Erratum: *"Even OpenAI has embraced this idea to an extent. As I cite in my talk"
Sorry, that should read OpenCog. I don't think OpenAI has embraced it. It would be nice if they did. On Sun, Nov 14, 2021 at 7:52 AM Rob Freeman <chaotic.langu...@gmail.com> wrote: > Hi John, > > I probably should have read this thread earlier. > > I agree with your insight. I have been pushing this idea that cognition, > or at least specifically natural language grammar, is lossy, for some time > now. Matt Mahoney may remember me pushing it re. the Hutter Prize to > compress language, when that came out. > > And yes, this relates to the idea that "true and false don't purely exist > as crisp booleans". Which actually has become a big theme in philosophy, > and is tearing society apart right now. > > But I suggest a re-brand. More recently I've started expressing it not so > much as the idea that cognition is lossy, but more that cognition is an > expansion. > > If you think of cognition as an expansion I think you'll get most of the > lossy compression insight you are seeing. In short, if cognition is an > expansion, details matter. > > There is now a handful of work which I think you can interpret this way: > > Tomas Mikolov - "We can design systems where complexity seems to be > growing". > Bob Coecke - "Togetherness". And a thread of quantum cognition emphasizing > subjectivity of category. Which is maybe not quite expansion, but it has > the rejection of abstraction aspect. > > And of course I have such a model which I presented most recently at > AGI-21: > > Vector Parser - Cognition a compression or expansion of the world? - > AGI-21 Contributed Talks > https://youtu.be/0FmOblTl26Q > > Even OpenAI has embraced this idea to an extent. As I cite in my talk: > > Vepstas, “Mereology”, 2020: "In the remaining chapters, the sheaf > construction will be used as a tool to create A(G)I representations of > reality. Whether the constructed network is an accurate representation of > reality is undecidable, and this is true even in a narrow, formal, sense." > > Technically, to avoid arguments about what is lossless and what not, I > suggest you focus on the decidability result. > > Personally, as I describe in my talk, I think it simplifies AI > tremendously. Roughly comparable to taking all the stuff we have now, but > turning it upside down. At which point it ceases to be a lot of confusing > detail, but becomes instead some rather nice, compact, productive > principles. > > Which is nice and inclusive. Because it means that nothing which has been > done in AI up to this point is really wrong. We've just been interpreting > it wrong. We can use most of it. And don't need to do a lot of work > starting from scratch. > > But it does mean we need to change the way we think about the problem. > > -Rob > > On Thu, Nov 4, 2021 at 11:50 PM John Rose <johnr...@polyplexic.com> wrote: > >> While performing thought experiments on an AGI model I realized that >> there is no purely lossless compression. Something is always lost. For most >> practical purposes yes lossless exists. This might sound trivially obvious >> and non-obvious but it does impact the theory in the model. >> >> In other words, I could not imagine any purely lossless compression, it >> might physically exist I just can't imagine it as I'm not a physicist. So >> maybe it does exist? or perhaps we just prefer it to be so... I suppose >> it's the same as saying true and false don't purely exist as crisp >> booleans. And, exists doesn’t purely exist…so everything is relative. But >> the implications are enormous when dealing with chaotic and complex systems >> models. Thus it being trivially obvious and trivially non-obvious or, >> non-trivially non-obvious... or... >> >> Net effect? Zero. Oh wait zero doesn't fully exist now does it. WTH? >> >> https://www.youtube.com/watch?v=JwZwkk7q25I >> >> >> *Artificial General Intelligence List <https://agi.topicbox.com/latest>* >> / AGI / see discussions <https://agi.topicbox.com/groups/agi> + >> participants <https://agi.topicbox.com/groups/agi/members> + >> delivery options <https://agi.topicbox.com/groups/agi/subscription> >> Permalink >> <https://agi.topicbox.com/groups/agi/T5ff6237e11d945fb-Mce05fa9f1ab04ee9cd87e46f> >> ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T5ff6237e11d945fb-M7ee24374cc77dabf9bd52fdf Delivery options: https://agi.topicbox.com/groups/agi/subscription
