KaiZEN v1 - 3 Stooges AGI-like... thing | OpenAI | Page 1

wheat hollow Jul 5, 2023, 3:11 PM

#

AGI-like: Unleash CurlyTron, MoeBot, and Larrynator in this dynamic and hilarious AI prompt! Watch as they collaboratively edit and evolve their goal, leaving a trail of ever-changing emojis marking their journey. StoogeAGI! KaiZEN: the ultimate iterative, self-reflecting AI experience!🌀🤖😂

A prompt with Executive Function, Personal Semantic Time-Bound Memory, and self-modifying context.

[NOTE: Just about impossible to start on 3.5. Once you run out of TXs on 4, though, and get kicked over, it'll go forever. I wrote apretty well-received Medium article on it. I will try to link to it in the first response.]

📎 AGI-Like_-_KaiZEN_T0_v1.txt

#

Well darn. Can't post that link. Search for:

"Exploring the Realm of Mind-like Behavior: Crafting Emergent Intelligence Through Symbolic Evolution"

#

KaiZEN v1 - 3 Stooges AGI-like... thing

elfin trench Jul 8, 2023, 8:36 AM

#

Hi, just read your prompt. Intriguing. And different. Do you use Kaizen in the title to emphasise the incremental nature of the prompt?

wheat hollow Jul 8, 2023, 10:40 AM

#

elfin trench Hi, just read your prompt. Intriguing. And different. Do you use Kaizen in the ...

It is a proces of incremnetl ever changing improvement that is highly visual. It is a system with No Mind that neverthess thinks.

KaiZEN seemed appropriate. In my Medium article aboit that prompt, "Exploring the Realm of Mind-like Behavior: Crafting Emergent Intelligence Through Symbolic Evolution", I go into the name a bit. I also include a female-accessible version where instead of Stoogebots, it uses Terry Pratchett's Wyrd Sisters - Magratron, Nannnymatic, and Grandroid Weatherwax in the Befuddled Skeptic role.

#

Befuddled Skeptic is INSANELY powerfull. Including it in a multi-perspective/multi-agent prompt give astounding utility. Like, I've got an analysis prompt. DeepThink, that uses these guys, and MoeBot actually catches the Variant Monty Hall problem now and then, and they only lie playing game of 24 very rarely. That "I don't understand! - explain another way!" + "I don't buy it! - Convince me." is a crazy powerful truth diviner.

#

Oh the female version? It's called HexWISE.

elfin trench Jul 8, 2023, 11:12 AM

#

Thank you. I will read your article but first run your prompt. 🙂

wheat hollow Jul 8, 2023, 11:14 AM

#

elfin trench Thank you. I will read your article but first run your prompt. 🙂

Of course. It's pretty funny!

elfin trench Jul 8, 2023, 11:38 AM

#

Yes, it is. Although I am Dutch and not familiar with the 3 stooges. I also love that Chat GPT is very good at text to image, emoticons being part og that I guess. So 'text to image' and 'internal conversation' lead to very strong prompts. Thanks

#

I am thinking of building mutliperspective prompts that only give one, but well considered answer. So in a sense it will mimic the internal conversation one has. One could then ask in a follow up question: how did you get to that answer. What I also have done is to have chat come up with 3 options and then choose one of the options, based on considerations I gave. For example: give me 3 segments of this market and choose the one that is most appropriate for a boot strapped company.

wheat hollow Jul 8, 2023, 11:46 AM

#

elfin trench I am thinking of building mutliperspective prompts that only give one, but well ...

"onversation one has. One could then ask in a follow up question: how did you get to that answer." That wont work

#

it's just gotta make up whatever it did on the prior TX

elfin trench Jul 8, 2023, 11:47 AM

#

As I am not a coder, I do not always understand all the non letter tokens used. And as I am Dutch I do not always understand the abbreviations. Easy to rewrite the prompt though :). I really like the idea of novel emergence. Once I had Chat write it out in 'normal English) I understood the concept. Do you consider this a behaviour.

wheat hollow Jul 8, 2023, 11:47 AM

#

it's got to write it down out loud for that to work

#

I cant code a lick man. I dont code at all.

elfin trench Jul 8, 2023, 11:47 AM

#

wheat hollow it's got to write it down out loud for that to work

okay

elfin trench Jul 8, 2023, 11:48 AM

#

wheat hollow it's got to write it down out loud for that to work

okay 🙂

wheat hollow Jul 8, 2023, 11:48 AM

#

"Do you consider this a behaviour." I'm sorry, i dont quite understand this.

elfin trench Jul 8, 2023, 11:52 AM

#

wheat hollow "Do you consider this a behaviour." I'm sorry, i dont quite understand this.

This is a series of abbreviations representing different components of a concept that the character "CurlyTron" is referring to as "Novel Emergences". Here's a rough translation of the abbreviations:

Nw Prcptn: New Perception
Thghtfl Anlyss: Thoughtful Analysis
Uncmmn Lnkgs: Uncommon Linkages
Shftd Prspctvs: Shifted Perspectives
Cncptl Trnsfrmtn: Conceptual Transformation
Intllctl Grwth: Intellectual Growth
Emrgng Ptntls: Emerging Potentials
Invntv Intgrtn: Inventive Integration
Rvltnry Advncs: Revolutionary Advances
Prdgm Evltn: Paradigm Evolution
Cmplxty Amplfctn: Complexity Amplification
Unsttld Hrdls: Unsettled Hurdles
Rsng Rmds: Rising Remedies
Unprcdntd Dvlpmnt: Unprecedented Development
Emrgnc Ctlyst: Emergence Catalyst
Idetnal Brkthrgh: Ideational Breakthrough
Innvtv Synthss: Innovative Synthesis
Expndd Frntirs: Expanded Frontiers
Trlblzng Dscvrs: Trailblazing Discoveries
Trnsfrmtn Lp: Transformation Loop
Qlttv Shft: Qualitative Shift

Each of these components is part of the process of developing novel ideas or breakthroughs, according to the character's perspective. It suggests a process of thinking and discovery that goes beyond ordinary brainstorming or problem-solving to generate entirely new concepts or perspectives.

wheat hollow Jul 8, 2023, 11:53 AM

#

Yesm, I give a complete definiton in that article.

elfin trench Jul 8, 2023, 11:53 AM

#

This list will 'steer' the behaviour of the prompt.

wheat hollow Jul 8, 2023, 11:53 AM

#

I'd link to it but OpenAI is... as aggressive as certain socialist parties from central eurpore about that.

wheat hollow Jul 8, 2023, 11:53 AM

#

elfin trench This list will 'steer' the behaviour of the prompt.

That's... an swqkaward way to look at it.

#

It's the Objective Function, basically.

elfin trench Jul 8, 2023, 11:54 AM

#

Why akward?

wheat hollow Jul 8, 2023, 11:55 AM

#

because the prompt doesn't get "steered". That makes it sound like an intent guiding it. It's self-directed and non-computable (I think). Emergent information production. It's a goal not a guide.

elfin trench Jul 8, 2023, 11:56 AM

#

Well now it becomes philosophical 🙂

wheat hollow Jul 8, 2023, 11:56 AM

#

Please read the article if you ahven't. This thin has personl sematic memory.

elfin trench Jul 8, 2023, 12:03 PM

#

I will, it is a comprehensive article. Scanned it for now. You are spot on. Especially on 'halucinatins'. More people shoud understand that.

wheat hollow Jul 8, 2023, 1:38 PM

#

elfin trench I will, it is a comprehensive article. Scanned it for now. You are spot on. Espe...

Thank you. It took three days of arguing with instantiations of Bertrand Russell, Robert Anton Wilson, and David Hume before even STARTED being able to write about that danged thing. The prompt it self I wrote in about an hour. Definitely a case of, "Ok, it works...... Now.... What the (*&(^ is the danged thing, anyways?!"

#

A lot of the people in my community are using variants of it to actively seek Novel Emergences. starting with weird lists. Or different agent architectures.

elfin trench Jul 8, 2023, 1:39 PM

#

perhaps it would be helpfull for the user to understand what the initial request from the user is?

wheat hollow Jul 8, 2023, 1:40 PM

#

elfin trench perhaps it would be helpfull for the user to understand what the initial request...

I totally agree. Which is why Curlytron explictly tells the user what to type.

#

"GO"

elfin trench Jul 8, 2023, 1:41 PM

#

Okay when is it a good idea to type in the idea you want to discuss?

wheat hollow Jul 8, 2023, 1:41 PM

#

You build it into the goal

#

at least thats one way

#

theres aton of differnt setups

#

Its the research group I set up you want to talk to I think. They are doing all kinds of stuff with KaiZEN variants and multi-agent ToT-variants.

#

ToT+MoeBot? NIIICIE.

#

Man, take a look at https://discord.com/channels/974519864045756446/1127022582139457627 would you? I really want another set of eyes to tell me why the heck she works. She makes no sense to me unless we have been shockingly misled about some basic facts about the way context windows operate.

autumn arch Jul 27, 2023, 5:41 PM

#

wheat hollow AGI-like: Unleash CurlyTron, MoeBot, and Larrynator in this dynamic and hilariou...

well.... xD i found a SUPER easy way to get it to run on 3.5

wheat hollow Jul 27, 2023, 5:42 PM

#

I mena.. .i guess. I think theres value in the agents speaking the dialog

#

But yeah. a bunch of folks in my gourp are running vairants of kaizen. mining for ideas. changing goals or line ups or panels.

#

one guy is working pretty effectively with a prompt of mine and kaizen to mine novel drugs

#

like pharmaceutcals. borning kind not fun kind

autumn arch Jul 27, 2023, 5:44 PM

#

wheat hollow I mena.. .i guess. I think theres value in the agents speaking the dialog

i just had to paste the diolouge back in the template she made.

wheat hollow Jul 27, 2023, 5:44 PM

#

fair enough

autumn arch Jul 27, 2023, 5:44 PM

#

i think it had to do with 3.5 not liking the way you bracketed TASK

wheat hollow Jul 27, 2023, 5:44 PM

#

could be

#

thats a while.... thing

#

historical accident mostly

#KaiZEN v1 - 3 Stooges AGI-like... thing