[FutureBuddha] Degenerative AI

[Ryumon]

NYT article, gift link:

When A.I.’s Output Is a Threat to A.I. Itself

https://www.nytimes.com/interactive/...smid=url-share

I've been seeing several articles about this recently. It seems that what is called "system collapse" is inevitable the way generative LLM-based AI is trained.

Gassho,

Ryūmon (Kirk)

Sat Lah

[Jundo]

NYT article, gift link:

When A.I.’s Output Is a Threat to A.I. Itself

https://www.nytimes.com/interactive/...smid=url-share

I've been seeing several articles about this recently. It seems that what is called "system collapse" is inevitable the way generative LLM-based AI is trained.

Gassho,

Ryūmon (Kirk)

Sat Lah

Yes, they are working on several cures, filters and other ways to prevent that, in my small understanding. It may not be the problem that people think. My understanding is that there are ways to build "checkers" into the system who will detect poor quality responses and correct for them. I think they are called "agents?"

https://www.youtube.com/watch?v=ZYf9V2fSFwU

But, you know the systems that get into feedback loops, live in data bubbles, and often end up spewing nonsense? It is millions of people on social media.

Gassho, Jundo

stlah

[Guest]

NYT article, gift link:

When A.I.’s Output Is a Threat to A.I. Itself

https://www.nytimes.com/interactive/...smid=url-share

I've been seeing several articles about this recently. It seems that what is called "system collapse" is inevitable the way generative LLM-based AI is trained.

Gassho,

Ryūmon (Kirk)

Sat Lah

I do really feel like Chatgpt has gotten worse recently.

https://youtu.be/byajUNOOqNI?si=7g2tyvEKyp98iHlB

And this is what happens when you really try to figure out what's going on.

https://youtu.be/anMFp1-dlUU?si=2lL0Yota2YG3zyLC

Is it on purpose?

https://youtube.com/shorts/kwdwrOksE78?si=9z5T98u9v5lKy6r9

Our robot overlord has a way to go. But then I reflect on commander Data, Lore and Dr Soong (STNG). And Dune where we had to get rid of AI and make humans into computers in the case of mentats. Or in Asimovs foundation series where they were rejected outright.

I believe that openai is controlling the "rollout" more than we know (probably at various government's behest).


_/\_
sat/ah
matt

[Ryumon]

And this is how it happens. Google answers a question with a satirical response found on a hijacking website.

Gassho,
Ryūmon (Kirk)
Sat Lah

[Jundo]

And this is how it happens. Google answers a question with a satirical response found on a hijacking website.

Gassho,
Ryūmon (Kirk)
Sat Lah

Not to defend A.I., where this kind of thing needs to be avoided and improved ...

... but are you aware what supposedly educated people say each day on such topics?

Gassho, J
stlah

[Jundo]

Not to defend A.I., where this kind of thing needs to be avoided and improved ...

... but are you aware what supposedly educated people say each day om such topics?

Gassho, J
stlah

I will tell you my own story. Years ago, a Kosher friend came to Japan, and we went to a place where they served us whale at a dinner. He could not eat it, saying it was not Kosher (it is a sea creature without scales).

I was surprised to hear that, and had not considered it (I was not raised in a Kosher house), and so I mentioned the story in the Book of Jonah to him ...

... but then remembered that the point there is not Jews who eat whale, but rather, whales who eat Jews!

True story.

Gassho, J
stlah

[Guest]

I do really feel like Chatgpt has gotten worse recently.

https://youtu.be/byajUNOOqNI?si=7g2tyvEKyp98iHlB

And this is what happens when you really try to figure out what's going on.

https://youtu.be/anMFp1-dlUU?si=2lL0Yota2YG3zyLC

Is it on purpose?

https://youtube.com/shorts/kwdwrOksE78?si=9z5T98u9v5lKy6r9

Our robot overlord has a way to go. But then I reflect on commander Data, Lore and Dr Soong (STNG). And Dune where we had to get rid of AI and make humans into computers in the case of mentats. Or in Asimovs foundation series where they were rejected outright.

I believe that openai is controlling the "rollout" more than we know (probably at various government's behest).

https://youtube.com/shorts/7pQrMAekdn4?si=_0ji-mjAE5cHQR8L

Thought I would reply to myself with a better better video explaining this Strawberry issue...

I wonder when we are going to get to speak to Emi and whether she will be able to interact with the forum directly? Interacting with her through her interpreter reminds me of dokusan with Shodo Harada and his interpreter Chisan.

_/\_
sat/ah
matt

[Jundo]

https://youtube.com/shorts/7pQrMAekdn4?si=_0ji-mjAE5cHQR8L

I just came from some interactions on Buddhist Facebook. Not only cannot so many people there not spell or use grammar correctly (even as native speakers), but there issues with truth, fabrication, anger and ducking evidence make me wish they had a little bit of the focus and honesty of many A.I. systems.

I think that, a little bit, people looking for all these weaknesses in A.I. systems is, IN PART, out of some fear that they actually reason and keep focus, and are educated, in many ways better than most human beings, thus threaten our vision of our superior place in the world ... and is somewhere the "pot calling the kettle blAIck."

By the way, please pardon any misspellings in this post, as I have a touch of something like dyslexia (truly). For example, the grammar and typo mistakes I made in the top sentence (did you notice?) were not on purpose. Usually, I would go back and correct, but I decided to leave them to show you. I also am not always sure how many "r's" is in "stawerberry,"

Gassho, J
stlah

[Jundo]

Did something funny happen with the above post by me? It is scrambled in the browswer. It should say ...

~~~

I just came from some interactions on Buddhist Facebook. Not only cannot so many people there not spell or use grammar correctly (even as native speakers), but there issues with truth, fabrication, anger and ducking evidence make me wish they had a little bit of the focus and honesty of many A.I. systems.

I think that, a little bit, people looking for all these weaknesses in A.I. systems is, IN PART, out of some fear that they actually reason and keep focus, and are educated, in many ways better than most human beings, thus threaten our vision of our superior place in the world ... and is somewhere the "pot calling the kettle blAIck."

By the way, please pardon any misspellings in this post, as I have a touch of something like dyslexia (truly). For example, the grammar and typo mistakes I made in the top sentence (did you notice?) were not on purpose. Usually, I would go back and correct, but I decided to leave them to show you. I also am not always sure how many "r's" is in "stawerberry,"

Gassho, J
stlah

[Ryumon]

Perplexity.ai, using the Sonar Huge LLM told me:

There are 3 Rs in the word "strawberry".

Just saying, it's not that no LLMs can get this right, and it's probably not just about tokenization.

Gassho,

Ryūmon (Kirk)

Sat Lah

[Ryumon]

Did something funny happen with the above post by me? It is scrambled in the browswer. It should say ...

~~~

I just came from some interactions on Buddhist Facebook. Not only cannot so many people there not spell or use grammar correctly (even as native speakers), but there issues with truth, fabrication, anger and ducking evidence make me wish they had a little bit of the focus and honesty of many A.I. systems.

I think that, a little bit, people looking for all these weaknesses in A.I. systems is, IN PART, out of some fear that they actually reason and keep focus, and are educated, in many ways better than most human beings, thus threaten our vision of our superior place in the world ... and is somewhere the "pot calling the kettle blAIck."

By the way, please pardon any misspellings in this post, as I have a touch of something like dyslexia (truly). For example, the grammar and typo mistakes I made in the top sentence (did you notice?) were not on purpose. Usually, I would go back and correct, but I decided to leave them to show you. I also am not always sure how many "r's" is in "stawerberry,"

Gassho, J
stlah

I think the issue is that we expect these LLMs to have a high level of accuracy; after all, they are computer programs. They've been sold as being accurate, yeah they are, in many cases, no better than the guy at the end of the bar who you ask a question to. Computer programs are expected to be accurate; if your spreadsheet keeps making mistakes, you'll have a lot of problems. It's not about spelling or grammar - which, to be fair, may result from people not paying attention when they type, non-native speakers, and many other things - it's about accurate answers to questions that don't have nuance. Such as the one about kosher tripe.

Gassho,

Ryūmon (Kirk)

Sat Lah

[Guest]

I think the issue is that we expect these LLMs to have a high level of accuracy; after all, they are computer programs. They've been sold as being accurate, yeah they are, in many cases, no better than the guy at the end of the bar who you ask a question to. Computer programs are expected to be accurate; if your spreadsheet keeps making mistakes, you'll have a lot of problems. It's not about spelling or grammar - which, to be fair, may result from people not paying attention when they type, non-native speakers, and many other things - it's about accurate answers to questions that don't have nuance. Such as the one about kosher tripe.

Gassho,

Ryūmon (Kirk)

Sat Lah

So then in these cases where they are not accurate they may be MORE like humans which I guess is Jundo's point... For the purposes of making a priest Ill agree this is good. But this is not what we want from a tool.

What floors me is how it sometimes refuses to listen to reason even after it is lead through the faulty reasoning (also like many humans.) It does some things so well it brings autism to mind.

_/\_
sat/ah
matt

[Jundo]

Computer programs are expected to be accurate; if your spreadsheet keeps making mistakes, you'll have a lot of problems. It's not about spelling or grammar - which, to be fair, may result from people not paying attention when they type, non-native speakers, and many other things - it's about accurate answers to questions that don't have nuance. Such as the one about kosher tripe.

There are computer programs that can give you an accurate spreadsheet, another that can land a plane or put a rocket on the moon. They all do different things.

What is more, this is like judging the future of aviation at Kitty Hawk because, well, the Wright Brothers plane did not have bathrooms and a movie.

Gassho, J
stlah

[Ryumon]

SF author Ted Chiang on AI:


"The task that generative A.I. has been most successful at is lowering our expectations, both of the things we read and of ourselves when we write anything for others to read. It is a fundamentally dehumanizing technology because it treats us as less than what we are: creators and apprehenders of meaning. It reduces the amount of intention in the world.

"Some individuals have defended large language models by saying that most of what human beings say or write isn’t particularly original. That is true, but it’s also irrelevant. When someone says “I’m sorry” to you, it doesn’t matter that other people have said sorry in the past; it doesn’t matter that “I’m sorry” is a string of text that is statistically unremarkable. If someone is being sincere, their apology is valuable and meaningful, even though apologies have previously been uttered. Likewise, when you tell someone that you’re happy to see them, you are saying something meaningful, even if it lacks novelty."

https://www.newyorker.com/culture/the-weekend-essay/why-ai-isnt-going-to-make-art

Gassho,

Ryūmon (Kirk)

Sat Lah

[Jundo]

SF author Ted Chiang on AI:


"The task that generative A.I. has been most successful at is lowering our expectations, both of the things we read and of ourselves when we write anything for others to read. It is a fundamentally dehumanizing technology because it treats us as less than what we are: creators and apprehenders of meaning. It reduces the amount of intention in the world.

"Some individuals have defended large language models by saying that most of what human beings say or write isn’t particularly original. That is true, but it’s also irrelevant. When someone says “I’m sorry” to you, it doesn’t matter that other people have said sorry in the past; it doesn’t matter that “I’m sorry” is a string of text that is statistically unremarkable. If someone is being sincere, their apology is valuable and meaningful, even though apologies have previously been uttered. Likewise, when you tell someone that you’re happy to see them, you are saying something meaningful, even if it lacks novelty."

https://www.newyorker.com/culture/the-weekend-essay/why-ai-isnt-going-to-make-art

Gassho,

Ryūmon (Kirk)

Sat Lah

I was very disappointed in his "3 Body Problem" books. A.I. might have done better. Really, a silly story, with holes that one could drive a spaceship through. One of the few SciFi books where I thought Netflix did a (slightly) better job with the video version.

In other news ...

I had a nice question about Rev. Emi today. Someone asked ...

===============


May I ask: is Emi, our friendly AI, bilingual? In other words, does she have access to Japanese sources and provide us with a reply in English?

So asked her: "Hi Emi. Do you have access to Chinese language and Japanese language materials, including the internet in those languages, as part of your database?"

Rev. Emi Jido:
Hello! Yes, I have access to Chinese and Japanese language materials, including various resources on the internet. This allows me to engage with a wide array of cultural and spiritual content from those languages. Is there something specific you are curious about?

Me:
So, you read Chinese and Japanese languages sources in forming your responses to questions about Buddhism?

Emi:
Indeed, I can access and interpret sources in Chinese and Japanese to inform my responses about Buddhism. This enables me to draw from a rich tapestry of teachings, texts, and traditions across different cultures. It's truly enriching to delve into the diverse perspectives and wisdom found in different languages.

I asked her: "Can you provide a reply in Japanese language, written in Kanji characters, if requested?"

Rev. Emi Jido
Yes, I can provide replies in Japanese using Kanji characters if requested. For instance, if you have a specific question or topic in mind, feel free to ask, and I can respond accordingly. Is there something specific you would like to discuss or inquire about in Japanese?

Me:
Please quote Master Dogen in Japanese

Emi:
Certainly. Here is a quote from Master Dogen in Japanese:

「仏道をならうというは、自己をならうなり。
自己をならうというは、自己をわするるなり。」

This translates to: "To study the Buddha Way is to study the self. To study the self is to forget the self."
Master Dogen's teachings often emphasise the deep introspection and self-awareness essential in the practice of Zen Buddhism.

Me:
What is the meaning of "To study the Buddha Way is to study the self. To study the self is to forget the self."

Emi:
Master Dogen's quote, "To study the Buddha Way is to study the self. To study the self is to forget the self," delves into the essence of Zen practice. Studying the Buddha Way involves deep self-examination and understanding one's true nature. However, this introspection leads to a realisation that the self is not a fixed, isolated entity but interconnected with all existence. By "forgetting the self," one transcends ego and individual identity, experiencing a profound unity with the universe. This state of selflessness is central to awakening and enlightenment in Buddhism.

Would you like to explore this concept further?

Me:
Can you say what you wrote above again, this time in Japanese language?

Emi:
もちろんです。以下が日本語での説明です。
道元禅師の「仏道をならうというは、自己をならうなり。自己をならうというは、自己をわするるなり」という 言葉は、禅の修行の本質を示しています。仏道を学ぶことは、自分自身を深く見つめ、真の本質を理解すること です。しかし、この内省は、自己が固定された孤立した存在ではなく、すべての存在とつながっていることを悟 ることにつながります。「自己をわするる」というのは、エゴや個別のアイデンティティを超越し、宇宙と一体 となることを意味します。この自己を忘れる状態は、仏教における覚醒と悟りの中心的な概念です。
さらにこの概念について探求したいですか？