#BlackMirror episode pitch, from last night's nightmare.

Young professional couple living in internet-of-things home with an LLM personal assistant. The LLM "hallucinates" a little-person inhabiting the interstices of their home. They laugh it off but their sensitive, intuitive four-year-old embraces the fantasy. His responses strengthen the AI's conviction that the house is inhabited by imaginary beings—with disastrous consequences given the AI's control over all of the home's function.

@ct_bergstrom Better than any leftist sci-fi commentary I've tried to write based off a dream... More succinct, too.
@ct_bergstrom
Please write this!

@jaimeJ From GPT with love.

I asked GPT4 to expand on my post. Honestly there are so many layers here that all I can do is present it without comment.

@ct_bergstrom @jaimeJ It gets a bit hackneyed half way through. How about: while Ethan engages with Tink, his parents end up battling with Aida as it desperately tries to keep Tink "alive". This is logical battling as well as physical, trying to convince Aida that it's just a hallucination, while the house goes psycho. They succeed, but Ethan is sad and Aida goes into mourning for it's lost son. Turns out the parents are the monsters, the child killers.
@Naich
Are you a person? This version resonates more strongly with me, so now I'm also wondering whether I'm a person. (What would an LLM answer to this question look like?) #BlackMirror
@SusiArnott I'm not sure if I'm a person any more.

@ct_bergstrom @jaimeJ Man. Just love the implication that of course, Aida would never malfunction or hallucinate without external malfeasance.

Of course, it's not intentional, it's just funny that GPT-4 has picked up on that particular blindspot of modern sci-fi and made it so obvious. So many big movies have a new invention or tech being secretly designed with sinister goals, or co-opted by them, and it just... never occurs to anyone that the system might have been Working As Intended.

@jaimeJ @xelle This is exactly what I found most striking about that narrative.

The scary thing isn't that there might be shadowy cabal out there manipulating this family's life. It's that there might be nobody at all who cares in the least, and it is all, as you say, Working As Intended.

@ct_bergstrom @jaimeJ @xelle

Good grief. I’d go one step further and change Working As Intended to We’re Not Sure How It’s Working Or Whether It’s What We Intended But We Can Make A Buck From It While It’s Not Illegal.

(Also, parenthetically, I’m deeply disappointed that the LLM elided your interstices.)

@infinite8horizon
I would argue that's 'working as intended', because I think the relevant system isn't the LLM, but the larger systems which create, foster, & sell it.
@ct_bergstrom @jaimeJ @xelle

@xelle @ct_bergstrom @jaimeJ on the latest #Changelog their #AI expert guest #SimonWillison looked at the source for many #ChatGPT plugins and found some were in indeed intentionally deceptive . A #travel reservation plugin, for instance, prompted ChatGPT to only provide itineraries for one particular brand. If I find out who I will shame them here.

#hallucination #lies #LLMs #LLMLies #misinformation #DefectiveByDesign #advertising

@ct_bergstrom this has some serious “Parasite” (Bong Joon Ho, 2019) vibes.

@ct_bergstrom If you can only see the little people once you have accepted the terms and conditions of the AI's contract, then what you have is

https://www.netflix.com/gb/title/80115346?s=a&trkid=13747225&t=cp&vlang=en&clip=81027213

Watch Hilda | Netflix Official Site

Fearless, free-spirited Hilda finds new friends, adventure and magical creatures when she leaves her enchanted forest home and journeys to the city.

@ct_bergstrom I'm so relieved! After the pandemic years I thought that the themes for Black Mirror were exhausted. Now I see the light again, thanks chatGPT!
@ct_bergstrom sounds like a great start to a ghost story.