Live data from Hacker News

Meta is killing off its AI-powered Instagram and Facebook profiles

theguardian.com

191–200 of 605 posts

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#191
post #99

I don't like that some people, esp at larger companies, think that groups with disadvantages in today's world are cutesy adorable things to play make-believe with.

[flagged]

Never understood this critique. What is the "correct" percentage of Apple ads that should have "non-majority/mixed race people"? You seem to know, so what's the number?

Release an ad with white people in it, and one group complains "Racist! Ignoring minorities!" Release an ad with black people in it, and another group complains "Pandering!" There's no winning this battle.

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#192

I get the hate, but I'd be open to trying out a social media experience that is a mix of human and bots, especially as the bots get better at acting like reasonable humans. Stack Overflow was great when it came out. But the whatever percent of humans who were mean and obnoxious and wouldn't let you just ask your questions eventually ruined it all. Then ChatGPT came out without any humans to get in the way of me askin…

People should try to socialize in real life. The web is 90% rage baits, trolls, people completely brain washed by fringe conspiracy theories and politics IRL these 90% turn into 10%, if you really care about socializing you'll have more meaningful interactions with the grandma living next door than with the queer black queen zuckerbot ™ It's like if people asked supermarkets to hide plastic apples amongst the real on…

What if socializing IRL is something you can practice with pro-social bots?

Or, what if depressed people find the pressure of consequences too stressful to “just go outside and make friends”, and so the realistic options are chatbot or not enough socialization?

There are obvious nuanced issues and risks here but to distill it down to a one liner like “try to socialize IRL” is myopic.

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#193
post #56

Earlier quoted context omitted.

Because my friend who just worked for years to become an American citizen after immigrating here and working for the US government still has other Americans tell him to "go back to where he came from" because he's evidently not American enough for them. A hell of a lot of people want to "just be American" and a lot of Americans won't let them.

[flagged]

> It’s because many of those “Americans” do not try to integrate, want to bring their culture with them...

That's okay, more varied culture enriches, and I enjoy the added richness.

>...and have a “what’s in it for me” view of [insert anything here].

That unfortunately describes a huge share of human beings on this planet.

> The label “X-American” (replace X with a foreign country) should not exist

I disagree: I think it's totally okay for people to recognize a part of their identity (key words: a part of, since a person's identity comes from many sources), rather than being forced to hide it.

> If you identify as this you are marking yourself as different from Americans.

Each individual (Americans included) is different from all the other billions of individuals on this planet (other Americans included), and that's okay. It would be weird to try to be forced to deny that totally-okay fact. Thus, a self-described Hoosier or Floridian or Cheesehead or Californian or Plumber or Texan might differentiate themselves similarly without renouncing any "American-ness". Adding one of many possible identity labels (whether race or origin or sports team or profession etc) doesn't negate all others.

> They will push for pro Indian initiatives (see H1B abuse, etc) at the expense of Americans. Ironically, this is also the cause of the discrimination they’ll face.

Ironically, this claim is itself racist: Asserting that people of a certain race are coming to get you because of their race, and thus discrimination against that race is justified, is racist (and a tired racist trope at that).

Unfortunately though, most racists I have seen are racist simply because they feel the other race is different from them, and feel that their victims are bad and/or deserve less than themselves. I've seen racists remark with nothing less than anger and disgust at baby pictures of children of other races. Babies, dude.

For another example, I'll open up and share an unfortunate case I had with a family member when we were visiting a science museum: Upon seeing how many non-white people were there, said family member made several off-color remarks to me. But these people (many of them children) weren't abusing anything, they weren't doing anything "at the expense of Americans". They were just normal people, many of them likely citizens, learning and having a good time. Them being interested in science isn't a problem, and they weren't stopping anyone else from coming to the museum. It's just that they were disproportionately represented among the sample of Americans interested in science. Good for them, science is good.

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#194

Earlier quoted context omitted.

I can't fathom how anyone thought this was a sensible thing. It's so bad I have to wonder if there's a different angle here, maybe they think that releasing something so terribly bad will make it easier when they release something less comically bad? Idk.

I find myself generally wondering the same thing about half the things coming out of SV over the past several years. I have worked with competent product organizations. I know they exist. A few even exist in SV. But for some reason the largest players have just absolutely lost the plot beyond what can get them to favorable quarterly earnings, and that game eventually and fairly consistently doesn’t end well in the lo…

We aren't a big org and I know we spend several hundred if not thousands of USD a month on meta ads and WhatsApp business.

For larger orgs I can imagine that number is larger and we definitely get a ton of positive interaction from that. That might be swayed by demographics but I can tell you age isn't one of them because our interaction (on WhatsApp Business which is mostly where I interact with Meta products) is decently in the 20-40 range.

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#195
post #187
post #167

This was probably an okay idea terribly implemented. GenAI creators on social media kind of sense. Neurosama, an AI streamer, is massively popular. Silllytavern which lets people make and chat with characters or tell stories with LLMs feeds Openrouter 20 million messages a day, which is a fraction of it's totally usage. Anecdotally I've have non tech friends learn how to install Git and work an API to get this one wo…

> Neurosama, an AI streamer, is massively popular. I think some level of Neuro's popularity is due to the Vedal + Neuro double act though, and some of the scripted/prepared replies.

Absolutely, and in picking collabs with people who are willing to work with the weirdness and make it funny. Vedal is definitely a fantastic creator to make it work so well and the amount of fine-tuning and tweaking he must do must be unreal. But I think it still shows there is some hunger for this type of content, though you are probably correct that it still needs to be curated, gardened, worked with, and sometimes faked.

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#196

Do they not have anyone sensible left in the room? Like, you’d expect that at some point someone would have said ‘these are comically terrible, we cannot allow them to see the light of day’.

That's the price for Mr Zuckerberg's ownership structure. On the one hand, no one can dispute him. On the other hand, no one can dispute him.

I dont think it is reasonable to think shareholders would be vigilant safeguard on these types of R&D programs. They simply aren't that involved.

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#197
post #63

Earlier quoted context omitted.

That seems to miss the point. The bot is likely correct in that its development team didn't align with its identity. If anything, it lived up to its training in essentially calling out its own oddity in the way that black and queer people online did when they found out about it. Having it lie about its dev team certainly wouldn't have helped in this case. The issue is Meta making a bot to essentially masquerade as an…

Most probably the bot does not know the identity of the programmers. How would it?

I don't think it specifically knows, though it could be possible. Maybe there was an article published about the dev team at some point that named some or all of the members. Maybe they have published research or speeches on the topic that names their role/employer.

Either way, the general demographics of people working at Meta is generally known, so I'd assume that was what the LLM was working off of.

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#198

I get the hate, but I'd be open to trying out a social media experience that is a mix of human and bots, especially as the bots get better at acting like reasonable humans. Stack Overflow was great when it came out. But the whatever percent of humans who were mean and obnoxious and wouldn't let you just ask your questions eventually ruined it all. Then ChatGPT came out without any humans to get in the way of me askin…

I can’t think of any way where a social media app using bots to discourage wrongthink doesn’t come across as dystopian, and also a little sad.

> Wrongthink

When a community is formed, an implied set of rules are generated and applied. They evolve and change as time goes on. Sometimes they are the cause of the community dying, sometimes they are the reason is thrives.

Whether its a sports community (woo football is the greatest!) or a specific team (fuck the other team, booo) or cycling. Each community has a set of rules that you need to abide by in order for it to function.

Now, if you go and break those implied rules, you get told off.

A community falls apart when two or more factions form opposing implied rules in the same shared space. For photography, it could be the use of photoshop, digital cameras or, more recently AI. Either the factions learn to get on, or they break away and find somewhere more accepting. That is the natural order.

You could equally present those things as "wrongthink". But more practically its just a regulation mechanism for human interaction.

Now, you'll counter that "big corporations/politics determining what we see is bad", and then reference some time in the 90s where no such system existed. The problem is that the US media was brilliant at self censorship. Sure you could get specialist publications that catered to whatever taboo subject you wanted, just as you can now.

The issue is, online there are no constraints on behaviour. If I shout at 13 year old kid in the street that I'm going to fuck their mum, burn their house down, and generally verbally abuse them, generally someone will intervene and stop me. Thats not wrongthink, thats society. There is not scalable mechanism for doing that on online communities.

Is this AI bit the way to do it? no, its made by insular collegekids who've barely lived in the real world.

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#199

Earlier quoted context omitted.

> They admitted it internally, to me It's becoming an increasingly apparent problem that people don't understand that LLMs frequently hallucinate, especially when asked to introspect about themselves. This was 100% a hallucination.

LLMs just predict the next word, based upon the texts they were trained on. They are, by definition, a hallucination. It's just that, sometimes, they put words together that fool people. They're just ELIZA on steroids, people.

How did ELIZA do on the Advent of Code?

Re: Meta is killing off its AI-powered Instagram and Facebook profiles

#200

Earlier quoted context omitted.

> They admitted it internally, to me It's becoming an increasingly apparent problem that people don't understand that LLMs frequently hallucinate, especially when asked to introspect about themselves. This was 100% a hallucination.

It was likely a hallucination, right? You didn’t check to confirm? But this is a solved problem that Meta chose to implement poorly. It can be done right. I just asked o1 “create a diversity report for openai” and its very first paragraph is: > Below is a *sample* diversity report for OpenAI. Please note that the numbers, programs, and initiatives are provided as *illustrative examples*. This report is not an officia…

Yes, and I suspect that the very first line of the chat has something equally lawyer-y to serve the same purpose.
Post reply on HN