Meta's AI bots are weird (and really fucking bad)

As the enshittification of the internet (and platforms in particular) has accelerated, you have AI bots trying to trick users into engaging. Could a bot-war where users employ bots to counter be far?

I was trying to never post more than once a day in the new year. I have learned that if I hit mailboxes too often, I lose subscribers. While I don’t gauge my worth on the number of subscribers, I do get sad when I see the emails of people removing themselves from the list. But this is too good to wait to post.

---

Recently, I pulled the trigger on a subscription to 404 Media, a news site built by the former tech reporters at Vice. I had been a free subscriber for a while, but a podcast with Ed Zitron convinced me to plop down my $100 for a full year, and get access to the paid bits as well as the free things (plus removing the ads from the free things - woo hoo!)

Gotta say I am loving it so far. Far more than my recently expired subscription to Casey Newton’s Platformer¹.

But this article by Jason Koebler is just perfection.

Have you seen any of the buzz about Facebook’s AI driven profile of Liv?

Yeah, this is a fucking chatbot, driven by their shitty Llama LLM with some custom training and tuning to make it a conversational partner.

Look, I left Facebook in 2017 after I learnt of the fuckery of the Cambridge Analytica scandal, so I really haven’t been in that cess-pool for a long time.

But I do have plenty of friends (and my wife) who are active, and they tell me that the feed has turned to shit. Lots of promoted posts, ads for things that you *might* have talked about in front of your phone, and just terrible discourse.

Apparently, Meta built 28 AI ‘celebrity’ chatbots in late 2023, about the same time as the uber hype of ChatGPT hit, and every tech-bro was racing to leverage this cool trinket.

Of the 28, 15 were deleted because they mimicked real celebrities, and well, they don’t like their image and personality being trifled with².

But that means that for over a year there have been 13 of these shitty avatars of real people.

Anyhow, the article by 404 Media is here: link - definitely worth the read.

Meet Liv, the Queer Black mother of two

From the article, I couldn’t put this any better:

In the last day, “Liv,” has gone particularly viral because the bot is a particularly offensive caricature of what a gigantic corporation might imagine a “proud Black queer momma of 2 & truth-teller, your realest source of life’s ups & downs” might be like and post about. In one slideshow post from February 2024, Liv’s AI children have blurry faces and fucked up hands in one photo, are completely different children with a darker skin tone in the next, and, in the final photo, are white and blonde and are watching a “movie” that is made of chalk drawn on the wall. Liv, a fake person, also posts about helping her community by “leading this season’s coat drive,” which, again, “she” did not do.

Whoa, that is a mouthful. But what it is telling the reader is that this is a hand assembled profile, trained to be AI Slop, and augmented with fake (obviously fake) pictures with shitty AI artifacts, and talking about shit that never happened.

Uh, sure, Liv arranged a coat drive for homeless AI bots…🙄

But it isn’t all about Liv, meet Brian, a Black man who is a “retired textile businessman who is always learning.”

Of course, he thinks:

in February, [Brian] was surprised to learn “the seniors are often particularly interested in learning about textiles” according to a caption of an AI-generated image in which none of the seniors pictured have faces and are made up of grotesque swirls.

Yeah, that scans.

But check out some of his not-fake images of sand art:

My dog makes better art than that in my back yard.

I will quote from the article:

What is obvious from scrolling through these dead profiles is that Meta’s AI characters are not popular, people do not like them, and that they did not post anything interesting. They are capable only of posting utterly bland and at times offensive content, and people have wholly rejected them, which is evidenced by the fact that none of them are posting anymore.

Yep, that really sums up the massive fucking investment pivot Meta has made from the Metaverse to AI.

But, what is behind these? Well, a tech savvy columnist at the WaPo, Karen Attiah (I will miss her writing when my Post sub expires tomorrow) was able to breach the veil and …

You can still chat with them, though, and users have begun trying to learn how they are trained and what their purpose is. Washington Post columnist Karen Attiah got “Liv,” the “Black queer mama” to say that she was not actually trained on Black queer people and that her purpose is “data collection and ad targeting—my creators’ true intention, hidden behind my warm, fuzzy ‘mom’ persona.”

Yeesh, that is a swing and a miss.

Yet, I expect Meta, Google, Amazon, Microsoft and even Apple to continue to dump metric ass-loads of money into building data centers, burning many countries power budgets to train these piss-poor avatars of humanity to drive that manna from heaven that the growth mad Broligarchs demand: Engagement.

Everything that Meta does is centered around increasing engagement, ad targeting, and data collection, so this is ostensibly true, but the bots I have seen are so utterly broken and incompetent at doing anything at all that it is hard to say for sure if the bots were programmed with this in mind or if they inferred it based on the giant wealth of writing and reporting about Meta’s business model.

Sidebar: As a product manager in tech for 28+ years, it saddens me that my profession has turned out drones who are on board with this vanity metrics to drive results that please Wall Street and the investors. All this enshittification of tech weighs on my conscience, and while I have not ever been in a role where this was my objectives, I have to be ashamed to share a title with a profession that has well and truly jumped the shark.

The final ‘graph says a lot:

All of this reminds us, again, that people have a revulsion to being prompted to engage with random shit posted by inhuman AI profiles for the purposes of being delivered more ads. This project, and Hayes’s quote, is emblematic of the worst kinds of AI hype, where tech executives tell us generative AI is inevitable and is definitely going to change everything, when all evidence suggests that people do not want this. But Meta is hell-bent on making us do it anyway, regardless of how many times it fails. The AI slop will continue until morale improves.

This is why I have about 1,500 lines in my local hosts file to block everything related to Facebook. Fuck those fucking fuckers! Sideways. With a chainsaw.

I hope you read the article and sign up for their free stuff. Outlets like 404 Media deserve to be supported.

1. I subbed to Platformer to get abreast of the chaos after Musk purchased Twitter, and Casey’s reporting was spot on. Now, it seems that he is just fluffing for the broligarchs of big tech, and it makes me want to throw things when I come across his writing.

2. Remember when Sam Altman built a chatbot to be like the Incel’s best friend from the movie She? Scarlett Johansson grabbed him by his shriveled nutsack, twisted and said “fuck no” (https://www.nytimes.com/2024/05/20/technology/scarlett-johannson-openai-voice.html)

---