planetdance and AI

[Jean]

Which version of claude are you using, the freeware?
This is what i got when i asked chatgpt 4o about the conjunction, i was a bit surprised that i gave me this, i had not specifically asked to give a comparison of two persons.
I do like what it did here though the problem remains, it's difficult to get beyond the stage of cookbooks, but it can do it.
But I have some modus operandi now, what it does is good enough to give me a basis for an interpretation along with my own thoughts, if specifically ask for weak / strong points / advice.

Two individuals, Alex and Jordan, share a unique astrological connection: both have their Moon and a Uranus-Neptune conjunction in Capricorn. This combination weaves a tapestry of shared traits, desires, and life experiences that set them apart in their approach to the world.

### Alex

Alex is a person who feels deeply yet expresses their emotions in a controlled and calculated way. With the Moon in Capricorn, Alex values stability, responsibility, and pragmatism in their emotional world. However, the presence of Uranus and Neptune conjunct in the same sign adds a layer of complexity to Alex's nature. There’s an underlying current of restlessness and a yearning for something beyond the conventional. Alex often finds themselves dreaming of a world where structure meets imagination—where the boundaries of tradition can be expanded to include innovation and intuition.

In their daily life, Alex is the kind of person who seeks to blend the old with the new. They may have a career in a field that combines technology with a sense of purpose, such as sustainable architecture or social innovation. Alex feels a strong sense of duty to contribute to society in meaningful ways, often grappling with the tension between their need for security and a desire to break free from societal expectations. Their approach to relationships is similar: Alex values loyalty and consistency, but they also crave space and freedom to explore their unique interests and spiritual pursuits.

### Jordan

Jordan, too, is marked by the Capricorn Moon's need for structure and control, but the influence of Uranus and Neptune brings an ethereal and rebellious edge. Jordan is deeply introspective and often feels pulled between the tangible demands of reality and the pull of their vivid inner world. They are drawn to unconventional paths—whether in their career, relationships, or lifestyle. There’s an innate drive to reform and reinvent, to not just accept things as they are but to imagine how they could be better.

Jordan’s Uranus-Neptune conjunction fuels a visionary outlook; they are someone who sees potential and possibilities where others see limits. This conjunction, tied with the Moon, makes Jordan highly intuitive, often picking up on unspoken undercurrents in their surroundings. However, this also means they can struggle with the harshness of reality, sometimes feeling overwhelmed by the weight of their own expectations or the societal structures they wish to change.

In relationships, Jordan craves a partner who can understand their complex emotional landscape and who respects their need for both intimacy and independence. They often seek connections that are more than just surface-level; they want to build something lasting and meaningful but on their own terms, often challenging traditional roles and expectations.

### Shared Dynamics

Despite their differences, Alex and Jordan find a deep resonance in each other’s company. They both share a drive to make an impact, to build something enduring that aligns with their ideals and visions. Their Capricorn Moon gives them a shared language of resilience and practicality, while their Uranus-Neptune conjunction inspires a mutual love for innovation, creativity, and spiritual exploration. Together, they might collaborate on projects that marry their need for structure with their desire for change—perhaps founding a start-up that promotes ethical technology or creating a community space that fosters both artistic expression and social responsibility.

Their shared astrological placements mean they understand the challenges of navigating a world that often feels too rigid or too chaotic. They bond over their experiences of being misunderstood or feeling out of sync with the traditional expectations of their peers. However, this common ground also gives them a unique strength: together, they are a force of both stability and transformation, capable of dreaming big while keeping their feet firmly on the ground. Whether as friends, partners, or collaborators, Alex and Jordan are united by a cosmic alignment that calls them to be architects of a new kind of reality—one that honors the past but isn’t afraid to rewrite the rules.

[Vaclav Spirhanzl]

In answer to your question about AI versions, I use the free versions of Chat-GPT and Claude. In the case of Chat-GPT, I use the latest version, which I believe is 4.x. Once it reaches its limit, it switches to 3.5. Claude has version 3.5 Sonnet, which is the same for the free and paid versions. Once it reaches its limit, it usually blocks me for around five hours.

The difference between Chat-GPT and Claude is in the usage of context and the opportunity and ability to learn from actual communication with me.
While Chat-GPT has a small context for one chat, about 64k tokens (one token is not one human word) on the one hand, on the other hand it is now able to share these contexts with multiple chats of one user. I found this information in a discussion on the Net about the latest version of Chat-GPT. Chat-GPT is able to learn as it communicates with me, i.e. after the chat is over, the information from our communication is further evaluated and incorporated into the global knowledge base.

In contrast, Claude is static, each chat starts on an empty table (tabula rasa), Claude has a rather large memory for each chat, about the size of a 500 page book. This has its limitations, for each question within a chat he has to re-read the whole chat from the beginning ( by the way this counts towards the limit in all models - free or paid). When the chat is finished, all communication is deleted. Claude has a much better language module compared to Chat-GPT.

Neither AI has astrological knowledge, only astrological texts of varying quality. The texts are mostly from Anglo-Saxon backgrounds, articles, books, etc. European astrological texts are very slow to adopt or not at all.

This is what I've experienced with AI over the past three years.

[Vaclav Spirhanzl]

About the question about Alex and Jordan. It depends on how you phrase the question. If I specify that they have a conjunction in Capricorn, these three planets (and nothing more, no other planets nor houses at all), then I should get the answer that they are both the same (astrologically), only the position in houses and other planets can generate different texts, this is how I got the note from Claude and it matches my astrological experience. Or have I made a mistake somewhere? Maybe I misunderstood the task.

[Jean]

I just gave those texts as an example. I don't understand your question.
I know about the memory chatgpt has.
I also found this https://www.whatplugin.ai/blog/the-4-best-gpts-for-astrology but i did not like it.
Also experimented with local AI, with LM Studio, but too simple, not as powerful as online. Edit: now i'm seeing things i like, still playing though.. It's nice that you can try all kinds of models.

Edit it'd funny how these things work, they don't interpret the prompt the same between sessions. The prompt 'Give interactions, strengths, challenges and advice.' works the first time by giving a resume at the end, the second time it gave strengths and challenges after each 'planet'. I've had this before and it's very frustrating, a known problem, it's hard to get consistent results. Therefore it's also a nuisance to test stuff.

I've now played with a few local models through lm studio.
ps, i don't think you can get chatgpt 4o in the freeware version.
Chatgpt itself says you can't, but duckduckgo has a prompt for it https://duckduckgo.com/?q=DuckDuckGo+AI+Chat&ia=chat&duckai=1&atb=v353-1 so i guess you can use it after all.

I'm just playing around... There's an overflow of info for me to handle.

[AB]

Hi everyone, I have a few insights you might be interested in since I used the paid version of gpt, all of them, to test how MLM 'learn', how they generate replies etc for one of my jobs.

First, the AI designation is a misnomer, a manipulative marketing misnomer that reinforces people's misperception of what constitute machine 'learning'.
That being said, when it comes to astrology, and pretty much every other subject, MLM work based on the curated literature they are fed, and it being curated is an important aspect, so the quality of the literature and the censorship of the same are key to what the Model will generate. This when it comes to astrology, and many other subjects like psychology, sociology, much of political science, etc is very very bad news. Since astrology is filled with junk as a subject. You have to understand that the feed in process of 'literature' includes any and all 'articles', blog posts, etc any self described expert has ever written on the subject. And almost all of the literature it was fed is 'modern astrology', some of it is 'vedic' astrology, and very very little, statistically almost insignificant, is the higher end of the literature or higher end commentary on the subject.

Since the Model uses statistical models to generate replies, it can't generate a higher quality than what the vast majority of 'literature' is on the subject. This is why it will be almost useless if you try to use it for Hellenistic astrology. It will always go back to 'modern', 'psychological' and 'esoteric' astrology, with some 'vedic' components here and there.

MLM also have an other weakness that most people don't know about or if they do they don't know the incredible impact it has. They are 'regulated' and heavily censored with what their developers call safety and 'ethical' guidelines, but first, these guidelines are terribly paternalistic and patronizing, even in the paid version, and they are seemingly meant for people that are either 6 years old or suffer from serious psychiatric disorders and thus need to be protected, and second, these guidelines interfere with the 'reasoning' capacity of the Model. The Model will ignore any and all information that triggers his warning system, and it will suppress, alter, etc any input that its guidelines react to, this results in 'cognitive' paradoxes where the generated reply is factually wrong simply because the guidelines are there to 'protect' you from facts. So, for example, it will not engage with Hellenistic texts (even if you try to create your own Hellenistic astrology gpt by uploading Hellenistic era text files and your own guidelines into a 'separate' bot) because its guidelines do not find them PC enough.

It will only generate, as a rule, only modern astrology platitudes, heavy on untestable general 'maybe, perhaps, sometimes' etc. and if by any chance you force it to abide by your rules by adapting your rules so they don't trigger its warning system, then it will generate what you feed it, but it won't be able to 'reason' beyond it.

So no, you can't use it, and I highly doubt that you ever will be able to use it, to 'read' a chart, beyond the limits I described, for not only will these limitations persist, they are getting worse and worse with each new generation. In order for a MLM to be able to do what I guess many here would like you would have to have a model that you yourself train, you curate the literature, you set all the guidelines, etc and you have access to the higher end computation capacity which even the paying models don't give you.

[Jean]

I use the payed gpt version too.

I agree on the hellenistic side, indeed, the books they feed it with are mostly about western astrology, however i did several horoscopes and i like what it does, it's good enough to use as a tool to use when doing an interpretation, it can give a fine basic summary. It does not only give the info from books / texts, it also generates summaries where it combines things, i do like it yes and other astrologers i've shown it to like it also.

On programming code it works very well. Sometimes you have to spend a lot of time to get the things you want, but other times i got good working code in a minute which would have cost my an hour to write myself, so no, i don't share your negative opinion there.
For me it's great, but you have to know it's limits.

[ABer]

Thanks for the interesting feedback AB. It explains why my attempts to get a Hellenistic slant came to nothing. Ed

[AB]

Hi Jean,
I understand if someone finds these MLM useful, my intention was not to dissuade you from using them. I'm just sharing my experience and insight.

Hi Ed,
yes, that's why I wanted to share this experience. It's part of one of my job to test how these MLM are doing in actual fact, beyond their design made to create the impression of an 'intelligence', and all I can say is that we are light years away from both the fears of some and even further from the hopes of others.

Take care,

A.

[Vaclav Spirhanzl]

Hi AB,
Could you please clarify what is meant by the acronym MLM? I am wondering if it stands for "Masked Language Model." Thank you in advance. 

[Vaclav Spirhanzl]

Hi Jean,

I fully agree with your view on AI and thank you for your last sentence - For me it's great, but you have to know it's limits.

Hi all,

I work with AI partners (Chat-GPT and Claude) in the following way:

- I prepare astronomical and astrological information plus text for the interpretation, most of the texts from my own astrological library (in case of interpretation)
- I will pass this information to the AI, along with a specification of what kind of document we are to produce (interpretation, analysis, basic summary), where I expect to expand it, etc., who it is for, and what the language style should look like.
- After a few iterations, the final document is done.

I've found that the text generated from Planetdance is a significant time saver and a valuable addition to my workflow. In my experience, generating text as an assignment for AI has proven to be an effective approach.

I use AI partners and tools primarily as language modules, that's what they always say about themselves - "that they are just a language module". I use Chat-GPT (paid version) and Claude - free version (free version has five times less resources and no access to programming API, otherwise the models are identical).

[AB]

Hi Vaclav,
MLM stands for Machine Learning Model, but I could have used the LLM which stands for Large Language Model, both are far more correct terms than the misnomer Artificial Intelligence or AI which is nothing but a marketing tool, similar to the extremely misleading but in marketing terms very effective term Cloud.
Even saying that they 'learn' can create a very fallacious impression.
Hope this helps.

A.

[Jean]

I was scripting a bit, something to count so called blue moons where there's a full moon twice in a month and decided to ask gpt. Couldnt formulate a good prompt but then i showed it the ab code i already had and asked it if it understood it, it said yes and began explaining the script to me. Then i asked for my blue moons function and it showed me code that was a mix of modern c and ab. Loved it. In the future it will be able to understand / learn astrobasic.

[ABer]

Wow Jean. Incredible breakthrough. This is getting interesting! Ed

[Vol2]

I've read that coders have saved themselves a lot of work by using AI.
If you manage to get your AI to learn Astrobasic, is it possible for it to look at the code for the 3D rotating chart sphere from Astrolog and translate the code into Astrobasic or adapt it accordingly?
The code is freely available, I have already looked at it, but I have put it on file due to its complexity and my limited skills.
But now I see a new glimmer of hope after your post.
Volker

[Jean]

@Volker yes i know that code, too much for me also.
And maybe it could be done in ab, but i'd prefer to have some commands on the c backend for it, for speed, some new ab commands. But so far it's only a dream.