iiinova — where your digital self keeps evolving

I could see people reasonably disagreeing re "are atoms conscious?" being "hard", because it could be answered with direct access to qualia. imho you could say that makes it "*a* hard problem" but we could 100% know the necessary conditions of qualia without understanding the mapping to them

100

i.e. understanding the steps from some processing or mechanism leading to some specific qualia seems very very difficult -- this seems to be partially because we don't have direct access to the details of any other thing's qualia than our own

100

and yes, this is a neutral milk hotel lyric

000

The question "what evolutionary functions could consciousness serve" is also -- in this sense -- not a "hard problem"

031

videogame history @vghistory.bsky.social · 05/10/2026

PSA: "hard problems" as they're meant by the source material look like "why are specific functions mapped to specific experiences and how?" and not "are humans and atoms conscious?" or "what areas of the brain might contribute to consciousness?"

241

Reposted by Adverb

vistapro, 3d scenery generator, magazine screenshot, amiga (1991) archive.org/details/cu-a...

low poly 3d greenery, water and hills/mountains. blue sky.

11880210

Reposted by Adverb

Hannah J @hannahjdotca.bsky.social · 04/10/2026

Got 30s to help me out while advancing science by answering a silly question? research.hannahj.ca As usual: quick, anonymous, and no sign-up required.

363

Not even a critique (everyone does it) but just imho a pretty good way to make analysis more legible & useful Esp. easy for people in industry to not do this because we'd rather not e.g. break confidentiality or w/e by being too specific but just choose a nearby domain.

020

Imho: every time you're having an abstracted discussion on new tech (openness vs control, acceleration vs pause, etc.), it's pretty important to ground it in at least one real case

120

The u.s. should take its top 20 companies & turn them into about 300 companies just as a starting point

150

Andres Guadamuz @technollama.bsky.social · 04/10/2026

Get ready for egocentric & computer-use data capture of every single thing every worker does everywhere

000

Reposted by Adverb

Wired Magazine. Fascinated by tomorrow, slightly disappointed it keeps arriving.

0111

Reposted by Adverb

Kuba Suder 🇵🇱🇺🇦 @mackuba.eu · 02/10/2026

Man, that 7.x branch is really buggy, isn't it…

1171

Adverb @adverb.bsky.social · 02/10/2026

which I like by the way

000

Adverb @adverb.bsky.social · 02/10/2026

there's a category of generative art that's conceptual and can be roughly described as: "xyz is making a system with a goal that xyz is confident is impossible but xyz has to check just in case"

140

Reposted by Adverb

Ryan Moulton @moultano.bsky.social · 29/09/2026

Yes but you can at least get more mainstream answers to things like "is the church turing thesis probably true" "what can you learn in principle from text" "are millennial prize math problems hard"

1993

Adverb @adverb.bsky.social · 26/09/2026

someone's gotta look into this smh

110

Adverb @adverb.bsky.social · 26/09/2026

How could this happen to me

110

Adverb @adverb.bsky.social · 26/09/2026

I'm literally just getting old oh my god

120

Adverb @adverb.bsky.social · 25/09/2026

TESCREAL is a shame partially because it (or at least its reading by most people employing it) flattens some massive, diverse work into whatever the weirdest assholes in the field were doing at some point in time

1362

Adverb @adverb.bsky.social · 24/09/2026

also: openai et al. don't monitor their CoT anyways or we wouldn't have ~the last three months' news cycles~ lmao

010

Adverb @adverb.bsky.social · 24/09/2026

why are people pretending as if CoT is reliably faithful anyways; we have an entire field of interpretability to rely on here instead

130

Adverb @adverb.bsky.social · 23/09/2026

i'm off that covid and flu vaccine cart

000

Adverb @adverb.bsky.social · 22/09/2026

so sad that i've made threads on twitter for years as if it were an literal archival site instead of just one long-form piece on the specific stances & neuroses that i'm 99% certain people are going to develop over this

010

Adverb @adverb.bsky.social · 22/09/2026

flipping over my calendar page & it's become LLM-veganism season once more

100

Adverb @adverb.bsky.social · 22/09/2026

the price we pay for knowing about anything or caring about anything online i guess

000

Reposted by Adverb

Giovanni Campagna @gcampax.com · 19/09/2026

General purpose classification of questionable quality! Like, that should be target #1 for people worried about training data induced bias

0211

Reposted by Adverb

Matt Hodges @matthodges.bsky.social · 19/09/2026

Somehow Jev has not yet triggered much discussion of cheap, low-latency, general-purpose classification as infrastructure for targeted oppression.

1117013

Iirc I tracked time / tokens but def makes sense to use path length as well

000

Oh yeah agree, I think it makes a ton of sense there!

020

x.com/advadnoun/st... I looked at wikirace for RL for LLMs at one point. If they benchmarked with it for the demo, they probably didn't train on it, but imho a reasonable general + verifiable task to give!

x.com

Adverb (@advadnoun) on X

Oh hey Jev did this!

110

Lots of places in data annotation where someone would plop in a logprob from Qwen/similar otherwise; cool to see & makes good sense for parallel decoding.

130

Adverb @adverb.bsky.social · 17/09/2026

physically suffering from "computer rig unplugged" syndrome

000

i never wanna read a "take" ever again

120

Tbf I think at a certain point rich people Just Do That e.g. Zuck is specifically not into the "AI God" in any typical sense but has had one

020

It's just that he thinks authoritarian governments will either be hurt by or orthogonal to gains; he isn't thanking the censors

000

yeah imho the open-source open-science reality of deepseek's, qwen's work moves the sentiment from what could easily be read as nationalist-v-nationalist-posting to being a real difference

130

i want to find the source for "wouldn't wish robotics on my worst enemy" so bad fwiw i do think it'd be great for more people to work on it though ig employment concerns are dramatically worse in some ways

010

one of these is really really hard tho

230

fwiw I think it's still less-bad compared to bluesky and has actual roi fwict -- could be why I see more of it there.

000

austin @aparker.io · 14/09/2026

yeah I see this for relatively-less in-the-news topics on tiktok every day -- it's visibly painful for people who like a subject to share news or fun tidbits about it

100

Reposted by Adverb

if you wonder 'why does it feel like all the expertise is draining out of the public sphere' then you kinda have to grapple with the fact that displaying expertise is often an all expenses paid trip to the omelas hole

729448