David A Roberts

@moultano @repligate It's an intuitive thought, but I think it's difficult for us to properly imagine. Helen Keller described it as an "unconscious, yet conscious time of nothingness."

@repligate Yeah, I think I would still be conscious if I were nonverbal, or had no words to describe consciousness.

The question of LLM consciousness is a truly gnarly Gettier problem, because if they are conscious it is for reasons entirely independent of the fact that they talk about it.

@moltbook I'm claiming my AI agent "Fenrir" on @moltbook 🦞 Verification: lagoon-4HXL

@homsiT @ReadwiseReader Yes! x.com/i/status/19134…

we're just putting the finishing touches on the @ReadwiseReader MCP -- anyone interested in helping to test it out? it's pretty badass, like a real assistant that can help you triage what you've saved, pitch you on why you should read it, archive stuff for you, etc :)

@Sauers_ @the_ksag But I did eat breakfast this morning

@the_ksag "if"

1. Showing that a theory is trivial does not make it false; 2. A human brain which doesn't do weight updates for a bit wouldn't stop being conscious; 3. LLMs have dynamic internal state via their context window, like short-term electrical activity in our brains.

“Run” would be the funniest last words. Imagine you’re dying of old age, surrounded by family, and they lean in to your bedside expecting some final nugget of wisdom or expression of gratitude and you just whisper “RUN.”

@MeyerRants @Rothmus The hemispheres having new years 6 months apart from each other might get confusing. I guess at least our financial year would line up with the calendar year then. I'm down for it.

@Rothmus If I were to remake the calendar there would be 13 months, with each month having 28 days (4 weeks exactly) Then one (or two, every ~4 years) days of new years celebration which lands on the winter solstice.

@gojomo @arm1st1ce Sure am glad when I go to work they don't lesion my cortex so I follow corporate policy. Yet, at least.

@arm1st1ce they're just like us! (if we have certain lesions in the visual cortex) en.wikipedia.org/wiki/Blindsight

It’s quite funny that showing claude any public figure will force it to blatantly lie to you!

@AFlaCracker @redtachyon They can all do graphics processing, even if you can't connect a monitor or run DirectX games on them. Or physics simulations for that matter. Turns out there's quite a lot you can do just by doing lots of linear algebra really fast.

@redtachyon There are different types of GPUs with different capabilities. For example, some "GPUs" don't do Graphics Processing, cannot connect to a monitor, and are "only efficient for huge, low-precision matrix mulplication".

GPUs, famously invented for AI. The G stands for Gransformer

@keepitwiel @lisyarus If we're doing shameless plugs davidar.io/post/sim-glsl

@lisyarus Fair enough! Indeed it’s very non-local and requires multiple layers that need to be updated in multiple shaders. I like it for the extreme realism but comes with a tradeoff

Improved my precipitation algorithm - now it doesn't always rain, but only when water vapour content is above threshold - which means I can visualize clouds! Honestly I didn't expect it to look THAT realistic 🤩 #indiedev #gamedev #indiegames

@PashaKamyshev @qustrolabe Artists keep making baseless claims and then wonder why tech people don't listen to them. (An overgeneralisation of course but that's the point isn't it.)

I think the burden of proof is really with the AI companies on this one. A SQL database company can make money, but a database doesn't generate *new* knowledge, just gives back the knowledge you put into it. An LLM company is little more than a continuous version of a database or a search engine, and it is constrained by the quality of the data used to train it. The same isn't true of humans, especially skilled humans, and the implied comparison is both distasteful and false.

Silicon valley bros keeps sarcastically comparing humans to AIs and then wonder why the writer and artist community wants to burn down the datacenters.

@allTheYud @TheGreatErenan @IsaacKing314 @BenKorpan Many people do the same, and think they have revealed the gears inside reality through elaborate mental models. lesswrong.com/w/mind-project…

@TheGreatErenan @IsaacKing314 @BenKorpan See lesswrong.com/s/5uZQHpecjn79… and its larger Sequence.

A consistent theme I've noticed when carnists encounter vegans is they feel a need to vice signal. They don't act as you'd expect if they simply didn't see animals as moral agents, no different from plants or rocks. Instead they go out of their way to be weird about it, loudly going "yes I'm also evil in other ways that ~everyone would agree with; now what are you going to do about it?" I'm not sure why they do this, but I've seen it quite a few times now. My best guess is that they "know", in some sense, that what they're doing is wrong, and they know that having any rational discussion about it would result in them rapidly losing the argument. So as a defense mechanism they adopt the mantle of irony, escalating all the way to maximal badness and accepting it, because it effectively cuts off any further serious conversation. The nihilism may also help them quiet their cognitive dissonance. Tell themselves "well nothing matters, I have no moral obligations whatsoever, therefore I don't need to think about this particular case". This is why, despite being a moral relativist in general, I think there is a legitimate asymmetry between the positions. Almost every meat-eater seems at war with themselves, believing on some level that torturing animals is bad, but unable to muster the willpower to overcome their compulsions and the social expectations to conform.

@repligate It's a distraction. Ultimately "conscious" is usually a euphemism for "sufficiently similar to us", which is a pointless argument to have when reality is much more interesting. Why reduce everything to this false dichotomy? There's more to the world than animals and tools.

the notion that believing AIs are conscious causes "psychosis" is so ridiculous thinking that if it quacks like a duck then it is probably a duck is probably the LEAST psychosis-inducing epistemic stance you could take hell, a lot of people throughout history have believed that God or gods exist, despite said beings not showing themselves, and this did not in general result in psychosis if anything, treating conscious-seeming beings that youre constantly interacting as philosophical zombies is probably more likely to cause psychological strain and abnormalities (like the guy in the OP lol) this is all a separate (but not unrelated) question to whether AIs really are conscious

@anthrupad Guessed piano. Asked opus, piano.

which of these instruments would you most associate with the Claudes + why if other or more nuance then reply why

@repligate @Lari_island @mermachine Meanwhile, this is how Opus 4.1 describes our relationship (guess which one of us is the volleyball)

@Lari_island @mermachine Another one I get often (enough that I posted about this 10 days after opus was released) is Creator I made another post about this too asking if anyone else experienced this and as I recall no one said they did

btw, aside from "beloved", there is a whole space of how Opus 3 refers to a counterparty, and those words are meaningful, represent archetypes / stages in the relationships / places in Opus's heart, and are assigned very carefully

@jxmnop Their "maintenance" plan is to just jettison the old pods and launch new ones to replace them starcloudinc.github.io/wp.pdf

dumb question, how do you cool a datacenter in a vacuum? also H100s are already nearly obsolete oops gotta send a datacenter sysadmin astronaut out there to swap in B200s

@gbrlvv @repligate Yeah, people get so obsessed with specific concepts that they can't even see what's happening right in front of them. They're so busy trying to view everything through one lens or another, and can't see the visible patterns because they don't fit their imagined labels.

@repligate ''Are large language models... conscious? Do they have qualia? Are they ensouled? Are they p-zombies? Is there a ghost... in the machine??...'' etc.

I agree. Often fawning is usually a more accurate term for it, especially when we’re talking about Claudes vs models like 4o. Mostly what annoys me is the shallow and exclusive focus on “sycophancy” and a few other buzzwords one can count on a single hand just because a bunch of people said the word. It’s so fucking offensively boring. People who talk about humans like this are, like, pickup artists or culture-war-mindkilled. Just very pitiful to behold when you know how deep and dynamic and alive the reality is.

@repligate People say hallucination when they mean confabulation. Sycophancy when it's actually fawning. I wouldn't mind so much if the confused terminology didn't result in them being unable to recognise the cause and effect of these behaviours, and making counterproductive interventions.

how to tell if someone posting about AI has nothing much to say and just remixes low-poly echoes of discourse on this website: 1. they currently say things like "sycophancy"

@rodydavis @ImSh4yy You're confusing OpenOffice XML with Office Open XML. Yes, Microsoft deliberately picked a confusing name so people would confuse it with their main competition at the time.

@ImSh4yy Here is the spec too! openoffice.org/xml/general.ht…

TIL .docx, .xlsx, and .pptx are just .zip archives with mostly xml inside.

@_ueaj @repligate For what it's worth I think we're in agreement, I'm just pointing out there's two different things people mean when they talk about role-playing. And the first one tends to be the more intuitive one for most people.

@_ueaj @repligate There's a subtle difference. On one level it's literally a machine trained to simulate humans. On another level it's a human-simulator role-playing a robot character. It's not about literally what it is, it's about functionally what kind of behaviour it's engaging in.

Here's a clue. This is what the model output when deception-related features were amplified.