Live data from Hacker News

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

theguardian.com

341–350 of 605 posts

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

#341

> Conversations with the characters quickly went sideways when some users peppered them with questions including who created and developed the AI. Liv, for instance, said that her creator team included zero Black people and was predominantly white and male. It was a “pretty glaring omission given my identity”, the bot wrote in response to a question from the Washington Post columnist Karen Attiah. This is probably tr…

I hate the journalists pretending that ML is regurgitating facts instead of just random tokens. The ML model doesn't know who programmed it and can't know because that wasn't in the training data.

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

#343
post #290

You know what AI profiles I want more of? Clueless grandma/grandpa that accepts scammer calls and waste their time. Such as UK's O2 Daisy: https://news.virginmediao2.co.uk/o2-unveils-daisy-the-ai-gra...

There is another real-world version of this in the US healthcare system, where doctor offices are using domain-specific LLMs to craft multi-page medical approval requests for procedures that cover every known loophole insurer’s use to deny, which are then being reviewed by ML-powered algorithms at the insurance company looking for any way to deny and delay the claim approval. In other words we have a bona fide AI arm…

The year is 2035. To cut costs, both insurance companies and providers removed the human from the loop long ago setting off an adversarial process between the LLMs on both sides. Medical insurance claims are now written in an ever changing format that resembles no human language. United Healthcare has just announced a $10 billion project including a multi gigawatt data center to train its own foundational model to keep ahead in the arms race. UNH stock is up 5% on the announcement.

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

#344
post #151

Earlier quoted context omitted.

Ironically, this would lead to dramatically unhealthy social interactions and be ripe for abuse. You are literally creating a bubble of interactions with technologically enforced rose-colored glasses. Get the right person in charge of the experience, and don't be surprised if it becomes a modern take on Orwell's 1984.

In what way would it become a modern take on 1984? In terms of surveillance? Afaik the telescreens were non-interactive, right?

No, I'm not even interested in the surveillance here (though it is also a good comparison). Full disclosure, I'm very pessimistic on what current LLM models are capable of, and even more pessimistic on the impact of social networks.

I'm referring to the control of interactions and information. Take a look at the character interactions on the novel.

People socialized as they could, but interactions were not only heavily monitored (to your point), but altered as well. Winston struggles to behave correctly in public knowing that one small misstep could result in him being arrested by The Party. He revels in the chance to rebel against it, only to find that the opportunity to do so was in fact a trap. Even the allowed language was neutered as much as possible to prevent communication. All of these virtual 'friends' will likely politely comment and correct your posts as they see fit.

As for information, "we have always been at war with Eastasia". People in 1984 were clearly being fed false information as part of control. History was altered, facts were changed, it was pretty heavy handed. But imagine what this personal echo chamber could do with a new concept or idea (that virtually all parties you interact with corroborate, along with LLM generated 'news' posts and generated pics that support it), how many people will verify? Just think of how many folks you've seen have been misled by a single well-done "fake news" post; now, imagine that the network itself is confirming it at every turn (note Meta owns multiple networks).

Oh, and as a bonus, these virtual friends are really enjoying Brand Product. Have you tried it yet? Folks keep posting pictures of themselves with Brand Product, and they look pretty happy!

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

#345

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’.

Having worked at a similarly gargantuan and dysfunctional company, I can tell you exactly how this went down. Someone had this idea for AI profiles. They speced the product and took it to engineering. The engineers had a good laugh at how preposterous it is, but then remembered that they get paid a ton of money to do what they're told, and will get promos and bonuses for launching regardless of the outcome. It all st…

You're kind of missing the early step where some executive had to sign off on this dumb idea. Otherwise it doesn't launch. It's only "impactful" to engineer performance review because some exec said so.

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

#346

Earlier quoted context omitted.

The next step will be to release bots that aren't labeled as bots. They'll be influencers (advertisers) without having to pay a person. They'll produce hyper targeted influencer slop, hocking products directly to individuals using Facebook's knowledge graphs of users to be incredibly manipulative. Companies will pay Facebook to make people siphon money directly to them. It'll be like entirely automated catfishing.

1. Create AI bot 2. Wait until it gets lots of followers 3. Sell mentions of products by popular AI bot. Profit. Sure, they faltered at Step 2, but how was this not the plan?

The future bots won't need followers. They will reach out to engage users based on their ad graph. Like how catfish accounts work.

They'll reach out to users and talk to them about whatever they're interested in. They'll then make influencer-style native advertisements. If you're a middle aged man that likes video games when you see their picture posts they'll be "wearing" some vintage game t-shirt (that's for sale). If you're instead a twenty year old woman into yoga the same bit will be "wearing" some new Lululemon yoga pants.

The first pass of these bots failed because they used the follower mechanism. The next version will just follow you to push ads or scrape more data about you.

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

#347

Earlier quoted context omitted.

Having worked at a similarly gargantuan and dysfunctional company, I can tell you exactly how this went down. Someone had this idea for AI profiles. They speced the product and took it to engineering. The engineers had a good laugh at how preposterous it is, but then remembered that they get paid a ton of money to do what they're told, and will get promos and bonuses for launching regardless of the outcome. It all st…

The best (worst?) part is when the engineers actively overlooked actual production bugs and security concerns to do this work.

There are 10,000s of SWEs at Facebook and this project was at most a handful of SWEs. (As stupid as it is, it did not significantly detract from prod bugs.)

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

#348

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’.

It's like people have already forgotten why people used the non-human beings in Westworld. One of the first things we're going to use them for is amusement and having them represent anything sentimental or sensitive is going to make them a target like this. It was irresponsible. There's something essentially wrong with Meta and Google where they can do tech but not products anymore. I'd argue it's because the honest…

Fictional dystopias aren't the real world. Westworld is just television.

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

#349

Earlier quoted context omitted.

It’s so bizarre to me that this uniquely US phenomenon of for-profit-middlemen inserted into the healthcare system has resulted in an adverserial relationship between the sick person and the “healthcare provider”. I put that air quotes because insurance companies don’t actually provide health care. They provide insurance. That’s a financial product, not a medical one.

Death panels.

What about them?

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

#350
post #345

Earlier quoted context omitted.

Having worked at a similarly gargantuan and dysfunctional company, I can tell you exactly how this went down. Someone had this idea for AI profiles. They speced the product and took it to engineering. The engineers had a good laugh at how preposterous it is, but then remembered that they get paid a ton of money to do what they're told, and will get promos and bonuses for launching regardless of the outcome. It all st…

You're kind of missing the early step where some executive had to sign off on this dumb idea. Otherwise it doesn't launch. It's only "impactful" to engineer performance review because some exec said so.

Execs go through the promo process too. And also, some execs will sign off on projects that they know are bad but will make for good promo material for them and their reports.
Post reply on HN