Isn’t one of the big complaints about social media that only like 1% or .1% of users get promoted?

I think the problem is your last paragraph buries the lede. - the social media companies want to make this about content, because algorithmic feeds are what make meta a trillion dollar company instead of a 50 billion dollar company.

I do think there would be a fair bit of agreement with your thesis if your minor caveat at the end wasn’t the big huge problem.

I think most of these bans are because no one has figured out how to ban the algorithmic feed and SV isn’t self-regulating.

There would still be plenty of moral panics about the content, but I think the bans are picking up speed because of the form.

I don’t care that my kid wants to join a satanic death cult I care that Meta has figured out how to monetize satanic death cult membership in the algorithm.

The biggest problem to me isn’t the content or posting on a friends timeline. It’s that choices are pushed to the user rather than pulled. Change that one dynamic and I may not care as much that YouTube has content I consider toxic. Add a little friction my teen has to go actively discover content he’s curious about, not pushed an algorithmic feed.

I think the problem is though - the tech companies are incentivized to make this about _content_ “ I can’t ban Andrew Tate because it’s not my job to tell people what they can and can’t watch, first amendment, blah blah blah.

And Zuckerberg isn’t wrong here exactly, but per above the real argument I think is less about content and more about form.

But the reality is these companies don’t actually make their money from the content they make it from the form.

So Zuckerberg can’t concede “oh you’re right, making the newsfeed less adddictive doesn’t violate first amendment rights, at least we’re not banning Trump, because he doesn’t care about banning Trump except for the heat he gets from republicans, he cares about you messing with the money printer.

The thing we may all be able to agree they should do is the thing they’ll push back much harder against doing.

I think there are plenty of adults who think maybe there are certain types of social media patterns that should be banned outright.

It’s not whack a mole - I _want_ my kid to engage with a replacement technology like gaming. I’m not worried about my kid playing GTA vi all day. I’m not worried about content. I’m not even really worried about Andrew Tate. I’m worried about the specific way mindless algorithmic applications mimic cigarettes more than “satanic heavy metal rock concerts”

My kid needs more satan in his life! And less growth hacker.

Failure? Or 30% improvement.

“While social media can be weaponized for bullying or hate, over half of all teens turn to social media as a source of social and emotional support.”

Teens also turn to drugs as a source of of social and emotional support.

“ Adult saviorism is rooted in the belief that teens are, again, passive victims but who need to be rescued by adults.”

Ironically the thesis of this article. In fact I would say by banning teens from social media, we may be implicitly increasing their agency, since they’ll have to figure out how to interface with a set of technolgies that are not optimized for them.

There’s something really dystopian to me about “designing for teens in mind” and I think maybe the unintended consequence of this kind of thinking is just - the continued marketable capture of segments.

What this person is basically advocating for is a new way for tech companies to extract more value from our youth.

The whole point of being a teenager is to generate your own ways to rebel against the fact that the adult world isn’t quite built for you.

This “facebook but for teenagers” is just another opiate for the masses.

I'm sure I can sort of prompt it to try to give the answer I want but some people talk about "I don't want an LLM with a political perspective, just provide all the perspectives equally and without bias."

And to a certain extent they have to be constrained. Anything more than a 500 word response I'm probably not going to read, particularly if I want to go deeper on something I want to navigate the conversation. But it means that tuning things like sycophancy just create sort of extremely simple mitigations. Even when I say "May" or "slight chance of" I'm often corrected "Be careful to make too blanket of a statement." I wonder if part of this has been the move towards optimizing towards the agent experience.

Likewise a bias towards "Say things equally without bias" is an exceptionally poor way to actually explore the nuances of an idea, with this sort of regression to the mean.

I don't know if they'll be able to solve this problem broadly. I mean, LLMs probably shouldn't be therapists/coaches/philosophical gurus. But I could anticipate a growth in sort of groupthink and regression to the mean in terms of perspectives broadly speaking as a society.

I find them useful. Maybe it's better I find them distasteful to talk to as well.

It's usually the first step I take when writing research. I don't even read the analysis, just go straight for the papers.

They usually provide me with very relevant seminal papers and I can then easily expand from the bibliography in those papers.

It basically solves discoverability, whereas in the past I would spend days and would need to ask colleagues and librarians for help, to achieve what I can get in minutes now.

But again, I value the actual output of the LLMs at bellow zero factually wise because I need to check everything they say if I want to internalize it and use it for any end. Pointing me to sources is the perfect use case. Really, I cannot emphasize enough how little I trust the text LLMs produce, I actively avoid reading it because I really don't want to risk ingesting spurious information that I might later repeat or rely on for any reason.

But they're response is kinda always.

"Python? Have you considered that some people prefer, javascript?"

To your point, I think I'm asking too much, and would have a better time if I adjusted my approach.

What they are good at is finding, summarizing, and providing links to perspectives from other humans.

Least privilege and say only opening ports or installing applications an application needs to operate are extremely standard security practices.

We talk about a firewall blocking exultation of data, why not blocking exfiltration of your agent ?

I think in some ways we talk about Silicon Valley leaders in this like objective points based system. Elon musk being a trillionaire must have the highest IQ, maybe Steve Jobs was second highest, Dario Amodei is number three right now.

I’m not sure I think anyone believes this outright but a lot of the time we talk about it like we do. And that means because Elon musk has the highest iq he must be more right about most things than most people.

I think a better mental model is that these folks are a combination of somewhat remarkable ability (99.9th percentile? Maybe 99.99th?) combined with something that could be described a negative - an 99.99th percentile level of obsession with an idea. There’s obviously a huge difference between Elon musk and a guy who stands on the street corner shouting at people for sinning, or a cult leader to group suicide his cult - but maybe that difference is what happens outside the obsession, the trait is remarkably similar.

The point is - the level of obsession it takes to convince thousands of the smartest people on the planet to work for you, and collect billions of dollars from investors is the same obsession that gives you a blind spot into where the bottlenecks are between say here and fast takeoff.

Dario is probably not wrong about what he knows about AI, it’s probably everything around it is where his blind spots are.

“We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain.”

He’s basically just saying, I dunno, spend 20% of every sprint building defensive capabilities.

Because 75-80% still allows them to overhaul society.

And in their defense - I do think they honestly believe in the dangers of AI and think that defensive measures have to be built in, but yeah, as part of the overhaul of society.

Right later in the article he says:

“I believe that if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong.”

This is not a broad societal, maybe we don’t take full advantage of AI and destroy jobs. It’s just - let’s try not to have AI hack into all the critical systems and launch nukes.

In 5 or 10 years everyone will once again be “why did we let this happen?”

Like we can just do stuff. We can fine companies, take away licenses. We are capable!

But I also think - I’m not trying to be too negative here because certainly we should be investing in scientists and other discoveries, but early tech - xerox park, googles 20 percent time. They were playing in an extremely immature space.

You could throw 20 ideas at a wall and create a billion dollar business.

We should do what you’re saying but we should also build institutions and maybe also limit the extent to which we’re building Elysium. If we can build AI models that rival human intelligence we can create laws to protect its dignity.