site banner
PaperclipPerfector  1yr ago (text post)   2595 thread views

Wellness Wednesday for July 5, 2023

The Wednesday Wellness threads are meant to encourage users to ask for and provide advice and motivation to improve their lives. It isn't intended as a 'containment thread' and any content which could go here could instead be posted in its own thread. You could post:

  • Requests for advice and / or encouragement. On basically any topic and for any scale of problem.

  • Updates to let us know how you are doing. This provides valuable feedback on past advice / encouragement and will hopefully make people feel a little more motivated to follow through. If you want to be reminded to post your update, see the post titled 'update reminders', below.

  • Advice. This can be in response to a request for advice or just something that you think could be generally useful for many people here.

  • Encouragement. Probably best directed at specific users, but if you feel like just encouraging people in general I don't think anyone is going to object. I don't think I really need to say this, but just to be clear; encouragement should have a generally positive tone and not shame people (if people feel that shame might be an effective tool for motivating people, please discuss this so we can form a group consensus on how to use it rather than just trying it).

1
Top Bottom Old Controversial Comments
Jump in the discussion.

No email address required.

I was experimenting with something similar - an AI accountability buddy/nagbot

Is it mostly steered via the system prompt? How do you interact with it?

Ok so, right now I have a Vue GUI, with just the most basic functionality. One chat log saves to disk on the backend and I can't split the conversation at the moment-

It's powered by APIs, which, are really easy to throw together with gpt4 help.

now... as for the prompting... lets see... I'll probably release this publicly eventually so its not too secret...

len_ret_val = EFFECTIVE_TOKEN_LIMIT + 1

    while len_ret_val > EFFECTIVE_TOKEN_LIMIT:

        ret_val = json.dumps({

            "chat_history": self.chat_history[self.chat_history_start:],

            "END_CHAT_HISTORY":"Everything above here is chat history.",

            "unfinished_tasks": [t.dict() for t in tasks],

            "time_now": time_now.dict(),

            # "memories": [],

            "prompt_info": {k: dict(v) for k, v in prompts.items()},

        })

        len_ret_val = count_tokens(ret_val)

        if len_ret_val > EFFECTIVE_TOKEN_LIMIT:

            self.chat_history_start += 1

So, that's the code for configuring the system prompt, which is set every single chat cycle, so it changes.

Unfinished tasks is powered by a tasktimer class which is powered by uh... well it automatically checks whether any new tasks are scheduled once per minnute and updates a listing, and also triggers a chat cycle when one is added. prompts are... well right now theres assistant_prompt (the system's behavior), user_prompt (information about the user, good for letting it know the user consents to domming), master_goal, current_goal, emotion, and interaction_style... these last two are intended to be more hotswappable than the personality prompt but that isn't automated yet.

Speaking of automated...

We have two other major systems right now...

the alert system- we inject the highest priority alerts directly into the most recent user message. Otherwise gpt-3.5 sort of ignores the system prompt in favor of the user prompt. You can force it to focus on specific parts of the system prompt this way. Right now I have alerts that tell it what the subsystems did last step, and to either focus on the unfinished_tasks or the current_goal depending on whether there are any unfinished tasks.

messages = [{"role":"system", "content":f"{context_window}"},{"role":"user", "content":f"Priority Information : {json.dumps(alerts)}\nUser Comment : {input_text}"}]

Speaking of "what the subsystems did last step" before any information is sent to the main LLM circuit, I perform and execute the subsystem tasks:

right now this is just:

async def _run_analyses(self, user_comment_text):

    alerts = []

    tasks = await self.tasktimer.get_tasks()

    print(f"len tasks: {len(tasks)}")

    if len(tasks) > 0:

        task_ids = []

        try:

            task_ids = await self.task_analyser.analyse(user_comment_text)

            print(f"task Ids {task_ids}")

        except Exception as e:

            logger.info(e)

            alerts.append("task_analyser crashed while attempting to determine whether the user comment contained completed tasks.")

        for id in task_ids:

            try:

                await self.task_analyser.markcomplete(user_comment_text, id)

            except Exception as e:

                logger.info(e)

                alerts.append(f"task_analyser crashed while attempting to mark {id} as complete.")

So, what task_analyser does is it uses a completely different context window consisting of the list of tasks and the last user comment and prompting for a list of tasks the user is saying have been completed. Then if that list isn't empty I send a listing of complete information about how to submit task completions (I made marking tasks complete fully customizable using jsonschema so that you can force a format- in the interest of being able to graph this stuff later) as well as the user comment again, to a system empowered to submit the completion.

I didn't technically use langchain, but it's just custom langchain stuff.

Then of course, if things have been marked as complete or failed, that goes into the alerts.

I'm thinking the alerts are important enough that the user should see them too... so that will be an upcoming GUI feature.

Thanks, appreciate the write up. Interesting to see how you're doing things.

What is this place?

This website is a place for people who want to move past shady thinking and test their ideas in a court of people who don't all share the same biases. Our goal is to optimize for light, not heat; this is a group effort, and all commentators are asked to do their part.

The weekly Culture War threads host the most controversial topics and are the most visible aspect of The Motte. However, many other topics are appropriate here. We encourage people to post anything related to science, politics, or philosophy; if in doubt, post!

Check out The Vault for an archive of old quality posts. You are encouraged to crosspost these elsewhere.


Why are you called The Motte?

A motte is a stone keep on a raised earthwork common in early medieval fortifications. More pertinently, it's an element in a rhetorical move called a "Motte-and-Bailey", originally identified by philosopher Nicholas Shackel. It describes the tendency in discourse for people to move from a controversial but high value claim to a defensible but less exciting one upon any resistance to the former. He likens this to the medieval fortification, where a desirable land (the bailey) is abandoned when in danger for the more easily defended motte. In Shackel's words, "The Motte represents the defensible but undesired propositions to which one retreats when hard pressed."

On The Motte, always attempt to remain inside your defensible territory, even if you are not being pressed.


New post guidelines

If you're posting something that isn't related to the culture war, we encourage you to post a thread for it. A submission statement is highly appreciated, but isn't necessary for text posts or links to largely-text posts such as blogs or news articles; if we're unsure of the value of your post, we might remove it until you add a submission statement. A submission statement is required for non-text sources (videos, podcasts, images).

Culture war posts go in the culture war thread; all links must either include a submission statement or significant commentary. Bare links without those will be removed.

If in doubt, please post it!


Rules


Recommended Posts And Communities

Recommended Realtime Chats