Roly Perera
@dynamicaspects.org
220 followers 470 following 100 posts
Making software more open, explorable and self-explanatory https://dynamicaspects.org/research Also: subjectivity, agency, active inference, niche construction, enactivism @plrg-bristol.bsky.social @iccscambridge.bsky.social @f.luid.org
Posts Media Videos Starter Packs
Pinned
dynamicaspects.org
Language-Based Dependency-Tracking for Explorables, presented at LIVE 2025. Joint work with Joe Bond, Cristina David, @mcnuttandrew.bsky.social and Alfonso Piscitelli. www.youtube.com/watch?v=gds8...
LIVE 2025 - Language-Based Dependency-Tracking for Explorables
YouTube video by The LIVE Workshop
www.youtube.com
dynamicaspects.org
Wow. Looking forward!
Reposted by Roly Perera
dorchard.bsky.social
PROPL '25: Proceedings of the 2nd ACM SIGPLAN International Workshop on Programming for the Planet. Now online: dl.acm.org/doi/proceedi...
Looking forward to the workshop next week @icfp-conference.bsky.social/SPLASH
Proceedings of the 2nd ACM SIGPLAN International Workshop on Programming for the Planet | ACM Conferences
dl.acm.org
dynamicaspects.org
Language-Based Dependency-Tracking for Explorables, presented at LIVE 2025. Joint work with Joe Bond, Cristina David, @mcnuttandrew.bsky.social and Alfonso Piscitelli. www.youtube.com/watch?v=gds8...
LIVE 2025 - Language-Based Dependency-Tracking for Explorables
YouTube video by The LIVE Workshop
www.youtube.com
dynamicaspects.org
But evolution has already demonstrated it. It built a conscious machine without relying on any resources (other than a lot of time) that an engineer wouldn't also have access to.
dynamicaspects.org
Consciousness isn't something special happening *to* us, it's something special that we can *do*, and so any account has to be functional in nature and any role for the "substrate" can only be in delivering functionality. Difficult to make progress without agreeing on the explandum first.
dynamicaspects.org
The problem is that these stories are all a bit mysterian about consciousness itself, i.e. as an explanatory target it is itself mysterious and so it remains an open question whether there is some (correspondingly mysterious) role for biological "substrates" to play.
dynamicaspects.org
Evolving a machine (meat or otherwise) is just a one way of engineering one, so conscious organisms provide an existence proof that consciousness can be engineered.
dynamicaspects.org
“Biological realisers” are also computational mechanisms, though, so not sure I understand your premise.

And a conscious machine is “just” a machine that thinks it's conscious -- perhaps not something easy to engineer but certainly not a notion that depends on any particular substrate.
dynamicaspects.org
Realising that “as Immanuel Kant would say” is quite triggering for me
dynamicaspects.org
The bits i managed to catch of LIVE 2025 were great - an impressive and subversive set of submissions! Looking forward to more LIVE (recorded earlier) antics next year 🌟🌟
Reposted by Roly Perera
joshuahhh.com
Less than 24 hours till the LIVE 2025 Q & A / Discussion session!

Watch the presentations: www.youtube.com/playlist?lis...

Join us to chat about them: links at liveprog.org

Hope to see ya there!
dynamicaspects.org
Companion (Workshop) Proceedings for ‹Programming› 2025 are now available! Co-edited with @jonathoda.bsky.social and @tomasp.net. Thanks to the awesome OASIcs publishing team for getting this over the line.
drops.dagstuhl.de/entities/vol...
Companion Proceedings of the 9th International Conference on the Art, Science, and Engineering of Programming (Programming 2025)
drops.dagstuhl.de
dynamicaspects.org
pseudonymous double-barrelled dipshit
dynamicaspects.org
A person is vastly more capable in this regard -- almost every kind of behavioural “prompt” will reveal rich, broadly consistent information about a subjective life.
dynamicaspects.org
There's no reason at all for thinking Gemini or similar systems have a self-model. We can prompt them to generate plausible-sounding claims about being conscious, but we can also prompt them to spout gibberish, or to deny that an AI could ever be conscious.
dynamicaspects.org
Then as far as you are concerned (you being the subject matter of the self-model), it is information about *your* first-personal vantage point — how things look to you. I think of it as the body schema enriched to a full agent-in-the-world schema.
dynamicaspects.org
Ok, nice. It's not enough just to have access to perspectival information about appearances. You also need a self-model that takes that information about a first-personal vantage point and interprets it as yours.
dynamicaspects.org
(And if you have the patience to sit through a 20-min video, I have a talk on this very topic here: youtu.be/WtWfmBujxxU?...)
dynamicaspects.org
6/ In fact it's other way around: the images are the content of self-interpretative acts by which the agent comes to a rich appreciation of itself *as* an agent with a perceptual vantage point.
dynamicaspects.org
5/ These are internal subpersonal actions that serve to build up a self-narrative whose subject matter is “how the world looks”. A conscious agent thus delivers phenomenal content primarily to itself. This doesn't threaten a Rylean regress because we don't use these images to explain perception.
dynamicaspects.org
4/ The other crucial step to flesh out the story a bit is recognising that the organism is able to (and routinely does) deliver that imagistic information *to itself* (not, say, only when pressed for a verbal report). This is an example of what @metzinger.bsky.social calls “epistemic agency”.
dynamicaspects.org
3/ The vehicle is the organism, which can deliver all kinds of fine-grained imagistic information about appearances: for example how apparent shape transforms in response to movement. The totality of this perspectival subpersonal knowledge *is* the content of visual consciousness, on this view.
dynamicaspects.org
2/ So the usual enactivist adage that “the world is its own representation” is tempting but too quick. Vehicle vs. content is one of the crucial ingredients need to resolve this conundrum. Then the claim is that the *content* is a dynamic egocentric image, not the vehicle.