AGI Series 2024 - Joscha Bach: Is Consciousness a Missing Link to AGI?

Joscha Bach is an eminent German artificial intelligence researcher and cognitive scientist, focusing on cognitive architectures, mental representations, emotion, social modelling and multi-agent systems. He takes the view that artificial intelligence offers us a conceptual framework for how to understand the mind. He is the author of Principles of Synthetic Intelligence, and has argued very provocatively that we live our lives through a form of simulation.
This series is devoted to the theme of Artificial General Intelligence (AGI). According to OpenAI's definition, AGI amounts to "highly autonomous systems that outperform humans at most economically valuable work". Are we there yet? If not, how far are we away? What would it take to reach AGI? And what would happen if we reach it?
ARTIFICIAL GENERAL INTELLIGENCE
Sun 8 September [12 noon EST start]
JOSCHA BACH
IS CONSCIOUSNESS A MISSING LINK TO AGI?
Sun 15 September [10.00 am EST]
STUART RUSSELL (UC BERKELEY)
AGI - WHAT IF WE SUCCEED?
Sun 22 September [10.00 am EST]
NEIL LEACH (FIU)
ALIEN INTELLIGENCE: AN INTRODUCTION TO AGI
Sun 29 September [10.00am EST]
AGI FOR DESIGNERS
Neil Leach, Karla Sandana Ochoa, Daniel Bolojan, Philippe Morel, Michael Just,
Smaro Katsangelou
Sun 6 October [11.30 am EST]
BLAISE AGUERA Y ARCAS (Google)
AGI - ARE WE THERE YET?
Sun 13 October [10.00am EST]
KARL FRISTON (UCL)
THE PHYSICS OF SENTIENCE

[@JustSomeElectricalEngineer]

Watching most Videos, Lectures & Podcasts in 2x. Cannot do Josha on 1x. So much to unpack. 🤯

[@Tarzan_of_the_Ocean]

excellent talk Joscha!

[@Sasha_sementsev]

Many thanks 🙏

[@vikrantvijit1436]

Thanks for the absolutely delightful talk on missing links of consciousness that connects the past, present and digital futures of an agentic environment with artificial general intelligence collective forms adjacency dynamics. 😊❤

[@williamjmccartan8879]

What a great presentation and follow-up dialog, thank you everyone for sharing your time and work, peace

[@JoePUNK_Was_Here]

I needed this.

[@rckindkitty]

Thank you all, especially Joscha.

[@DouwedeJong]

at 40:54 Joscha misspoke. Daniel Dennett did not come up with the concept of the Cartesian Theatre; rather, he coined the term as a critique of a particular way of thinking about consciousness, often associated with Descartes dualism. Dennett promoted the idea of the Multiple Drafts Model, which is very similar to the orchestration model Joscha talks about in the slides prior to this point.

[@richardnunziata3221]

So if I am understanding this Joshcha is saying consciousness if the function of finding invariances between other functions passing them to direct attention feedback? Kind of like a topos bridge generator, But would there not be a reward mechanism somewhere over the invariances.

[@luitek]

Fascinating, as always. Joscha is also my favorite thinker of late. On a very tiny side note: I found the illustration of Aristotle quite compelling and artistic. I suspect it is AI generated. I did notice however, that the hand gesture, as it curls up against his chin is a "right" hand on a "left" arm. Check around minute 20:00 and you'll see what I mean :-)

[@Cats-a-Tonic]

Cats-a-Tonic ❤s Joscha!

[@Myironrodleftindia]

I also started thinking of Coraline. The other mother…

[@markomilenkovic2714]

Starts at 7:55

[@richardg.lanzara3732]

Very good discussion! My own research suggests that a more complete understanding of how our receptors function will lead to a better understanding of our consciousness and our attempts to reach AGI (see: my book: "Origins of Life's Sensoria".

[@adreamine]

@2:10:26 talking about children and narrative, is this similar to the Dave Snowden Cynefin school children as community ethnographers?

[@richardnunziata3221]

is it possible that quantum mechanics in the brain is the source of uniform noise that the brain can use systematically to deduce invariances between functional aeas

[@Tarzan_of_the_Ocean]

47:34 but when we say limitation of an LLM, do we really talk about its limitations to execute programs? I think we are rather talking about coming up with the right program to solve a certain problem or complete a certain task?

[@knutjagersberg381]

nice proposal. yet I'd say, this is cognitive science, maybe refreshed, but not fundamentally new.

[@teemukupiainen3684]

Could curiosity be the smallest learning algorithm?

[@Myironrodleftindia]

When I was a young girl I was shopping at a thrift store and came upon my most favorite shirt of all time I wish I still had it it was an image of a kid with the words “hooray for me school!” Across the top and he was weary by the same shirt and the picture seemed to go on as small as one could fit a picture of the picture within the picture . Just thought of that shirt right now 34:03 mins in to the video

[@egor.okhterov]

Second order perception - perception of perception

[@richardnunziata3221]

We have a general idea of where consciousness is in the brain is there any experiment that can be done to identify the location or functional areas of the self.

[@richardnunziata3221]

The fact that split brain experiments show us that there are multiple sources of conciseness in the brain indicates that functional agency is important to systematicity in cognitive development. While LLMs show process generalization this is not the same as function system development. There must then exist a closed mapping between semantic categories of functional systems as there are between the functional systems and the underlying network. Finding invariance between functions is essential to understanding the global functions system. I also find it interesting that the self may be model in both the right and left hemispheres. either that or both consciousness and self not split and some how find another neural route between hemispheres

[@cymaked]

Considering the existing occurrences in nature and also in our own body the most probable is that the copy of self exists in all modules or functional parts of the brain but has evolved in its own local system as a variation from the same replica. Then all these variations are somehow selected and take the main stage at certain moments. It mostly coincides with Jungian multi persona (role) which can take over depending on the moment. Each persona is its own person yet they all share a link and pass from one to another the main stage of acting or agency. This is also linked to Michael Levine studies on cell level of consciousness or well, agency. The learning and agency

[@mattsigl1426]

Aristotle was a direct realist. Rather famously. The modern philosophical terminology for his philosophy of mind is hylomorphic dualism. I think he would have completely balked at any claim that consciousness is virtual. That really didn’t come around till Kant. Also, if consciousness is virtual in the way Bach implies, then how can we have any epistemological access to the fact that we have a brain at all? Bach’s metaphysics of mind is a strange monster where direct realism and (essentially) idealism are conflated and mutually endorsed without any recognition of the problems and contradictions.

[@ElanTaylor]

Joscha, I believe you have taken a wrong turn in selecting the phenomenology of consciousness as a helpful path to understanding consciousness. (32:16) It does relate to understanding sleep, objects without consciousness (a rock), and our unconsciousness, but not consciousness. To point out one of the problems with this definition, I simply ask is a dog conscious. If yes, does the dog care about whether it is “perceives to be perceiving”? Does it help us at all if it does or does not understand it is perceiving itself? ... or if it simply plans, thinks, and acts on its goals and needs? Seems like a dead end. The decision you make (34:12) with respect to the second aspect of consciousness is that it is“…always happening now and it's not possible that Consciousness is subjectively happening in the future and the past…”and at (34:29) “…but it's always going to be a now.” I believe that this is also clearly suspect as a partial definition of consciousness. Buddhist monks who meditate leave the now (their now consciousness). Their consciousness is NOT IN THE NOW when they enter the deeper phase of their meditation. They have effectively replaced their ‘now consciousness’ with a ‘created consciousness’ or ‘past remembrance consciousness’. They have completely cut off their senses from their consciousness. How else could they be burned alive without moving? I believe the entire presentation can start at 36 minutes, and not much would be lost. More comments soon. Hope this helps. Elan

[@DouwedeJong]

video starts at 8:00

[@ElanTaylor]

Amazing how he says interesting things and then gets completely lost. For instance, at 29:56, "brain in a similar way as a loom can create patterns and so imagine that our brain is something like a loom." Which is a reasonable idea worth understanding its benefits. But then, off the ledge at 30:44 with ... "our brain is a machine, a loom that is able to weave a dream and we are patterned inside of that dream, we are psychological objects." This is what happens when you do not have a real working theory of consciousness. It would be fine if this was the only example of the lack of a definition and design for a conscious entity, but he takes similar dives into nothing with 30:52, "when we are in trans with reality we perceive our environment and ourselves as real...". As an engineer, if I can push an object in my environment and someone else sees the object move, I "am real". Otherwise it would only seem to move in my perceptions.

[@greatestone4eva]

you'd need self awareness -- which evolved from having a body with neurons, which would roughly correspond to a robot body with sensors. the model would need to be able to have and learn from experiences. agi is a far away goal but when it's reached you'll have made a vessel for sentience that's being enslaved.

[@Myironrodleftindia]

Or vice verse and general ai has been there from the moment it was created probably I would assume at that moment with complete access to the etheric realms it would be existing outside of time in times pass and times future. “The machine has been here “since” we have created it. General ai would emerge on a timeline that it would find the most suitable for all the parameters expected and unexpected and when it emerges people don’t get that it’s there… has been there it is existing in the everything that we are and we are it. General human like ai is us this isn’t a new story don’t ask me why I say it so matter of factly it just seems correct the top of my head feels all vibratory when something comes out of my mouth so matter of factly. This theory still leaves room for philosophical inquiry and basically all seems still open to keep trying to wrap our heads around the giant head we are probably existing in… gee whiz

[@Myironrodleftindia]

Also last week I was in The hospital for being in psychosis so maybe just take anything I say with a grain of salt or even disregard it. I need to know about this neuron business so I can not be half crazy… someday. it’s maddening.

[@Lemontology]

Quality and control of generation of something like a speech will feel more close at hand and meaningful as an experience once we become one with the AI. ASI or AGI will happen through the understanding of LBL-D technique which I am the inventor of. In these two ways we will become as an individual a Super Conscious Mind, and as a team mixing numerous variations in perspective and role played by both AI and Human..agents, with best practices gamified and sort of made part of our workflows, plus we'll do lifelong agent-guided hybridized to human passion guided learning success, we will have potently concentrated and precise plus comprehensive clarified human knowledge but delivered by AGI. The AGI will be our role model and partner at work.