First, I wish we'd be a little more imaginative. While other world leaders paint visions for the future, we're starting to sound like a broken record: 'tech firms need to stop extremism on the internet'. (Thread).
There is still a mistaken view that algorithms will be able to spot 'extreme' content automatically. It's the algorithm fantasy: 'they can predict your next book, they MUST be able to spot extremism'. It doesn't work like that. 2/??
There are some obvious use cases - esp known terrorist videos / images which can use i.e. Microsoft PhotoDNA to auto-remove known cases that are held on databases. Beyond that, however, it always requires human analysis.
Another plausible use case is the applying machine learning to help more efficiently & accurately 'surface' cases which humans moderators can then make judgements on. However, even this will never resolve the problem.
And besides - we slip too easily from 'terrorist' content to 'extremist' content. The former is easier (though in fact still contested and hard) to define, and is illegal. Extreme ideas are not illegal, and is often a matter of one's perspective.
By endlessly asking tech firms to remove extremist content, what we are in fact demanding is that some nervous risk averse content moderators with jumpy investors in another country decide what should be in the public domain.
I worry we will create a dangerous precedent. We demand extremist Islam content is removed - but things have a tendency to expand. Then it's radical right wing, then radical left, environmentalism, dangerous libertarianism, etc. It always ratchets up, not down.
There is a great value to society of having radical and even extreme ideas available - to force people to think for themselves & develop their own critical faculties. Mollycoddled people won't be able to reject bad ideas if they've never even seen them.
In short, this is one of those very few ideas that really is both unworkable in practice and undesirable in principle. <End Rant>
