armutie

Why is everyone so certain that LLMs are NOT conscious? Every argument I've read boils down to 'its just a programmed machine'. Why does that prove it's not conscious? There is no test you can conduct, on human or animal or machine, that they are or aren't conscious.

@bobthebobnes @DaveShapi Believe what you want but ‘because no’ just isn’t a good argument.

@DaveShapi Because it doesn't. And if you have a brain and any concept to how they (ai) work, you would not even remotely consider them to have consciousness. In fact, it's so absurd that it's guaranteed that anyone who says otherwise is engagement farming or just has no clue

An interesting question would be why do people think it doesn't?

@HouMuza @FredLambert I guess this is the chart you're looking for, then. Put another way: Grok costs $7.45 per AA point, and MiMo $8.55 per point. To me, Grok has essentially the same score (53 vs 54) while being 15% cheaper.

@ar_mutie @FredLambert Per token cost plus performance metrics would be more interesting to me because GPT 3.5 right now would cost less to run in total but would be useless.

xAI's new Grok scores worse than Xiaomi's Mimo model. Elon Musk just forced SpaceX to buy xAI for $250 billion. Xiaomi is valued at $100 billion, and it's not just an AI company; it is also a giant personal electronics brand and the fastest-growing electric automaker. Musk's real superpower is overinflating valuations.

@FredLambert The truth is more nuanced. There is a chart on AA where they measure cost to run benchmarks. As shown, 4.3 costs $395, where Mimo is $462. Even when Mimo on paper has similar per-token cost, it uses many more tokens when actually running. I hope this is definitive proof for you.

@ar_mutie It would be a good point if Grok were more efficient to run, but that's not even the case, so what's the point?

@sagewilyam @FredLambert You have a good point, but what I’m trying to say is that these ‘moral issues’ and him not actually wanting to ‘make things right’ cannot and will not be judged in court. You are free to make a judgment of your own making, though, that is true

@ar_mutie @FredLambert We are the public, yes we can judge off of anything we want.

Musk is obviously right here. But even then, it's hard to side with him because of the reasons he is doing it. Does anyone really believe that Elon has moral issues with OpenAI going from non-profit to for-profit? Of course not. There are plenty of evidence that Elon was himself pushing for the change to for-profit. The disagreement was only on control. Musk wanted himself to control the for-profit and the others weren't into it... for reasons that are now obvious. So Elon doing this right now has nothing to do with "making things right", but everything to do with trying to hurt a competitor as he desperately tries to gain control over AI.

@Stormy_Skies_36 @skyesmunroe Agreed, some criticism is still important. Let’s see what the end of season is like, I’m hoping it allows the show to transform into something more slow and brooding over action-action-action hype aura moments. I think they’re aware of this

@ar_mutie @skyesmunroe Ok that’s an excuse for last season and current season, not further seasons tho

do you guys ever think about how we had almost of an entire episode of matt losing his hearing in the og show and we’re never gonna get anything like that again cos we quite literally don’t have the time

What I’m seeing with #Daredevil discourse is people who fixate on issues that leaves them unable to enjoy the show. Yes, there are obviously issues, like with Matt, but this show is still so much fun as a whole. I’m thinking back to pre-overhaul drama and like, imagine that mess

@RomainHedouin You must take into account human reaction time WHILE DRIVING. Sure, on tests, human reaction time can be as low as ~150 ms, but this is frankly impossible in a driving situation with 15 cars around, focusing on pedestrians, focusing on the traffic light, etc

@T10AlarmClock Hopefully tonight's episode will be a bit better for you! I have a feeling the whole Karen/Bullseye thing will be a big deal.

@ar_mutie i think my issues lay both with Matt stretching his arm and Bullseye accepting it, and then also the very relaxed and smiling way Matt presents him to Karen. would’ve preferred a tiny bit more angst in those moments.

that shot of DD stretching out his arm to Bullseye, that doesn’t really make sense. why would Bullseye take his hand if he doesn’t want to be saved? and just cause DD wants to save him doesn’t mean he wants to reach out like that… feels like a shot designed just to look cool.

but it’s so so basic. If we follow this further, maybe the atoms themselves have this mini-experience that isn’t really useful, but is there. Augmented by memory and whatnot, it becomes us. So maybe underneath all of physics, *consciousness* is the real fundamental thing.

Think of a person with dementia or like even 7 second memory. We would agree that he is conscious. But it would be more primitive, would it not? Scale this down further; think of a person with 1 millisecond memory, is this person even conscious? Yes they might have experience,

This is random but the whole consciousness thing is really getting to me. I am starting to feel that it is inherent rather than emergent. Like, it makes no sense why we feel like we are ‘conscious’. What physics explains this? Nothing.

If you haven’t watched Daredevil, bless your eyes by binging through it right now

If you are someone who thinks LLMs are slop and that AI is a bubble, please try out Claude Code and Codex. It is simply impossible to end with that conclusion after knowing their capabilities.

@lefthanddraft @tenobrus Following from here, if we have no plausible explanation for what bridges physical matter to go from non-conscious to conscious, wouldn’t it be valid to say panpsychism has a hold here, and that *everything* is conscious, and that rocks have some primitive form of experience?

I'm not asserting that experience is a non-physical "thing". But if some physical things lead to experience and others don't, there needs to be a bridging principle to get you from descriptions of function and structure of physical things to experience. Physics and chemistry do not provide that bridging principle. Your solution is to claim there is an identity, without seeking any underlying explanation. By contrast: - Claude can understand that running code on GPUs will lead to inference simply by applying general principles of computer science. - It is also possible in principle to use the general principles of physics and chemistry to derive the properties of water from H20 and determine nothing else will have those exact properties. Even though there is identity, there is also an explanation and the identity is derivable from general rules. We don't stop demanding explanations about how the microstructure of H20 leads to macro-structure of water just because they are identical. A brute identity for consciousness is not solving the explanatory gap. It is avoiding it by stipulation.