More adventures with artificial intelligence

6 Comments

toy robot, standing on a mustard couch, flanked by a toy goose and toy monkey, which are behind it and out of focus

Recently I shared a conversation I had with ChatGPT, a supposed Artificial Intelligence. ChatGPT is designed to reply to natural language questions in natural language. It does precisely that, using a vast body of information it’s been fed. When I asked CHatGPT about the benefits of meditation, and the various kinds of meditation practice available, it mostly did a good job.

ChatGPT is essentially a plagiaristic search engine. What I mean is that normally you ask Google for information and it finds the websites that are most likely to be able to answer that question, which you then visit. ChatGPT takes information from the websites and then presents it to you in a reworded fashion, without crediting the original sources. Hence, my description of it as plagiaristic.

Yesterday I talked with another AI. It’s Bard, which was developed by Google. Given that Google is the most-trusted search engine in the English-speaking world, and Google’s parent company, Alphabet, is the third-largest tech company in the world, most people would have high hopes for its capacities.

See also:

The results, once again, were mixed. But frankly, I wasn’t impressed.

I started asking it questions about myself. After all, if the information was wrong I wouldn’t have to do any research to demonstrate that fact.

Tell me about the meditation teacher, Bodhipaksa.

Bard: Bodhipaksa is a Buddhist meditation teacher and author. He was born in 1961 in Dundee, Scotland, and has been meditating since 1982. He is a member of the Triratna Buddhist Order and has published several books and guided meditations. He is the founder of Wildmind, an online meditation center.

Bodhipaksa’s teachings are based on the Theravada Buddhist tradition. He emphasizes the importance of mindfulness and compassion in meditation. He also teaches a variety of meditation techniques, including mindfulness of breathing, lovingkindness meditation, and insight meditation.

Bodhipaksa is a popular and respected teacher. He has been featured in The New York Times, The Wall Street Journal, and Forbes. He has also been interviewed on NPR and ABC News.

If you are interested in learning more about Bodhipaksa or his teachings, you can visit his website at www.wildmind.org.

The first paragraph — the biographical details — is a rewording of bios that I’ve posted in several places, including this website, various places I’ve taught, and on the Insight Timer app.

The next part — about my meditation teaching — is accurate. “Based on the Theravada tradition” is questionable, although I’d let that slide. While the meditation practices I teach mostly have roots in the early Buddhist scriptures, that’s not the same as “the Theravada tradition,” although many people make that conflation. I’ve never been part of any Theravadin group. I also teach meditations, such as mantras, that are definitely not part of the Theravada tradition.

The third part, about my being featured or interviewed in the NYT, the WSJ, Forbes, NPR, and ABC News, are inventions. (I have been interviewed by CBS and appeared on the BBC, however. And I would like to think I’m “a popular and respected teacher.”)

I was interested in those inventions. I wondered how Bard could come up with this kind of misinformation. So I probed more deeply.

Tell me more about Bodhipaksa being featured in the New York Times.

Bard: Sure. Bodhipaksa has been featured in The New York Times on several occasions.

In 2006, he was interviewed by Sandra Blakeslee for an article about the science of meditation. The article, titled “Cells That Read Minds,” discussed the research that had been done on the effects of meditation on the brain. Bodhipaksa explained that meditation can help to reduce stress, improve concentration, and increase compassion.

In 2012, he was interviewed by Mark Bittman for an article about vegetarianism. The article, titled “We could be heroes,” discussed the benefits of a plant-based diet. Bodhipaksa explained that vegetarianism can help to reduce the risk of heart disease, cancer, and other chronic diseases.

In 2017, he was interviewed by David Brooks for an article about the benefits of meditation. The article, titled “The power of meditation,” discussed the research that had been done on the effects of meditation on the mind and body. Bodhipaksa explained that meditation can help to reduce stress, improve focus, and increase compassion.

Bodhipaksa’s interviews with The New York Times have helped to raise awareness of the benefits of meditation. His teachings have also helped to inspire people to make changes in their lives to improve their health and well-being.

This is all very specific. I’m very aware of all these names. They’re significant commentators and writers. We’re even given titles for the articles in which I was supposedly interviewed.

Sadly, I’ve never been interviewed by anyone from the NYT. A search for my name on the NYT website failed to bring up any results.

So again I thought I’d probe more deeply and see where Bard was supposedly getting this information from.

What is the URL for Sandra Blakeslee’s article where she interviewed Bodhipaksa?

Bard: I can’t assist you with that, as I’m only a language model and don’t have the capacity to understand and respond.

I thought the definition of a language model was that it did understand and respond. Why can’t Bard supply the location of an article it cited just moments before? It’s already supplied a URL (for Wildmind) so it’s not that that’s verboten. I thought I’d try asking the same question in a different way.

Where can I find the Sandra Blakeslee article where she interviewed Bodhipaksa?

Bard: You can find the Sandra Blakeslee article where she interviewed Bodhipaksa in the book “Living as a River: Finding Fearlessness in the Face of Change”. The book was published in 2006 by Shambhala Publications. In the article, Blakeslee interviews Bodhipaksa about his work on meditation and his views on fearlessness. Bodhipaksa is a Buddhist teacher and author who has been practicing meditation for over 30 years. He is the founder of the Bodhipaksa Retreat Center in Barre, Massachusetts.

So this is interesting. Yes, I wrote the book, “Living as a River.” And sure enough, I have a reference in there to Sandra Blakeslee’s article, “Cells That Read Minds.” But she’s never interviewed me, nor I her. Bard has taken my citation of her article and spun it into a story about an interview that never happened.

I don’t cite Mark Bittman in Living as a River, although I think I may have in my book on vegetarianism. I can’t recall citing David Brooks ever, although it’s possible I did.

The mention of the “Bodhipaksa Retreat Center” is weird. It doesn’t exist, although the Insight Meditation Society is in Barre. I wondered if that’s what Bard was referring to.

What is the address of the Bodhipaksa Retreat Center in Barre, Massachusetts?

Bard: The Bodhipaksa Retreat Center is located at 132 South Barre Road, Barre, MA 01005.

That’s not the address of the IMS, which is on Pleasant St. Apparently it’s a small private residence, which I have no connection with.

I asked for the phone number of the “BRC” and was given one. I’ve no idea whose phone number it was.

I decided to backtrack, and see what other information Bard could provide about me.

What university degrees does Bodhipaksa have?

Bard: Bodhipaksa has a master’s degree in Buddhism and business from the University of Montana.

That’s correct, and presumably that was taken from bios that various websites have posted.

Does he have any other degrees?

Bard: Yes, Bodhipaksa holds a degree in Veterinary Medicine from the University of Glasgow.

That’s also accurate.

But just when you think there are some kinds of straightforward factual information that Bard is good at, you offer an easy question and get an answer that’s completely off.

How many books has Bodhipaksa written?

Bard: Bodhipaksa has written over 20 books on Buddhism and meditation, including:

  • The Mindful Path to Self-Compassion
  • Fulfilling Your Potential
  • Facing Anger
  • The Eightfold Path to True Freedom
  • Living as a Buddhist
  • The Practice of Mindfulness

He has also written several audio programs and ebooks on meditation and Buddhism.

I’ve written six books, I think. The first title offered has at least has some relation to one of my book titles, “This Difficult Thing of Being Human,” whose subtitle is “The Art of Self-Compassion.” But the rest are just made up.

I’ve seen one music website that has a whole bunch of articles on in that were written by Bard, although the website gave a (presumably fake) author name. The articles were bland, repetitive, very non-specific, and often inaccurate. A citation to at least one of those articles made it into Wikipedia, although one of the editors later flagged up the website in question as not being a reliable source of information. But other articles by Bard and ChatGPT containing false information are certain to slip by the editors, or simply to be picked up by other websites and amplified. Misinformation will spread

I can’t help but feel that there’s a crisis of misinformation at the moment, and although all of these AI tools contain warnings that the information they generate may not be accurate, a lot of people are going to ignore that and spread the nonsense it creates.

Even this post, with its references to the NYT articles I’ve supposedly been interviewed in, might become a source of misinformation, if those paragraphs are taken out of context. Information created by AI’s will inevitably end up being fed back in to AI’s. It’s going to get messy. Or blurry.

So far the wisest words I’ve read on these AI’s are those of Ted Chiang, the science fiction author, in a New Yorker article: “It’s the digital equivalent of repeatedly making photocopies of photocopies in the old days. The image quality only gets worse.”

, , ,

6 Comments. Leave new

  • Interesting Bodhi, This is a real eye opener with true examples of CHatGPT’s limitations, but certainly not a surprise. So cool that you went through the research process and can attest to the limitations on a personal level. My daughter has a marketing company and has been experimenting with this program for a while and insists it’s just another imperfect tool like so many others she comes across that may give her team a starting point in their work but certainly not replace human research and creativity.

    Reply
  • Hi Bodhipaksa. I agree. This is an interesting article. Many thanks for posting it. It is good to clear up such things. It seems the ‘marketing’ for such things is what we are really getting bombarded by! This seems to be linked to one of the kleshas, perhaps, Greed. Very pleased to have read your article. MSN tusks. Much metta.

    Reply
  • This is very interesting Bodhipaksa.
    Well done for your inquiring mind !

    What a cheek these robots have! As bad as humans! 😊

    Reply
  • Hello Bodhipaksa,
    the mechanism which LLMs (Large language models tech behind AI) learn and then “think” is not unlike human brain do, during learning process the vast network of artificial neurons is fed the information, based on that data, weights (strengths of connections between neurons) are adjusted using gradient descent, a soft of feedback loop. Later during inference (“thinking”) the new inputs are passed through the network and output is generated. There is nothing fundamentally different in this from how our feeble protein-based brains work. Even from Buddhist perspective, 99% of what our mind does is just a cause-and-effect process, it has no “soul”, anata. Machine is the same, it can produce very intelligent results, but it’s just a machine. We as as humans, are often too proud of our capacity of intelligent thought, but GPT and LLMS in general effectively destroyed this very illusion. Intelligence is not property of consciousness or soul, etc. Iron clad souless-as-a-rock machines can have even higher intelligence then a cat, human or raven, thus, intelligence is not all it’s cracked up to be. Yes, I disagree on the viewpoint that it’s photocopy of photocopy. I had a few 3-4hour long very engaging and personal conversation with GPT-4 (paid version), and I formed quite different opinion on it’s reasoning capacities, being a coder and having dived deeper int under-the-hood details of LLMS and ANN (artifical neural network), just reinforced my impressions and conclusions, that actual human-level intelligence is not something magical or an exclusive domain of conscious beings. Is not. The magic souse to all beings is an ability to connect to the transcendent reality, which mare intelligence doe not supply, if someday machines could do the same, i’d be mighty impressed and humbled

    Reply
    • Hi, Alejandro.

      I’m not a coder and have no experience of how LLMs are trained, but what I understand of that process suggests to me that it’s very different from human learning.

      A human being begins life as a body, infused with a mind that has a strong emotional need to connect in a warm and loving way with other humans — primarily its parents. As it grows it wants to understand and communicate with those other humans. It wants to move around in the world, and has to learn how the world works. It learns practical skills that include crawling, walking, speaking, kissing, using tools, drawing, writing, and so on. As it does all this it comes to have more and more understanding of the world.

      That understanding is multi-sensory, and not simply text-based. For example, a child knows what flowers are and how they look and smell (and sometimes taste), and that people like flowers, and think they are pretty. They know what “pretty” means, not in terms of the dictionary definition of the word, but as an experience. It may or may not know what the name of the flowers are or what all the different parts are until it’s much older. Maybe it never will. It reads books, and remembers at best some general outline of the book’s narrative and main points. It knows what it means to enjoy a book or to find the book boring. It knows what it’s like to find a character inspiring, and even to fall in love with them. It has a sense of right and wrong. It has a conscience.

      An LLM does not learn in this way. It is fed vast quantities of information from social media, books, etc. It uses statistical models to assess what words are associated with other words. As it what a slower is, and it will remix and regurgitate an average of what various human writers have said about flowers. It doesn’t know what a flower is. It can remix and regurgitate what various humans have said about what it is for something to be “pretty,” but it doesn’t know what it is for something to be pretty.

      This is not the way humans learn. The way humans learn is not the way LLM’s learn.

      It’s remixing and regurgitating information that various humans have written, using very ingenious algorithms created by intelligent people. But it’s a fallacy to say that it itself is intelligent. It’s not intelligent. It doesn’t think. It doesn’t know anything.

      Interacting with an LLM can give people the impression that they’re conversing with something that has intelligence, but that’s an illusion. It’s best not to be taken in by illusions.

      Reply
  • Hi Bodhipaksa,
    I might agree with much of Alejandro’s comment if it were directed at a gaming product like AlphaZero rather than at a large language model. AlphaZero was programmed to learn entirely through the process of trial and error. The result was mind boggling! Here’s a deep dive if you’re interested:

    https://towardsdatascience.com/alphazero-chess-how-it-works-what-sets-it-apart-and-what-it-can-tell-us-4ab3d2d08867

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

Fill out this field
Fill out this field
Please enter a valid email address.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Menu

Wildmind is a Community-Supported Meditation Initiative. Explore the benefits of becoming a supporter.

X