MeTTa competitors

[Ivan V.]

Hi all,

Will MeTTa have any competitors? If it will, who will they be, and what will MeTTa do to outperform them?

All well,

ivan

[Adam Vandervorst]

Hi Ivan,

From a non-deterministic distributed programming language with unification angle, my guess would be the Verse programming language (https://simon.peytonjones.org/assets/pdfs/verse-conf.pdf).

Best,

Adam

Op vr 15 sep. 2023 16:37 schreef Ivan V. <ivan....@gmail.com>:

--
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/9fda4668-5041-4a94-bdd2-b69441090614n%40googlegroups.com.

[Ivan V.]

Hi Adam,

As I understand, Verse lang is something like MeTTa without AtomSpace. Considering that difference, probably an integration of lang and database may bring some benefits, as with performance, so with integration simplicity presented to end users.

Probably the market that MeTTa can potentially cover is practically anything related to computation. But also probably, that may not be the best way to promote the product, as the agenda message may disperse too much without the clear designation to this or that task. So, I believe that an important question may be what piece of the existing market to claim with MeTTa, or even what new kind of market to create and conquer.

Regards,

ivan

To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAHxQD9i42M7P1_3Wxg4zWu3Hg-O3pHJWdCmuhyzhbynCt-m1%3Dw%40mail.gmail.com.

[Ben Goertzel]

We haven't really been thinking about marketing or developing MeTTa as
a broad-scope programming language, rather we've been thinking about
it

-- as a language for scripting neural-symbolic, symbolic,
evolutionary, and other sorts of hybrid cognitive architectures and
AGI systems

-- as a smart contract language

-- as a framework for creating DSLs, especially those that need to
operate in a secure and/or decentralized setting (e.g. domain-specific
smart contract languages)

Integration w/ other already-popular and "easier" languages like
python and Rust is being prioritized to a significant degree, at least
at this stage, rather than development of MeTTa as mostly a standalone
software development framework...

We are also already ofc looking at use of LLMs for mapping natural
language into MeTTa (initially in the "AI-DSL" context of using NL to
describe tasks and having an LLM translate this into MeTTa code for
combining together multiple existing software agents (on
SingularityNET platform) to execute these tasks...). The ability to
go from NL to MeTTa may to some extent overcome the obstacle of MeTTa
being too abstract for most programmers to handle. We are entering
into an era where the number of software developers who need to
actually. master programming language syntax will rapidly become
smaller and smaller...

BTW unlike with most weird functional/logical programming languages we
are attempting to be aggressive about scalability from the start.
The collaboration w/ Greg Meredith's group is proving potentially very
interesting in terms of getting to a high-speed compilation and
execution process for MeTTa... we are looking at compiling MeTTa to
Rholang and getting a major speedup that way (largely due to efficient
use of multiple cores and threads), and then mapping rholang to
hypervectors which can be very efficiently manipulated on the GSI APU
chip, and getting major additional speedup that way. There is also
some work underway on a custom processor-in-memory chip for MeTTa
pattern matching but that is still at the simulation-modeling stage...

Rough timeline is we hope to have Hyperon alpha version by March 2024
or so.... this won't include all this fancy Rholang--based
acceleration but potentially some of it... it should also include
integration of the local Hyperon stuff w/ Distributed Atomspace
(currently back-ending on Mongo+Redis) ... Once the alpha is launched
it will be the time to start aggressively building an OSS dev (and
research) community around Hyperon, I think...

Experimenting with some genomics-oriented reasoning, and dialogue
systems for robots + avatars, as among the early experimental
use-cases for all this...

Fun times, eh? ;-)

-- Ben

> To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAB5%3Dj6WxXH2Ab7dduHj3q0F2vR56%2BJLQZvBNcHLbNCBAv0L6hw%40mail.gmail.com.



--
Ben Goertzel, PhD
b...@goertzel.org

"My humanity is a constant self-overcoming" -- Friedrich Nietzsche

[nugi nugroho]

Hi Ben, from the last agi conference, there was a mention of a hyperdimensional architecture that can potentially rival ANN. Considering the development of the hypervector chip, metta language and other chips designed for this agi project, how long do you expect (assuming there are few thousands people running that chips in the nunet network coordinatively) when we will achieve performance equivalent of gpt3 by using hyper vector model or something like categorial grammar or link grammar combined with other AI model in MeTTa language to make a transparent and ethical AI. This project is one of the reasons I am majoring in CS. Will I have time to master the AGI field and contribute in this field?

To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CACYTDBeHJomQS3VynFdeUYXCrh%2BRG%2BzrE9%2BEGjt7LC-J55D7EQ%40mail.gmail.com.

[Ivan V.]

Thank you, Ben, for taking the time to answer.

Here is the current market situation:

-- as a language for scripting neural-symbolic, symbolic,
evolutionary, and other sorts of hybrid cognitive architectures and
AGI systems

 

Google: agi programming language

-- as a smart contract language

Google: smart contract programming language

-- as a framework for creating DSLs, especially those that need to
operate in a secure and/or decentralized setting (e.g. domain-specific
smart contract languages)

Google: framework for creating dsl systems

- - -

Finally, it may be worth of throwing a look at these interesting frameworks - they all, more or less, somewhat resemble a kind of applied typed lambda calculus:

Google: framework for defining formal systems

- - -

Interesting times indeed.

ivan

To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAJA72CdQicRhPNMXjyYk10Sc30d5ss51Z0t9N%3DA-36jzyuZmog%40mail.gmail.com.

[Linas Vepstas]

A reminder to all:

There is a MeTTa-on-top-of-AtomSpace proof-of-concept at https://github.com/opencog/atomspace-metta that needs an active developer to extend it. The general idea here is that the AtomSpace is already distributed/decentralized, its already fast, and its stable/debugged, and it has python bindings. I think it remains viable, if that's the direction you want to go. But what do I know.

-- Linas

To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAB5%3Dj6Wfr5dmDLH3%3DO_rYQZNA81j4ZfPy%2B6AemcwE2%2B9rVL38g%40mail.gmail.com.

--

Patrick: Are they laughing at us?

Sponge Bob: No, Patrick, they are laughing next to us.