,,The ML Ethics, Transparency, & Accountability team was one of a kind''
Sure, something nobody has heared of and nobody knows why it should be at a company like Twitter.
Sure, something nobody has heared of and nobody knows why it should be at a company like Twitter.
If you don't know you're the problem tbh
We're living in a world where voting rights of pensioners have been delegated to big companies like Vanguard and Black Rock. They clearly are not keeping the best interest of the shareholders which would be their duty, and vote politically instead.
If you look in the medical industry for example ethical decisions are always managed by a different group than Product/Legal because it's a completely different set of skills and competencies.
Even more so at places like Twitter which are heavily dependent on ML models to make real-time decisions. And so you need a dedicated team that is proactive rather then reactive like a Legal team would be.
Putting the "what will our users think" question in an ethical framing is not obviously (to me) going to benefit users. (Especially when modern fashionable ethics centers around utilitarian normative ethics, but that's besides the point.)
Some companies, such as for instance Procter & Gamble or Apple care about their brand equity a lot, since they rely on it to charge above market premiums. Twitter needs to care about that too, since if they get caught doing something unsavoury they will turn toxic to their advertisers, or at least those who care about their brands. I am not a marketing exec but if Twitter now drops the pretense of caring about ethics you will see major advertisers pull out, and make a point of communicating why they do it. Thats the ROI on having functioning ethics teams.
I consider much of the practices of social media companies designing for dopamine hits to be unethical, but it had been very lucrative for shareholders for more than a decade. It’s still lucrative but shareholders are taking it on the chin for other macro reasons.
But that wouldn't be the crony capitalism that we're stuck with now. Also, things have been moving too fast for a long time, making it nearly impossible for governments to keep up.
As political discourse and more elections are swayed by rage-bait and disinformation pumped into voters retinas by these platforms, the political risks to these platforms and their bottom line will increase (and become wildly unpredictable).
ML ethics and accountability is beneficial to both society and to any company which has an interest in self-preservation.
[Stephan Livera Podcast] SLP421 Vivek Ramaswamy - The Antidote to Corporate Woke-ism https://podcastaddict.com/episode/146353684
Actually almost everyone who is in the Data/ML space knows about it.
It's one of the biggest challenges in ML which is making sure that models don't amplify biases in the training datasets and cause harm to users and the company. There have been many stories with TikTok for example recommending accounts which promote unhealthy eating [1] to users with a known eating disorder.
[1] https://journals.plos.org/plosone/article?id=10.1371/journal...
If you are going to be hosting petabytes of data and in the business of showing advertisment, a lot of recommendation algorithms work in the background. A lot of times they aren't perfect and end up showing information which could be very high correlated but culturally sensitive, offensive or avoidable. People working in fairness and accountability run a lot of experiments to make sure biases don't get amplified in service offerings.
Social networks like Twitter use ML to boost engagement.
It turns out this tends to lead to ML amplifying pretty destructive patterns... racism, conspiracy theories, ideological bubbles, anti-semitism, etc.
A company might wonder how to boost engagement without promoting these things.
Good points. I agree with all you write.
However, Musk's stated intent when buying twitter, was to make it a non-partisan town square. This comes with a context of the US currently being divided more and more into two competing "tribes" of similar size and power. (Ie. the perfect setup for a civil war.)
If you look at the "ethics" teams kind of like law enforcement, it's important that they are neutral to this partisan divide, and don't have a "My side good, other side evil!" attitude.
If Elon is of the opinion that Twitter has been engaging in a partisan way, it is natural that roles like this are torn down and built from scratch, where the new teams are thoroughly screen to make sure they have as little affiliation with each of the tribes as possible, and to make sure that any affiliation remaining is balanced between the tribes. (Kind of like one does when recruiting a jury for a high profile court case, and also how presidents should select supreme court justices)
Hopefully, Musk goes through with this the way he has stated, by aiming for non-partisanship, and not in a pro-Republican way.
My reply was,
"you're a global sentiment barometer"
Because Twitter, YouTube and Facebook have gotten under fire multiple times for the shit their artificial "intelligence" did - promoting content harmful to society (antivaxxers, flat-earthers, antisemitism, ...), sending them into outright radicalization spirals (YT autoplay), banning of people by automation based on shoddy input or without context...
If you're running AI that has a direct impact on wide society and you're not running an ethics oversight program, you should be arrested for endangerment of society.