iiinova — where your digital self keeps evolving

Elon musk belongs on trial at The Hague.

Reposted by Colin

Over a year ago I met an Anthropic employee at a conference and at one point she mentioned that one of the things she was working on was whether or not it was cruel to ever turn off an AI model. I thought she was joking. She was not.

561021181

Reposted by Colin

Hypervisible @hypervisible.blacksky.app · 08/10/2026

Coming on the heels of the reporting about lobbying the pope to say models could be conscious, it’s clear they are determined to impose their cultish worldview on everyone.

1323154

Colin @colinniloc.bsky.social · 08/10/2026

I mean, I guess it could still be a bit, and if so, bravo!

000

Colin @colinniloc.bsky.social · 08/10/2026

When I saw this I thought that surely this must be some kind of joke or mockery of AI techbros. But nope:

Alvin Sng
@alvinsng
My wife: Can you read our baby a book?
Me: I can just play an audiobook from my phone.
My wife: But then it won't be in your voice.
Me: I can train a voice model.
Parenting hits different when you work in Al.
Oct 7, 2026 • 12:30 AM UTC

Anirudh Sharma
@anirudhology
Replying to @alvinsng
Shouldn't your first instinct be to read your baby a book willingly as a father and to spend more time with your baby.
If it wasn't, I guess it's time to think deeply on what you will lose if you trade your emotions for technology.
Just a friendly advice.

Alvin Sng
@alvinsng
I do want to spend time with my baby. There's even a chance I'll choose to retire early and spend all my time with my kids!
I'd just rather spend that time cuddling, feeding, and playing, and automate the things I don't need to be around for.

141

Colin @colinniloc.bsky.social · 06/10/2026

This is only worth doing if you can say "COME" and it lets them in

010

Reposted by Colin

tante @tante.cc · 03/10/2026

So the "I left $AILab for ethical reasons" grift is starting up: I understand wanting that bag, wanting to be in on a big thing. But if you thought OpenAI or Anthropic were okay a year ago or so, I really am not willing to give you any credit for leaving now.

theatlantic.com

I Quit OpenAI Because Its Culture Is Broken - The Atlantic

The industry’s approach to safety will guarantee more failures unless something changes.

1225654

Colin @colinniloc.bsky.social · 01/10/2026

But regardless, you've mentioned that you're "agnostic" about a hundred times in these threads, without ever clearly stating what you believe, which I find exhausting. Also you've found your way onto a mute list that I follow so it's hard for me to follow. Thanks for the conversation.

100

Colin @colinniloc.bsky.social · 01/10/2026

I am using agency in the colloquial sense of the word here, feel free to substitute “sentience” if you wish.

200

Colin @colinniloc.bsky.social · 01/10/2026

My position is that this is not learning, and the AI companies branding the use of context as "learning" is an attempt to convince users that the LLM has agency when it has none.

100

Colin @colinniloc.bsky.social · 01/10/2026

A harness could run an LLM on a repeated basis, but it's still just running one prompt at a time, just perhaps more frequently than you would run it if prompting it manually. So it's still not continuous.

000

Colin @colinniloc.bsky.social · 01/10/2026

There is no way that I'm aware of which would allow the current crop of LLMs to change their weights in the way you suppose, but even if they could, there are other properties (mentioned in this thread and elsewhere) that would likely mean that this hypothetical "dynamic" LLM couldn't be sentient.

000

Colin @colinniloc.bsky.social · 01/10/2026

People with memory loss (short or long term) still have the capability to change. LLMs have no capability to change, so I don't think this is a valid analogy.

100

Colin @colinniloc.bsky.social · 01/10/2026

I also now realize that @wescontreras.com mentioned the lack of persistence in a message way up the thread, which it doesn't look like you substantively responded to, so I'd be interested in your thoughts on whether something without persistence can be sentient.

000

Colin @colinniloc.bsky.social · 01/10/2026

As I said, I've only thought about this briefly; feel free to poke holes in it!

200

Colin @colinniloc.bsky.social · 01/10/2026

C. LLMs cannot be sentient.

100

Colin @colinniloc.bsky.social · 01/10/2026

P4. LLMs are static and unchanging; once a given model is trained, it is fixed.

100

Colin @colinniloc.bsky.social · 01/10/2026

P3. LLMs do not exist continuously; they “exist” in any sense of the word only for a short amount of time when responding to each prompt.

100

Colin @colinniloc.bsky.social · 01/10/2026

P2. If an entity is sentient, it is necessary (but not sufficient) for it to have a mechanism for changing over time, so that it can remember the experience it just had, and think/respond/feel something about it.

100

Colin @colinniloc.bsky.social · 01/10/2026

Here's a crude propositional argument which I've literally only thought about for about 10 minutes: P1. If an entity is sentient, it is necessary (but of course, not sufficient) for it to exist continuously, so that it can, for example, experience something lasting a non-trivial amount of time.

100

Colin @colinniloc.bsky.social · 01/10/2026

Okay. My position is that we know empirical, verifiable things about LLMs, such that sentience (defined in however you would define human sentience) is not a possible property for them to have.

100

Colin @colinniloc.bsky.social · 01/10/2026

None of this is an argument for why the studies I mentioned are not valid evidence. Or do you think no scientific study provides evidence? Do you believe in germ theory? Gravity? The theory of the atom?

100

Colin @colinniloc.bsky.social · 01/10/2026

Any of the literally thousands of studies that has been done on the topic by neuroscientists and other biologists, psychologists, medical doctors, etc. Feel free to search if you’d like specific ones.

100

Colin @colinniloc.bsky.social · 01/10/2026

No, a Turing test can say nothing about sentience because that’s not what it was designed to measure. And again, I think you’re misusing the word “evidence” here. There is plenty of evidence for sentience among humans and other animals as well.

120

Colin @colinniloc.bsky.social · 01/10/2026

I don’t think any of this answers the question I asked, so I’ll bow out. Take care.

100

Colin @colinniloc.bsky.social · 01/10/2026

Turing tests are absolutely 100% not an indicator of sentience, and there were systems literally decades ago IIRC that were able to pass Turing tests.

200

Colin @colinniloc.bsky.social · 01/10/2026

If not, I think the only conclusion that’s rational is that it’s much more likely than not that LLMs can’t be sentient.

010

Colin @colinniloc.bsky.social · 01/10/2026

Likewise, in his original posts, Thomas provided evidence (which all looks accurate to me, someone with a CS degree) that LLMs don’t have the ability to do the things required of a sentient being (know, reason, experience, think, etc). Do you disagree with any of the statement he made?

210

Colin @colinniloc.bsky.social · 01/10/2026

I don’t think your “trump card” is even an accurate statement. You have evidence of my sentience, and I of yours, because we have the ability to converse and appear to be sentient beings to one another. Yes, we can’t know with certainty, but there is plenty of *evidence*.

200

Colin @colinniloc.bsky.social · 01/10/2026

LLMs don’t even exist continuously! They do literally nothing until you prompt them, at which point they produce (deterministically) a probable response to your prompt. Then they stop again.

010

Colin @colinniloc.bsky.social · 01/10/2026

It seems that in every discussion about how deterministic computer algorithms can’t be sentient, someone inevitably pivots to the philosophical problem of hard solipsism. The LLM can’t be conscious because it doesn’t think, or reason, or know, or observe. There’s no mind. Solipsism is irrelevant.

110

Reposted by Colin

Brett Karlan @oldjerryfodor.bsky.social · 01/10/2026

"I ran this piece of writing through an AI detector" oh cool I had a psychic tell me whether this was written by AI. We both have the same level of evidence now

3428

Colin @colinniloc.bsky.social · 30/09/2026

“This [unwanted/illegal action taken by an LLM] is not sentience by computers; it is negligence by humans” is the best single-sentence way to explain this concept that I’ve seen thus far.

seattletimes.com

WA needs to do these 3 things to protect us from AI harm | Op-Ed

Here's what Washington can do to protect us from irresponsible companies pushing as fast as they can, releasing unsafe products and expecting no consequences.

011

Reposted by Colin

Alexander Schmidt-Lebuhn @anschmidtlebuhn.bsky.social · 29/09/2026

It is really underappreciated to what degree the rogue agents discourse is about prefabricating an excuse for releasing predictably unreliable and unsafe products. We don't need "AI alignment", we need enforcement of consumer protection and felony hacking laws.

04513

Colin @colinniloc.bsky.social · 29/09/2026

This is not “deportation”, it’s human trafficking, and shame on you for not calling that out.

010

Reposted by Colin

Lia @liasae.bsky.social · 24/09/2026

I am now keeping this on speed-dial

Edited old IBM printout in black caps on off-white paper.
'A computer can never be held accountable
Therefore a computer must always do the hacking haha what are they gonna do, put a computer in jail? Haha "oops it's alive" lmao'

81102321

Colin @colinniloc.bsky.social · 26/09/2026

This has to be ragebait, right? Like, no one genuinely thinks like this??

020

Reposted by Colin

Ed Zitron @edzitron.com · 24/09/2026

I think that he should've had Altman arrested for running a company committing cybercrimes

261058182

Colin @colinniloc.bsky.social · 23/09/2026

This is not what “deported” means and using that term in this context furthers the fascist project. Do better, AFP.

011

Reposted by Colin

cait (and adonis) @cait.bsky.social · 22/09/2026

this stupid ass motherfucker simply does not have that dog in him

321469141

Colin @colinniloc.bsky.social · 20/09/2026

The “state of mind” part is harder to prove, and ends up drifting more into philosophical concepts like solipsism and the hard problem of consciousness.

010

Colin @colinniloc.bsky.social · 20/09/2026

There are plenty of scientific definitions of consciousness, most of which involve awareness of one’s state of mind and external environment and being able to reason about both. There are plenty of ways to prove the “external environment” part of that.

221

Reposted by Colin

Amyville Hoyrror @amyhoy.bsky.social · 20/09/2026

i think it’s incredible how people claim that if you say “llms don’t think” you are a “physicalist” or, even worse, a secret dualist it’s a basic manipulative trick, you know extraordinary claims require extraordinary evidence and they don’t have any there’s NO evidence for llms being conscious

611627

Reposted by Colin

Annalee @flowerhorne.com · 19/09/2026

This is why I think it's so important to push back against the personification of LLMs. If a person shoots someone while wearing a hand puppet, they can't claim the hand puppet did the murder.

21653501