.)

I imagine that the conversation beneath that umbrella will be flattened in a day.

>and racism

Excuse me, but the definition of racism in the west is anything but stable. The definition has changed dramatically in the last decade, and it could change again.

that merely reinforces my point. Though i accept it was inaccurate of me.

> You are basically paying people to shit on your own products.

In reality, if you want them to fulfill their true role, they are preventing you from producing wrong products (which should be good for the Company).

But Google understood that as you say (which is not what I think but this is beside the point), and that was their mistake.

That makes sense to pay someone to perform that function privately and internally.

It’s been a disaster for Google to have that happen in public.

I think you will find that hiring skilled researchers and telling them that they cannot say anything publicly is a hard row to hoe.

Apple seems to be doing that just fine and with huge success. They are exceedingly good at silencing every single employee to not speak about any kind of internal dissent or HR issues like this.

Perhaps, but hiring skilled researchers who do not think they need to say whatever they want is probably not as hard.

Importantly, Google offers great work conditions, access to their intellectual property and different evaluation criteria. Many prefer that and other benefits over the stresses of a proper academic career.

This is the most curious part: hiring those two individual researchers, who clearly didn't have an understanding of the nature of public communication, was incredibly naive and asking for disaster. There is no other outcome. It kind of implies a total lack of understanding of human nature on the part of those doing the hiring. "What were they thinking?" is the only thing that comes to mind.

It's a very, very tricky thing to do to hire your own public critics, that has to be done very carefully and with a lot of parameters.

> You are basically paying people to shit on your own products. Makes no sense.

They hired them to control them. And it's a diversion to keep people thinking about the wrong things.

Another useful (to Google) aspect of hiring an ethical AI team is also that is allows Google to shut down ethical concerns raised in other parts of its own organization by going "we agree that ethics is an important, but please focus on your own job", i.e. an internal AI team can act as a sort of "controlled opposition". The fact that the ethical AI team is now being undermined within the company after trying to do their job and making clear that they are not content to be relegated to such a role is pretty consistent with this motive.

Yes, and just to expand on what I meant with "thinking about the wrong things", they've got everyone thinking about killer robots and automated weapons systems and surveillance (by government entities). When in reality Google's AI is, and has long been, vastly deployed to understand human behavior for the express purpose of maximizing advertising revenue and modification of behavior through pervasive messaging, recommendations, addictive dark patterns, etc. They've basically built a completely automated real distortion system designed to trap people into spending their time staring at whatever the highest (fully automated) bidder thinks they should see. It's dystopian and all the local incentives are wrong now. They've got a crapton of money, computational power, and self-justification. The reality distortion field is powerful. We should be very wary.

Yeah, it was a PR move.

No, Google should never have messed with their researchers’ publications; it’s always the coverup that becomes the story. Contrast Google Research with Microsoft Research, which enjoys a stellar reputation and independence.

This is upside down.

There is no reason for Google to pay researchers and to not have some ownership over the process or outcome in some manner, in fact, it's unreasonable to contemplate otherwise.

Google was exceedingly gracious by sponsoring the work and having it make public in the first place.

The demands made of the researchers in question were quite minor, and completely appropriate, and frankly immaterial to the thrust of concerns anyhow.

To boot, accusations of 'racism' are vicious and repulsive.

If researchers want to be completely independent in the terms they might expect in academia - that's entirely understandable - perfectly fine - but they'll have to find jobs elsewhere.

Google has a Search Engine and bunch of other products, and to suggest that their major investment in research should entitle them to some kind of fair parameterize of the process is very reasonable.

To remedy the situation, and an obvious failure on Google's part, they should spend some time thinking about where that line is drawn, and what kind of expectations should be in place. In retrospect, it seems like a management failure for there to have been any misunderstanding on that front. But the point is -> there is a line. Researchers that don't want to with those parameters can make the choice not to. That's why we have both private and publicly funded research.

Absolutely, that's the only sane stance towards it. The original poster seems to imply that it's impossible to do sustainable ethical business because of conflicts of interests, but there are many ways to deal with such conflicts of interests without firing researchers or suppressing papers.

This case was really just mismanagement by Google. They could have just as well taken those researchers seriously, allow them to investigate whether there is systemic discrimination and then try to correct it, as was the whole point of hiring them in the first place. AI ethics researchers all over the world are publishing articles about the dangers of serious discrimination by large language models. Not addressing this and firing researchers who were working on it on their behalf is not going to help Google in the long run at all.

How do you know that the publication wasn't pulled because they genuinely thought it didn't meet their quality standards? When this happens at Microsoft research, the author makes a few revisions and moves on rather than accuse Microsoft of racism.

Review of the quality of publications should never be made by your indirect superiors. If you want to have an internal quality review process, fine, but the people have to be peers or immediate superiors, and the criticism must be public and must not be related to the company's bottom line, in order to allow for revisions to increase the quality before peer review.

Gebru and Mitchell were both previously employed by MSR in the same context without issue.

Google didn't allow the team to make revisions. They stated the paper couldn't be published, period. The difference here is very clearly the management, not the researchers.

The paper was submitted late. Because of this, and probably previous actions, there was a special go/no-go review instead of a weeks-long back and forth.

If the paper was good it could go through the process and be ready for the next conference. Nobody is seriously suggesting that those words were axed forever.

This is factually inaccurate in a variety of ways.

> The paper was submitted late.

This is not true. The paper was submitted for internal review with about the median amount of notice. The normal internal review process was and is usually very quick, it is not normally a weeks-long back and forth. Importantly, the paper actually passed the normal review process before being submitted to the conference.

> there was a special go/no-go review

It's unclear to what extent the special review was a go/no-go review. It may have been. The review however was not due to the paper being submitted late. As far as I know the special review was actually only done retroactively, after the paper had been submitted externally and approved by the normal internal process.

> If the paper was good it could go through the process and be ready for the next conference.

Even after submission for conference review, there was ample time for back and forth and review internally. Timnit asked if there was feedback she could incorporate into the paper, as even after the paper was submitted to the conference, they could revise it. Google's response was initially to not provide any specific feedback, and then to say that the feedback could not be addressed, the paper could only be unsubmitted.

> Nobody is seriously suggesting that those words were axed forever.

Actually yes, that appears to be precisely what happened.

> This is not true. The paper was submitted for internal review with about the median amount of notice.

This is true. There was a requested amount of time and it was not met. To show discrimination you'd also have to show median quality and median editing work required. This paper had a lot of flaws that would be hard to correct without writing it again.

> Importantly, the paper actually passed the normal review process before being submitted to the conference.

But not the corporate review where the people paying the money decided if they thought it was good to attach their name to.

> It's unclear to what extent the special review was a go/no-go review.

Not at all. There was a strong 'No Go' message given. That's how you know.

> The review however was not due to the paper being submitted late.

The "don't bother fixing stuff, we're just kiboshing it" part was.

> after the paper had been submitted externally and approved by the normal internal process.

The normal internal process is about writing quality and fact-checking. The corporate review is entirely separate and always last in the process in order to see the latest version.

> Even after submission for conference review, there was ample time for back and forth and review internally.

No, not for the level of rework they thought it should receive. If it had passed then sure. The reason to meet the deadline is to handle the situation where it doesn't pass on the first go.

> Timnit asked if there was feedback she could incorporate into the paper

And her bosses decided that the paper wasn't worth the last-minute work. I imagine they felt she submitted it late to game the system.

> Google's response was initially to not provide any specific feedback, and then to say that the feedback could not be addressed, the paper could only be unsubmitted

Right, because it wasn't feedback to her. That ship had sailed. It was confidential feedback to her boss about their belief in the quality of the work. To share that feedback would put innocent employees in a Damore position, attacked for feedback they'd been asked to write. Given how she attacked her boss by name it's reasonable to think she'd have included the coworkers, where they'd be pounced on by an angry twitter mob.

> Actually yes, that appears to be precisely what happened.

No. The preprint was released. Google just said "not in our name". And now that she resigned she has the freedom.

Exactly this. Indeed, looking at Timnit Gebru as an example of someone who worked at both, I'd be very surprised if her paper that was censored at Google would have had any issues in being published if she were still working for Microsoft. Despite the research being, by implication, critical of a growing market segment in Azure, regarding its environmental impact and so on.

> You are basically paying people to shit on your own products. Makes no sense.

Self regulation is preferable (to a private company) to industry-wide regulation, which in turn is preferable to government regulation.

If you want to avoid government oversight, either have internal oversight with the appearance of independence. Alternatively, have an "independent" overseeing body for your industry (whose behavior you may still indirectly control via nominations, seats on boards, funding, etc). Having an actual independent body providing oversight, appointed and controlled by an arm of government body is the worst outcome, yet it is inevitable when lapses arise, and voters demand "something be done" and you failed to provide a veneer of oversight, like Google is doing here.

> funded by government funds, maybe even additionally with an AI regulation fee

Ahh, rent-seeking at it's finest. Do we really want to use the tax payer's money to fund this type of behaviors[0]?

[0] https://syncedreview.com/2020/06/30/yann-lecun-quits-twitter...

It's good for Yann LeCun to quit Twitter if he's going to say silly things like "AI bias can always be fixed by collecting data" which is not true and not how his own employees fix issues - it's certainly not how the other FAANGs solve problems involving bias. Problems are perfectly capable of happening even if your data is the best it can be!

> he's going to say silly things like "AI bias can always be fixed by collecting data"

Except he never said that. He pointed out one algorithm was lazily trained on a non-representative dataset and that it was, therefore, producing biased results.

The model was also new, so you can't be sure improving the data without doing anything else would fix it. You have to actually try it first.

> Google should have never started hiring ethical AI researchers. They should have hired AI researchers and made their stuff available to independent ethics researchers.

If you're going to submit your work to an external, independent review board, it's even more important that you have internal people working to address related concerns.

And what qualifies anyone to be an 'AI ethicist' anyhow?

This is a heavy question - it's problematically like elevating an individuals personal moral position, with all their little trivial baggage to something more than it is.

The challenge with these roles is very real: self awareness, objectivity, and processes for arriving at that are probably key. I suggest it would require a lot of engagement, feedback, possibly legal knowledge, reference etc..

AI credibility aside, the individuals chosen by Google previously were not remotely up for the job on a professional level, this was apparent from their public posture and even writing skills, and that's not slander, it's hard job.

As the OP indicates, some degree of objectivity as well, which may be difficult from within the company.

Finally, I would say it's entirely doubtful if AI requires it's own special branch of ethics. Everything Google does touches on ethical issues, and 'plain old search' much more so than perhaps anything else - certainly more than any of their core AI. It's a business and technology issue, not an AI issue. It just may require some insight from those familiar with the company.

All of that said, Google may lose the 'PR cover' of having a woman or PoC in the driver's seat.

While I don’t personally think her race or gender should matter, the new head of their AI ethics org is Marion Croak, so they’ll retain the PR cover you’re describing.

The last sentence here makes no sense - this is Google RESEARCH we’re talking about, and the issue is around the suppression of research, the sneaky modification of the papers by lawyers and then doubling down on the ensuing shitshow of recriminations, dirty-laundry-airing and anger of 2700 employees by firing people. Including the founder of the AI ethics unit.

What’s their top priority?

Optics? Or better AI? People doing research are not automatically critics, but it is their literal job in this area to ask the hard, hard questions, this is not Harry and Megan go Oprah, this is technology with far-reaching consequences.

Well, Google says they didn't surpress research, but that the researchers didn't follow process. Now we are all having an argument - we wouldn't have that if the researchers were truly independent.

not just that, they fired a toxic employee who was bullying her colleagues on social media and even tried to bully her bosses by throwing ultimatums.

This whole drama has nothing to do with neither AI nor ethics.

> You are basically paying people to shit on your own products.

I mean, the only way to fix bad stuff is to find it?

Sure, but who is going to have a fair look? Someone who is employed by the company or someone who is independent?

VW faked their emissions - EPA uncovered it. Google is harming people with bad AI models - Google is uncovering it?

This works if the company is actually interested in finding the mistakes (i.e. MS finding security bugs in Windows), but not if the findings threaten your business.

MS has an interest in fixing security bugs, because it makes their product better.

No big corp has an interest in fixing their ethics, it's usually bad for their profit margins.

Actually it wasn't even the EPA that uncovered it. It was academia. Corporations and governments just don't have the curiosity built in to their culture that academia does.