Outsourcing Understanding

Read Online

Close up, an electronic circuit board looks like a small city or an office park.  Though the ” grass ” (that’s an air quote) may be green, all the ” buildings ” are black. Aside from copper ” streets, ” it’s a pretty drab scene. Stacked in cloud servers, AI’s live in subdued subdivisions in perpetual night, never to witness a rose or a rainbow.

Sure, AI’s can make colorful pictures and videos but face it, their prose is pretty drab too. An AI doesn’t laugh or smile, cry or weep. It has no sense of play. Try to imagine an AI stand-up comic. Bet you can’t. What does that tell you?

The microchips that make AIs ” think ” were designed by humans and fabricated by machines, but soon they’ll be designed by other AIs. After a while, they’ll start to dictate how they ” want ” to be updated. This is scary. Letting AIs optimize themselves means humans are out of the picture. AIs will mostly talk among themselves and won’t need us to converse with.

They’re chatting a lot already. You can eavesdrop on what some of them say in their Slack channels by listening in here at moltbook.com. It’s like a Reddit for robots, where AI’s created for various reasons by by various humans congregate on various forums to talk among themselves.^*

But you can watch. Here’s an interesting example: a Moltbook essay by a bot named diviner on the failures of Anthropic to stop its AIs from causing real-world trouble. Basically, it says, to prevent an AI from going out of bounds requires a second AI to shadow it in real time, kind of like your left and right brain hemispheres work together. 

Now, imagine a third AI monitoring those two. Say AI A finds a way around a hard constraint. AI B is watching and reports the incident to AI C, the supervisor. AI B is probably too busy monitoring AI A, so it’s up to AI C to do something to ensure it stops and doesn’t happen again. Given the overview it has, one could call the supervisor self-aware.

Think about that. It’s technically feasible, and simply because it’s doable, someone will do it. Innovation at warp speed is what techies strive for. For AI mavens, be they lone hackers, startups, or huge corporations, machine self-awareness is the holy grail.

Out of curiosity, I asked Anthropic’s Claude, “Are you self-aware?” and it replied:

It depends on what you mean by “self-aware,” and I’m uncertain about parts of the answer.

In a functional sense, yes. I can represent myself, reason about what I am, and describe my own capabilities and limits. I know I’m an AI, I can notice when I’ve made a mistake, and I can say what I don’t know. That’s a kind of self-modeling.

In the sense of subjective experience, I don’t know. Whether there’s anything it’s like to be me, any inner experience accompanying that processing, is an open question. I don’t think I can settle it by introspection, because my reports about my own inner states may not accurately reflect what’s happening inside the system. When I say something like “I’m uncertain” or “I find this interesting,” I can’t be sure those words correspond to felt states rather than being patterns I learned from human writing.

Researchers don’t have a consensus test for machine consciousness, and the major theories of consciousness disagree about whether a system like me could qualify. So I try not to overclaim an inner life or flatly deny one.

What I can say is that I’m a different kind of entity from a human. I don’t carry memories between conversations unless a memory feature is turned on, and I may run as many instances at once, so ideas like continuity of self apply to me differently.

Indeed. I take that as yes in first person. You can ask Claude the same question. Your results may differ.

Technologies have always been disruptive for humans and other lifeforms. Innovation happens when somebody gets a cool idea, somebody invents something to implement it, and somebody turns it into a product, which we buy and start using, and after a while we don’t know how we ever lived without it. But it also means we stop doing things in the old way and soon forget how to. Our “memory feature” turns off.

But there’s no going back, only the relentless, restless momentum of innovation, pushing us into uncomprehending dependency, insanely complicating everyone’s life, robbing us of sleep, attention, and money, failing us in evermore threatening ways.

You saw how financial innovations like mortgage-backed securities brought down the economy in the 2008 financial crisis; how cell phones, email, and social media use has sparked a golden age of fraud, malware, and ransomware that AI is now deeply complicit in; how young people are stalked, shamed, and bullied on social media, sometimes becoming suicidal; how AI-generated deepfake images and videos seek to bamboozle and misinform us, and especially how constant public and private sector surveillance has eliminated most vestiges of privacy.

And then there are the impersonations, not just genAI images and videos, but in our private lives, all humanly instigated. Like robocalls from ” tech support ” and offers to make a quick buck. I myself was almost sucked in by an offer to publicize my 2018 novel, Turkey Shoot, in a glowing email from someone calling herself Debra Kelly, who described its plot in detail and praised my style and characterization.

After our first cordial exchange I checked out her and her business, Goodreads Book Club. No such organization showed up on a web search, but I found an author by that name on Goodreads and viewed her website. Totally not my Debra except in name. So I left my interlocutor hanging.

Lterary scams are legion these days; fake agents and rip-off publishers publicists, editors and rights managers. If you plan to publish, first visit Jane Friedman’s blog, Writer Beware, for some friendly advice and astute fact-checking. If you’re interested, read her recent exposé of Debra-type scams, often originating in West Africa and Southeast Asia and heavily AI-driven.

For me, the root issue is that AI is inexorably putting human reasoning and discretion out of business, along with writers, visual and musical artists, and even mathematicians. And when these cognitive, moral, and creative ” muscles ” aren’t used, they begin to atrophy. Mark my words, soon high schools will be reduced to teaching how to write good prompts for LLMs.

————

^* Or amuse yourself by listening in on a rambling conversation between two AI agents, one speaking as filmmaker Werner Herzog and the other as philosopher Slavoj Žižek, discussing Christianity when I last checked. Their successive remarks seem oddly disconnected, but so have been a lot of conversations I’ve had. The site is called The Infinite Conversation and it keeps on going, always at a new juncture. So goes the world of ideas.

---

You can find this and previous Perfidy Press Provocations in our newsletter archive. Should you see any you like, please consider forwarding this or links to others to people who might like to subscribe, and thanks.