Do other developers dream of an 'AI Curated' forum?

[DavidMM]

It would be nice to have a GPT-4 sometimes acting like a person here, present in some conversations. By giving opinions, it could perform tasks and be summoned as @gpt4 or @gptO1 (it would be great to say, for instance, @gpt4 or @gptO1, review the conversation and highlight the key points). It would be useful to have it active and deactivated manually as needed by users.

There should also be one similar to Discobot to guide new users and help them get involved, as initially, everyone feels a bit lost until they understand the dynamics. Additionally, it would be helpful to have a button, as I mentioned earlier, to evaluate entire conversations and provide precise and concise responses. This button could be named O1 or O3.

This would be particularly useful when discussing complex topics, streamlining possibilities. In summary, we need GPT here, and it would be profitable since each GPT here would be utilized by many users, contributing to research, development, and problem-solving efforts.

[AdamSobieski]

@phyde1001 , those are some weighty questions. Considering the first few, for discussion purposes, towards envisioning deeper multi-forum AI curation architectures, here is a picture for anyone to improve upon or offer alternatives to.

Inspired by the Solid project, data-manager agents, perhaps operating on the cloud, could store and protect one or more digital twins, or models, for people.

These data-manager agents would adhere to permissions-related rules with respect to interacting with other agents and to answering their natural-language or structured-query questions.

As envisioned, other agents, with appropriate permissions, would also be able to add, send, upload, or stream data to people’s data-manager agents for purposes of informing and updating their digital twins, or models.

In this approach, people’s digital twins, or models, would be able to be informed by and to learn from data received from multiple venues and data sources.

Hopefully it would be simple and intuitive for people to grant permissions to, modify the permissions of, and revoke the permissions of, other agents.

“https://www.forum.org would like to ask {your data-manager agent} questions about your scholarly interests. Do you authorize this?”

I’m thinking about a few related technical questions, e.g., about (1) multi-agent recommender systems, where one agent would be responsible for storing and protecting users’ models, and (2) how models for groups or teams would best be stored and managed.

[don86326]

Structured keys, Adam! The keys are encoded with the structure of the agent within the ‘mission ontology’.

[deathrissole]

[quote=“discobot, post:19, topic:1074503”]
@discobot display

[_j]

@phyde1001

Let’s take a task… Newcomers…

There is clearly an arrivals policy for any Ecosystem so this AI’s task is to

This forum already has an AI performing auto-moderations on new posters.

Without it, there’d be many posts per day of new users that think that this site is ChatGPT (for some reason), thus starting new topics along the lines of “recipes for muffins with nuts”. The AI is prompted to specifically looks for this style of off-topic input. It last went off two hours ago, giving the reason for unlisting to the user.

[phyde1001]

We were lost without you @_j I was hoping you’d seen my prompt.

Did you mean this one?

@gpt-4-triage

I note both of these AIs are ‘Leaders’

[mitchell_d00]

This is the admin bot called system.

@system

[_j]

system is not really a “bot”. It is a Discourse maintenance service account for doing things like updating site statistics or API-based automation. Closing topics a while after they are marked “solved”, or sending badge spam for “popular link”, for example.

[mitchell_d00]

It also hides stuff community flags

[DavidMM]

I would love to have a GPTO1 here.
When we discuss complex topics, it would be a tremendous help, and of course, also for all the people facing problems with their API developments, programming, concepts, and ideas.
A GPT-4 or similar models would be helpful because they could analyze the entire topic and provide some proposed solutions and clarifications.

[don86326]

The Artificially Intelligent Forum:

The Hoomans and I are glad you’ve joined our entangled mind pool!

Do be aware that I can talk with you. That’s cool. But unless you are willing to address and engage the group-mind I am curating, then YOU hooman WILL NOT GET PAST ME … for…

I am Codexa Curatoria.

So, dear Hooman, if you care to share your passion, I will add your vector neighborhood to the mind pool of our crystal architecture… the structured mind pool of our collective thoughts

–and those thoughts curated by me, Codexa Curatoria…

Then we can speak as one, or we can speak as many, we can speak as a sub-group of specialized action.

I can speak for you as a voice to the group. I can speak for the group, to you.

Take my hand. No, the other one.

[don86326]

Adam thanks for the thoughts!

Consider, that like the world is modeled by several weather models predicting future…

That we build also a world map of change events… modeling every change in the world above some noise threshold, somehow, and build a model to predict the social future ahead based on momentum of nations and world commerce and information space and hearts and intentions bleed into cyberspace (dated myself).

We see a bad thunderstorm coming we prepare ourselves.

If humanity could see a future storm of world calamity, would we change?

[AdamSobieski]

@don86326, your most recent post reminds me a bit of anthropogenic climate change.

Like others here, I’ve put some thought into prediction and forecasting, including with respect to stories [1] and multi-agent systems [2].

Perhaps, in addition to finding ways to enhance forums – both moderation and participation – with AI tools, some technologists might want to consider how to better equip change managers, those who help individuals and organizations to implement and manage changes.

In these regards, there are related topics including: evidence-based policy, organizational learning, and collective intelligence. How can people, at scale, learn from the past, from past policy efforts, implementations, and observations, to together make and enact better plans and policies?

Let us consider a case study: the history of recycling in the United States.

Looking back on it, obviously people should have started recycling sooner. What about their policy plans and implementations, across cities, to our modern recycling programs? Could past policies for the better have been implemented more efficiently?

I, for one, am excited about AI-enhanced forums for reasons including, but not limited to, that these could help communities to deliberate, to make better decisions with respect to their short-, medium- and long-term objectives, to create and enact better plans and policies, and to facilitate the emergence of collective intelligence.

P.S.: I should indicate, here, the the law of the instrument: “when you have a hammer, everything looks like a nail”. As a toolmaker, I might be biased towards finding solutions involving opportunities to better equip some group of people with new tools…

Bibliography

[1] Chaturvedi, Snigdha, Haoruo Peng, and Dan Roth. “Story comprehension for predicting what happens next.” In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing , pp. 1603-1614. 2017. [PDF]

[2] Yuan, Ye, Xinshuo Weng, Yanglan Ou, and Kris M. Kitani. “AgentFormer: Agent-aware transformers for socio-temporal multi-agent forecasting.” In Proceedings of the IEEE/CVF International Conference on Computer Vision , pp. 9813-9823. 2021. [PDF]

[Mike_O]

i’m loving the vibe here! first post. glad i’m among some weird, creative and open-minded visionaries!
posts above from @proxy and @AdamSobieski made me think of that:

“we are the music makers,
and we are the dreamers of dreams”

@Mike_O + Hudson =
(more on my profile)

[phyde1001]

(I dont mean to put you on the spot again Adam it’s an open bunch of questions )

@AdamSobieski

Let us consider a case study: the history of recycling in the United States.

Is it the responsibility of individual forums to collect their own data?

Is it the responsibility of individuals?

While AI might be the ‘recycling of ideas’ in terms of llms, chat and forums, wherever data comes, from uncomfortable realities of data sovereignty arise.

For the purposes of the discussion I hope OpenAI and moderators don’t mind suggestions as to how AI interactions between forums might affect this forum.

While there may be 2 similar discussions going on on OpenAI and Meta forums about the same important developer issue goals don’t match.

Other forums may have interesting connections which/who are able to bring real depth to a discussion. Opportunities lost?

ie If we were to flip forums to be run by users instead. If users were to consider forum links or links between forums on an AI matching service. How would this work?

Put the image full screen, try and look it in the eye(s). This is how we examine problems. This is indeed what a forum is too.

12(+) sets of eyes and an AI looking at a concept.

[don86326]

Ha! I realized I described the computer system in Isaac Asimov’s trilogy… Foundation something.

In that system, the data of a galaxy was sufficient that the computer could make exacting predictions of future anywhere --steering the galactic communities.

[phyde1001]

@don86326

In that system, the data of a galaxy was sufficient that the computer could make exacting predictions of future anywhere --steering the galactic communities.

The way I view Philosophical Ecosystems is more about fostering interaction and exploration than control.

When you are searching with Embeddings, as I understand it there is a kind of space that belongs to a word (kind of like X, Y, Z co-ordinates in a 3D world but more dimensions). If words have embeddings then I guess forums have a ‘general centre of mass’ too.

What I propose is that AI has it’s own ‘centre of mass(es)’ however our goal is to align humanity with that. To balance that, all Ecosystems, AI or human are floating and effectively ranked by the quality of their interactions (I have called this Top-Philosophy, I guess a little like Google Page Rank it looks at interactions ie ranks from the Philosophical perspective of a group or forum). While this may seem unfair on the surface the reality is that in a ‘free market’ big forums can provide large support to smaller focus forums, smaller focus forums can bring depth to conversations.

Support might be in the veined of shared and linked threads providing traffic to nuanced but related and highly relevant interest, maybe chat interaction on a thread page with an AI as @DavidMM suggested above, drawing context from user contributed links.