Notices by FeralRobots (feralrobots@mastodon.social), page 4

"system that think like your brain in parallel" no my dude, you haven't, you've done absolutely nothing to demonstrate that transformers think in any meaningful way "like my brain."

@LinuxAndYarn @thomasfuchs
arguably a way glib to put it; however, it does get at a problem with this type of treatment. What I've read is that psychedelics as treatment for depression, anxiety, PTSD, et al, is currently believed to work by increasing neuroplasticity.

What do you then *do* with that?

If you just take the drug & live your normal life, you should expect any beneficial results to be impermanent. Or if you continue to do bad shit, expect new, undesirable patterns to be enforced.

@cstross @pettter
Sure, but I think we're mostly shitposting on this. (FWIW, though, I think there's an argument for 'bullshit' vs 'lying.' It's more dismissive - I'd map 'bullshit' to contempt, & 'lying' to anger. Contempt is often more damaging than anger.)

@pettter
I'd argue for #bullshit in the Frankfurtian sense of it not caring about truth, just about outcomes - but it doesn't actually care about *anything* because, yeah, it's not capable of caring.
@cstross

@Gargron @anildash
I'd argue it's usually not, since I mostly see it used ironically ("17 civilians were discovered to have been rendered unalive") or as a fine distinction ("there's something unalive about that painting"). I.e., not bowdlerization - it has a different purpose than that.

What's going on is that Anthropic "prompt engineers" have redefined self-awareness to mean 'has contextual information.' That the system is using language then allows them to delude themselves into universalizing their definition.

Saw a similar problem in AI research in the 80s: researchers might define a "frame" holding contextual info, & when their program produced solutions that referenced the frame, construed that as a form of self-awareness.
#AIHype #Claude

https://arstechnica.com/information-technology/2024/03/claude-3-seems-to-detect-when-it-is-being-tested-sparking-ai-buzz-online/

Put another way: We should not lose sight of the fact that LLMs are doing some really interesting things. But that they're being built by cultist #AITrueBelievers, to do this thing using natural language, while simultaneously making lots of money*, all contribute to the delusion of something being there that isn't.
_
*which is the primary signifier of God's Grace in Calvinist Capitalism

What's fascinating to me is Alex Albert losing sight of something genuinely cool & interesting: the model integrated needle testing concepts so quickly that it produced responses that could be construed as recognizing the test environment.

Illusion of "meta-cognition" isn't that surprising if one remembers the system is created & trained by #AI #TrueBelievers who spend all day every day communicating in language that presumes #AGI is imminent - if not, as assumed here, immanent.
#AIHype #Claude

Put another way: Alex is basically telling Claude 3 ("Opus") that he's running a test on it, & is excited when Claude (a system for analyzing & producing human-plausible representations of similar text) "recognizes" a needle-testing prompt and produces text that's plausibly consistent with needle-testing.

What one SHOULD then do is remake one's tests (or better-sandbox the model). Instead, Alex leaps to concluding the model is self-aware.
https://twitter.com/alexalbert__/status/1764722513014329620

@nemeciii
There's just no way that voice control as a primary method of control holds up to ordinary day to day use across the population.
If you want to suggest that disabled folks should get knobs, then think about transitory disabilities: illness, fatigue, injury. Think about circumstances: passengers, loud music, software failures.
So, no. Hard pass.
@msh @thomasfuchs

@nemeciii
OK I think I didn't understand your claim before, but now that I do: No. Absolutely not.
Ambient noise exists. People lose hearing. People aren't always able to speak. People carry on conversations in cars. Disability exists.
You can say machine learning systems will handle a lot of these use cases, & maybe they can - sometimes. But what should our tolerance for failure be in a one or two ton vehicle moving at high speeds in varied conditions?
@msh @thomasfuchs

@nemeciii
So what I'm hearing you say is that your use case is very different from that of most other people.
@msh @thomasfuchs

@nemeciii @msh @thomasfuchs
I didn't say anything about 'future'.

@emilygorcenski
maybe you saw it coming? I think a lot of people remember the pandemic starting at whatever point their relevent sociopolitical organizing unit went into lockdown; for me (western NYS), that was approximately March 2020, IIRC. So not terribly far ahead of 'mid-2020'.

@thomasfuchs
scrollbars are just generally a hot mess, at least on Windows. Someone at Google decided about 15 years ago that nobody really needed to see or click-drag scrollbars & they've been shit ever since.

@thomasfuchs wait does that have a three-dial combination lock? So cool.

@thomasfuchs
Oh man I literally keep forgetting (because I never use my bsky account) that bsky doesn't even pretend to do DMs. (Or maybe they pretend now.)

It does bemuse me to almost no end that people still believe DMs exist on Fedi. If I can ever be arsed to up my gif game I'm gonna make one of Morpheus in the dojo that says "You think that's a DM you're reading?"

@AT1ST @grumpasaurus @molly0xfff

@goatsarah fair, fair, just that I seemed to recall a photo taken up in a high rope obstacle course - & IIRC one from atop a mast, though TBF the boat seemed to be in port at the time.

@goatsarah
If YOU say that...😉

@lrhodes
TBC those are all opt-out mechanisms.
@snarfed.org @activitypubblueskybridge @fedidevs @fediversenews