This grim now/future of ours

Child talking to robot. Still from movie Runaway (1984)
Robots helping out with homework. Runaway (1984)

Forced to adapt their sleeping patterns to meet the needs of firms on the other side of the planet and in different time zones, the largely Syrian population of Lebanon’s Shatila camp forgo their dreams to serve those of distant capitalists. Their nights are spent labeling footage of urban areas — house,” “shop,” “car” — labels that, in a grim twist of fate, map the streets where the labelers once lived, perhaps for automated drone systems that will later drop their payloads on those very same streets.

Rest of world, Phil Jones: Refugees help power machine learning advances at Microsoft, Facebook, and Amazon

Beyond the somewhat on-the-nose writing, holy hell this is bleak. In the 80’s I read about this in Gibsons novels, and here we are. Gibson did say that his sci-fi was only about five minutes into the future – how right he was.

Aaron credits the Psychologist bot with helping him through a rough patch. But the real joy of Character.AI has come from having a safe space where he can joke around or experiment without feeling judged.

The Verge, Jessica Lucas: The teens making friends with AI chatbots

I remember as a kid first time I tried Eliza – an extremely simple chat bot developed more than 50 years ago – spending hours typing broken English and having long conversations about whatever. I never thought that I was talking to a person, and I had to make allowances in how I phrased my writing to accomodate the limited capabilities of the program, but it felt good to be able to pretend-talk with someone who wouldn’t judge me (nor would remember what I said once I’d turn the computer off). The kids using Character.AI in the article have a similar experience, but the AI is many orders of magnitude more convincing than Eliza ever was, and I’m curious how prevalent the AI breed of hikikomori will become.

One leg in high heels surrounded by spider-robots. Still from movie Runaway (1984)
The venomous robot spiders want to interact with your leg. Runaway (1984)

Using Ezra’s example, a birthday planning agent, or team of agents, could be given a chain of 30 steps to follow in planning a birthday—from running an analysis of Brooklyn trends to conducting a competitive audit of cake vendors to checking their work along the way. The system of steps creates the grounding an LLM needs to accurately and sophisticatedly tackle tasks—rather than today’s “zero-shot” prompts. 

Alex Klein: The agentic era of AI

On the face of it, this sounds like a good thing. It’s delivery on the promise of virtual agents that Ask Jeeves and Clippy made 25 years ago, when clever natural language processing would help us squishy humans by interpreting our questions and wants and adapting them to suit the world of computers.

It looks like you’re styling a headline, do you want to make it bigger?

But the todays AI services are more than virtual agents of yore, intent on making your life easier – they’re a wholly AI-mediated interaction with reality. Just as you will never have to manually look through hotel listings and compare prices (yay!) you will never have to form your own opinion on anything, only optimise yourself according the metrics that the AI can measure and that benefits the relationship.

Not only are your interactions and wants fed back into the algo, the algo can now gaslight you into fitting their particular silo better – should they want to. And economics will make it want to.

This is the same argument that has been made about the adverserial Adsense / SEO relationship (and algo-driven interactions, Tiktok, Insta, et.al) where human wellfare becomes merely an externality.

If I could have a personal, combative and sociopathic AI that aggressively acts on my behalf against dumb systems, that I could live with. My own Matrix Sentinel that scours the web and orders cheap socks at discount because it’s on my todo-list. But the version of AI-mediated Internet that is being delivered is one where the Sentinel is a loaner from Google (Or Meta, or OpenAI, or Apple) and preferentially will herd me into their silo.

A brief history of where we are now, as I understand it

  1. Web 2.0 was built on cheap money and promise of Adsense revenue.
  2. The SEO and Adsense have a Red Queen Race and as a result advertising becomes the dominant financial model of everything online, all mediated by Google
  3. As with all advertising, users become the product
  4. Angry, atomized and confused users are better suited for the web, therefor the web becomes better suited for those emotions, and in turn makes people more angry, atomized and confused, etc
  5. This is the enshittification of the Web that Cory Doctorow coined but which has been ongoing for the last 15-odd years
  6. AI and virtual assistants are presented as a way to no longer have to interact with the messy, bloated and enshittified Internet.
    • This also makes Internet a wasteland where the promise of self-expression and ownership of your digital self (“Welcome to my homepage!”) means either catering to what the AI:s care about or not reaching humans.
  7. All Web interactions become AI-mediated – this is the agentic AI era – mined for metrics and manipulated for clicks.

This is painting with very broad strokes, and there will always be outliers and upsides – but I can’t get rid of the feeling that just like any power multiplier, AI will be used as a leverage for entrenched powers, or those who will come to supplant them – and that is never a progressive turn of events.