AGI discussion group, Sep 10 7AM Pacific: Characterizing and Implementing Human-Like Consciousness

[Ben Goertzel]

Theme for discussion this week: Characterizing and Implementing
Human-Like Consciousness

See

https://wiki.opencog.org/w/AGI_Discussion_Forum#Sessions

URL for video-chat: https://singularitynet.zoom.us/my/benbot ...

Background reading:
https://www.researchgate.net/publication/275541457_haracterizing_Human-like_Consciousness_An_Integrative_Approach

Pretty soon we will have some new Hyperon design documents/ideas to
discuss in the AGI Discussion Group -- lots of progress occurring on
that front, but much of it not quite yet fully baked enough for
public-ish discussion ... so for this session we're going to explore
some highly relevant but more general topics... ;)


On Thu, Aug 26, 2021 at 9:22 AM Mike Archbold <jazz...@gmail.com> wrote:
>
> Is there a discussion tomorrow?
>
> On 8/16/21, magnusw...@gmail.com <magnusw...@gmail.com> wrote:
> > Brute forcing is about reducing the search in a way, to spread it out
> > further, the same amount of computation power.
>
> ------------------------------------------
> Artificial General Intelligence List: AGI
> Permalink: https://agi.topicbox.com/groups/agi/T5b614d3e3bb8e0da-Me75a2ed05b46ea5198406b29
> Delivery options: https://agi.topicbox.com/groups/agi/subscription



--
Ben Goertzel, PhD
http://goertzel.org

“He not busy being born is busy dying" -- Bob Dylan

[Ben Goertzel]

On Thu, Sep 9, 2021 at 5:06 PM <magnusw...@gmail.com> wrote:
>
> I dont think Man as a group has much chance in getting that to happen. Need to get more down to earth I think.

Our OpenCog/SNet team is spending a lot of time on down-to-earth
stuff, some of which we'll talk about in some future AGI Discussion
sessions

Mainly

-- design of a new programming language (MeTTA = Meta Type Talk)
designed to serve as the type system and formalism behind the new
version of OpenCog Atomspace. A lot of math/CS work has been done in
the last 9 months leading up to this design...

-- a new architecture for a high performance distributed Atomspace store

-- a prototype of current "OpenCog Classic" controlling characters in
Minecraft via a pure experiential learning framework

-- use of current "OpenCog Classic" together with neural nets for
controlling the Grace humanoid robot, and analyzing genomic and
clinical medical data for personalized medicine

The first 2 will be good AGI Discussion Forum topics, but we're aiming
to slightly more fully bake our design specs first...

However, these are also very nitty-gritty topics, and for fostering
discussion among folks pursuing AGI via different paradigms and
approaches, it can also be good to dip into some higher level topics
that aren't so particular to anyone's special approach...

FWIW I do think we can achieve human-like consciousness via Hyperon
design, but I also think it's not necessarily the most interesting
thing one can do -- there are fairly severe limitations to human-like
consciousness and in some ways I think our AGIs will be able to do
better...

ben

[Ben Goertzel]

>> -- design of a new programming language (MeTTA = Meta Type Talk)
>> designed to serve as the type system and formalism behind the new
>> version of OpenCog Atomspace. A lot of math/CS work has been done in
>> the last 9 months leading up to this design...
>
>

> Meta Type Talk...
>
> Is it a fair guess that there will be an element of Category Theory, sheaves, and the like, in MTT?
>

MeTTA has def. been designed with the triad of (programming languages,
categories, logics) in mind...

Working out the specifics of the Curry-Howard mapping from MeTTa to
intuitionistic logics, and from there to categorial semantics, is one
of the things on our plate for the next couple months

This will then allow a more rigorous and MeTTA-flavored version of the
stuff I sketched out here

https://arxiv.org/abs/2012.14474

as well as formulation of e.g. stuff from modal HoTT in MeTTA terms

https://golem.ph.utexas.edu/category/2020/02/modal_homotopy_type_theory_the.html

ben

[Dave Xanatos]

Thanks Ben, Nils and all who participated, I appreciate the thought enrichment, I have many trails to investigate after this.

I would enjoy it very much if, at some point, those of us interested could go "down that rabbit hole" of the hard problem of consciousness. And possibly to explore why it may, or may not, matter - if an artificial system is having an experience as we seem to, or if it is completely unaware, just flipping switches based on the position of other switches, so to speak. And why we even care...

Lastly, a quick comment on "Intention": I have been working on the conversational ability of my robots for a few years, which was very limited at best. I have also played extensively with GPT-2, which improved its "conversational" ability, but also was prone to going off the rails. What made the biggest difference is giving the robot a hard-coded "intention" or goal for the conversation. This allowed it to redirect conversation to an expected direction and subject via more -or less- elegant segues, and prolonged the length of time a conversation continued to make reasonable sense before it got way off track. I think there's something to the idea of perhaps encoding systems with a group or hierarchy of goals for interaction, an artificial "intentionality", right out of the starting gate, if the goal of the robot creator is to have their robot interact in a more "human-like" fashion.

Thanks again all,

Dave

--
You received this message because you are subscribed to the Google Groups "opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email to opencog+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CACYTDBfnBV-29VTEBP3rr_WsEb_Wh0E%2B6n%3DsOCCmJi7bTSTtgQ%40mail.gmail.com.