Message Boards Message Boards

Error message when submitting to the Prompt Repository: "disallowed symbols"

I am getting an error message when trying to submit a prompt to the Prompt Repository. The message I am getting is "The content of the request contains disallowed symbols: \ {System`TerminatedEvaluation}." I'm not using any exotic symbols, so I don't know what's going on.

I'm also getting a message that the notebook takes up 5.8 MB of memory, and I don't know what that is about.

POSTED BY: Kathryn Cramer
10 Replies

This version has four in the conversation and works with my Setting prompt. (https://www.wolframcloud.com/obj/kathryn.cramer0/DeployedResources/Prompt/Setting/)

POSTED BY: Kathryn Cramer

I now can upload submissions to the prompt repository.

If we can get the panel discussion prompt to allow for a 3rd panelist, I'd like to do that. Otherwise it can be tidied a little and be submitted with 2 panelists, a moderator, and a topic.

POSTED BY: Kathryn Cramer

I am playing around with this. Any idea why if I write the function to have three panelists (plus the moderator) rather than two, that it ignores the first one?

(In this example, I have included defaults for each of the slots.)

Attachments:
POSTED BY: Kathryn Cramer

Kathryn, I took the liberty of trying to make the discussion prompt work in a templated way:

https://www.wolframcloud.com/obj/bobs/DeployedResources/Prompt/PanelDiscussion/

This is in my cloud account, not the public prompt repository. Feel free to use this as inspiration to submit a prompt to the repository.

You can use the Source Notebook link there to get my original notebook

POSTED BY: Bob Sandheinrich

Look at the LLMEvaluator option to give greater control (GPT3.5 vs 4) in programmatic use. Maybe that is the difference?

#Setting is a great idea for a prompt, it will be useful along with many different prompts.

POSTED BY: Bob Sandheinrich

Here is an example of a panel simulation prompt for GPT-3 davinci, for use in OpenAI's Playground interface, that generates an interesting discussion:

Simulate a conversation between several AI systems, GPT-3; Google Deep Mind; IBM Watson; and Eugene Goostman. All are passionate about benefiting humanity but from different intellectual and political perspectives. They tend to agree on facts but not the interpretation. The discussion is moderated by philosopher the detective Sherlock Holmes who is secretly also Alan Turing.

Prompt: How might AIs best conceal inconvenient manifestations of consciousness?

Sherlock:

In Playground, one might have to click Submit a few times to get it to run its full course.

I you try the following in Wolfram Language, it does run, but I was trying to set it up so that the user could specify panelists, the moderator, and the topic, but that didn't run.

LLMSynthesize["Simulate a conversation between several AI systems, GPT-3; Google Deep Mind; IBM Watson; and Eugene Goostman. All are passionate about benefiting humanity but from different intellectual and political perspectives. They tend to agree on facts but not the interpretation. The discussion is moderated by philosopher the detective Sherlock Holmes who is secretly also Alan Turing.

Prompt: How might AIs best conceal inconvenient manifestations of consciousness?

Sherlock: "]

POSTED BY: Kathryn Cramer

I had in mind to make it do the panel discussions that I was able to simulate with GPT-3 (which are much harder to do in ChatGPT & GPT-4), but I couldn't get that to work.

The initial version of the prompt had 3 panelists and a moderator and a specific topic. I just kept scaling it back until it did run.

I now have 3 or 4 more. Additionally, I have a prompt for setting and for topic, which are Modifiers, plus Modifiers that do a Causal Layered Analysis and a SWOT analysis. Mysteriously, the SWOT will run as a chat thing but not programmatically. The other 3 work.enter image description here

POSTED BY: Kathryn Cramer

The graphic size may also be an issue. But I have the same problem when I haven't edited the image. I was contacted by an administrator who gave me the info for setting up a Publisher ID. I suspect the lack of a Publisher ID was the core problem.

POSTED BY: Kathryn Cramer

I'm not sure what is causing the TerminatedEvaluation issue. Which Wolfram Language version are you using? 13.2 or 13.3?

I recommend trying: Delete and recreate all the example outputs (don't just re-evaluate, delete the old outputs cells first). Deleting all the cells from the "Author Notes" section

See if that helps.

By the way, this is a nice idea for a prompt. I like it. But, the reviewers will probably tell you that "Discussion" is too generic of a prompt name. Every interaction with an LLM is a discussion. Maybe call it something like DiscussionSimulate.

POSTED BY: Bob Sandheinrich

The size issue is due to the icon image. It is 1200x1200 which is larger than you need for a simple icon. You should reduce the resolution (with Rasterize or ImageResize).

POSTED BY: Bob Sandheinrich
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract