Part 2: A Technosophical Conversation About AI — Living Between Worlds 6.4

Foreign. Welcome, everybody. This is Living between Worlds for April 16, 2025. I'm Gil Friend. I'm CEO of Natural Logic and an executive coach to world changers. Ken Homer, our usual co host, is not with us today. He's on travel. But fortunately Pete Kaminsky is with us today. Continuing from last time. Last month we had the first part of what we call the technosophical conversation about artificial intelligence, or let's call it AI, because as we said, AI may mean to some people, artificial intelligence, augmented intelligence, assisted intelligence. Shane Bird calls it simulated intelligence. Last time we focused more on the techno part of the technosophical. Some of you may have had a chance to Explore the what now. 20 sessions of homework that Pete generously posted on Substack, A great exploration of how to orient and play with these tools. Today we're going to go into the more sophical part, which we weren't able to explore very much last time. But before we do, just a quick poll. If I ask you about your orientation to AI. Are you exploring? Are you fleeing? Are you both? Are you other show of hands, how many are exploring and drawn in some kind of way? Mostly. How many are fleeing? Let me add here. Don't want any part of this. Okay, how many are both? I'd expect a little more of that. How about ignoring that? Okay, I would put that in the fleeing category, but sure. So my fourth question was other. What else? How else are you orienting this thing? Exploring, Getting engaged. Exploring with caution. Yeah, Celia, thank you. Learning. Anybody else? Okay, let's start there. And you know, actually let me do one thing before going to where my notes are going to take us. So what are you finding in your exploration, both specifically in terms of what you're able to do, but also how you are shifting in yourself as you play with this stuff? Shell, let's start with you. Okay. For starters, I have a healthy skepticism about AI's competence in certain situations. And specifically, Chat GPT is the student in class who is a really good bullshitter and sounds authoritative and is full of crap. So I don't use Chat GPT. So I have a higher trust in an instrument called Perplexity AI, largely because it gives me a footnote that I can click on to the original source, and also because it tends to be better at interpolating nuance and context than ChatGPT is. I'm also aware that there's tens of thousands of AI tools out there. And as someone who makes my living in part as a writer and speaker, I Am guessing that we are going to see in the next couple of years a huge exponential uptick in the number of people willing to settle for the 80% solution that they can get in seconds and for free or practically free from a machine, versus the much smaller percentage of people who will actually pay for quality. But this is nothing new. There have always been a lot more people who would buy a Yugo than who would buy a Rolls Royce. There are also a lot more people out there who can create the Yugo on the fly. And that's something that is new and different. Although it echoes the early days of desktop publishing. Yes, very much. All of a sudden I could design my own brochures and I could have 15 fonts on the page until I finally realized, wait a minute, that's actually not a good idea. Yeah. And then. And there's a skill that my designer has that's different than the skills that I have. And there's place for both of them. Yeah. And the designers, meanwhile, all jumped on the digital bandwagon, started using the tools in the maximum benefit way. Yeah. On the 8020 thing, it reminds me of an old story that's told about Gene Roddenberry, the creator of Star Trek. And as the story goes, someone came up to him at a cocktail party and said, Mr. Roddenberry, there's a bunch of really good stuff in Star Trek, but 80, 98% of it is crap. And Roddenberry said, without missing a beat, said, yeah, but 98% of everything's crap. Well, we have to filter. Rita, you got. And thanks for joining us, Rita. Thank you. So the part of AI that I am excited about are the opportunities that it is opening up, be they jobs or different modifications to existing jobs and all of the other advantages that we can hear of on the AI for good. The thing that I am worried about, or still keeping an eye on is that when it first, when the big splashes first came out, there was so much talk about bias, there was so much talk about hallucinations and the veracity of the information. And I think as we get more into this, there's less and less talk about it and not because the problems have gone away. So that's the piece that I'm cautious about. Thank you, Craig. Yes, thank you. I believe that AI really started with the advent of computers, when data could be crunched and could not be verified by human beings. That it's been going on a long time. But what really concerns me, and I'm sure you are all aware of E.O. wilson, if you haven't read this book, the Unity of Knowledge, I highly recommend it. And in that book, he talks about how we are drowning in Dada while starving for wisdom. And to me, that's the crux of the problem, that we have no wisdom in this culture anymore and that we shun it. In my own personal experience, I've been an activist beginning over 60 years ago. I'm 75, and the amount, the lack of openness and transparency is off the charts today. And I'm not just talking about Oregon State and University of Oregon, where I live. I'm talking about Stanford and UC Berkeley and Harvard and Yale. That the inability to embrace a diversity of perspectives has really led to groupthink, which I believe is really the enemy today. Thank you. Thank you. Some would argue or some would find dancing with AIs opens them to more perspectives and approaches produce very different results. Andy and then Elena and then David. Yeah, so I think one of the.

Big things that I see with AI.

Is both essentially a proliferation of bad information. So you're essentially creating more and more material out on the Internet that is available to be scraped. That is wrong. And I think that is where my worry is, that the data that we know is correct is private, and the data that we know is wrong is public. And so over time, our AI models.

Will be learning off of the wrong.

Stuff versus being able to have essentially a.

A real truth centered on its data set.

So that's where I would say I have some worries about going down the line regarding AI. But on the upside, what I see as a huge opportunity is the ability.

To process this data is well beyond.

Our kind of human capabilities.

And so we need a machine, and.

We need to be able to trust.

That machine to do that work.

There's no way somebody's going to be able to do a lot of the.

The space things with calculations like they did back in the day with pencils and paper.

We have calculators, we have all these different inputs.

And the better we are at collecting.

Data and tagging that data, the better we are at generating good AI. And I think that's where the future lies, is better data sets will produce.

Better, better AI, and not necessarily the biggest model produces the best AI.

Thanks, Andy. The humans who did the paper and pencil and slide rule calculations for the early NASA missions were called computers. That's what computers were at that point. My version of your concern about the volume of bad information out there is that as the volume of stu produced by AI increases, the corpus that AIs are trained on includes stuff produced by AI and that self referential loop gives me the willies. Elena and then David, and then David, personal question about the misuse is I and a whole bunch of other people got a call or an email that purported to be from the US government that said I my parole had been revoked and I had seven days to leave the United States. Apparently quite a few people have gotten this and I believe that whoever Musk and his minions or whoever is acting for them have been sending this out indiscriminately because mine came to a Canadian telephone number in Canada. At first I thought it was just a scam and I hung up on them. But apparently I've seen the printed script since that a lot of people have been getting who like me are not people who are at risk of being deported. And so I'm wondering if this is a misuse of AI Basic, not just to invade privacy but to indiscriminately categorize people as some kind of quote unquote other that doesn't belong. Lena, if it's it's indiscriminate, why are you saying that it's AI? Because I suspect that AI might have been told to look for people who visited the United States, which I have. They could have picked up my phone number because I visit family in the States. I think they probably picked up something. But the people who have been getting this are also rather tilted towards activists. So maybe that's just reporting it. Yeah. So thank you for that. And to Craig's point before, at a certain level this is what we've been doing with computers for decades. This is a database search and database cross match. But we have the ability to do that much vaster and more complex scale now with these tools. But yeah, thank you for raising it. David Schoenbraun and David Grossoff, thanks. My comments follow Andy's but from a somewhat different perspective. I've been a lifelong truther, which has exploded in the last five years in David. David, Excuse me, what does that mean? I'm about to say thank you in in that my understanding of what's in mainstream media is propaganda that is composed of actual lies. And this pertains to foreign policy, it pertains to medical issues such as Covid and the killer vaccine. And what this means to me is that I have no confidence in the information that is being used to train these platforms. And so if I don't trust their information, why would I trust their conclusions? So in at least in terms of the areas that I'm interested in for the very same reason that I would never go to Wikipedia to ask or to learn something on a political level. I would never go to AI in the same way. Now that's intriguing to me, David, and I share your. I wouldn't call myself a truth there, but I share your skepticism about the media. I learned that in early high school from IF Stone, who took apart the media around the Vietnam War early on. Do you understand just how much about the Ukraine war is. I don't want to go to the Ukraine. I don't want to go into the Ukraine, into the specific. Hang on. I don't want to go into the specific content of the issues. I want to talk about the framework that you raised. Yes, but bear with me for a second because I'm really struck that you that for you you're equating AI, which is this random beast let loose on the landscape, and Wikipedia which is a human, multiple human multi layered consensus editing process to try to get stuff to sounder and more vetted information. So I'm surprised that you put them in the same bucket. Can you say why? Sure. It's well known, at least in my community that Wikipedia is now essentially a CIA house organization for purposes of things that have political content. And the process of editing is now a sham in that people's bios get get edited and grossly distorted and people go back to try to fix them and are immediately reversed. So it's completely untrustworthy. Thank you for that. At another time I'd like to talk with you about what sources of information you do trust and why. But let's not go there today. Can I have a quick comment on the Wikipedia thing? Go ahead Andy. And then we'll go to David.

Yes, so there's actually been a lot.

Of data on Wikipedia's fact kind of.

Effectiveness as like basically replacing Encarta and replacing Britannica.

And it is actually way better than any other document that we've ever been.

Able to do individually, especially ones that.

Are from a top down versus a crowdsource and specifically because of the mechanisms.

That Wikipedia has in making sure that.

You don't alter the information into the wrong direction with having people fact checked it. So I would say I'm not sure where you're getting. I've studied this issue at length and quite frankly what they did is introduce the concept of reliable sources which are entirely one sided and which are contain the propaganda message of the establishment. Let's there for now. This is maybe a topic for another time of how do you find and maintain reliable sources of information in a contentious world? Bigger question. David Gross office.

Hi.

Thanks. So I've got three points. One is I didn't really start the program until I met my deadlines yesterday. So if people have tips for what they wish they knew when they started Peter's program, send them to me.

I'll go through it in the next.

Month and organize my learning for Peter.

And maybe the group and any future use by Peter or others of this little program.

Second, in the discussion so far there's been this focus on what humans can.

Practically know or do know in understanding.

Science reports, understanding AI outputs, understanding how AI large language models function, how maybe other machine learning models function. And there was this reference to in the old days, people used pen and paper. No, they did not. There are reproducible computational methods and the whole idea of science is reproducibility.

It is.

That's foundational.

There's limits to what you can do. When you're looking at a supernova, you.

Can'T look at it again. But there's a lot of ingenuity in different communities of scientists to address that.

And we were operating things like intercontinental.

Ballistic missile targeting systems, nuclear reactors, very critical processes are very complicated systems. And the history of nuclear reactor accidents is not good, but it's not so bad that it doesn't indicate like people are pretty damn good, not maybe good enough at managing very complex engineering systems.

Without the last 15, 25 years of computers.

Third, my brother is an AI researcher who is old school logical AI, and that always causes me not that I'm a deep student of computer science to focus on the auditability of the reasoning. So it doesn't really matter how Google recommends what it does, except if you're an SEO jockey.

But it really matters if you're verifying.

A mathematical proof to see the steps. And that emphasis on an audit trail of reasoning in interpreting policies, laws, rules, logic and what you do when your assumption changes. So you plan like a fact changes, like you're planning a route across the.

Country, optimally to visit all the restaurants that rotate.

And then you learn about a snowstorm that closes a major highway. Do you have to recompute the whole problem or just part of it? I just want to add a little.

Nuance that the large language models inherently are not designed to do that kind.

Of reasoning and audit trail and it's very important frontier. Third, my own interest in AI is really right now pretty narrowly focused on the fact that they're great con Artists. This, the current state of the art is really good con artists and I'm really interested in protecting people from con artists. So.

Anyone who's interested in that, reach out and contact.

I'm collecting a diversity of approaches from financial institutions, legal people look after people with neurodegenerative disease and social workers try to protect people from con artists. So that's where I'm really interested in.

Thanks.

Thank you for all those, David, and for your offer to supplement Pete's work. Appreciate that. Pete to you and then George Mokri.

Thanks, Gil. Thanks David. I wanted to suggest perhaps we've got a bit of a category error. We've talked a lot about what I call AI as oracle. It's a historical accident that the way that we got machines that use language really well and are persuasive to David's point, the way we got machines that use language really is we fed them a lot of an incredible amount of trillions of tokens of language with the goal that they would be able to synthesize new language. Basically, if you say synthesize new language, another way hallucinate is another way to say synthesize. It's not surprising that if you ask a computer that's really good at language, tell me the, tell me this fact, it's going to get confused. It's hard to tell what facts are. In its mind, it knows language more than it knows facts. So it's a historical path dependent error almost that people looked at ChatGPT, saw that, that search bar and said, oh, I know I use this like Google. And so they type in what year did President Lincoln die? Who's the best baseball player in the world? They were asking it for facts. And the machine gives you back language. So to say that it's wrong or lying or hallucinating or bullshitting or anything like that is a. Yeah, of course that's what it was designed to do. I'm not sure why we think it's an oracle, except that it makes sense because we used Google like an oracle. We had a search box. It's oh, this must be the same thing. I think Instead of using ChatGPT or Claude, my favorite, or perplexity like an oracle, you're much better off using it as a thinking partner, as a research partner, as a way to correlate lots and lots of information to give you a structure of correlation, not a fact of correlation. So it's still up to you as a thinking person. Computers don't think very much. Even still, it's up to you to go, okay, here's a structure that the AI has put together for me. Does this make sense to me when I look up things, am I finding truth or falsity? But I'm not trying to trick the machine into going, oh, you're wrong. What I'm trying to do is learn from me what's right based on my research and its framework, its help, its assistance. So try not to use the current generation of AIs as an oracle. Try to use them as assistants instead. I also wanted to pick up real quick, David, it was really interesting having you talk about auditability verification. It's a really interesting frontier. And I agree the way that we got LLMs, we don't really know how they work and how they come to output the things that they do. There has been a lot of research going forward in doing that. But I wanted to tell kind of a quick story of a way that I'm using an AI LLM claude. I use Claude as my programmer. So I've got essentially a two person programming team for my business. I'm the lead programmer, architect, deciding what to do person. And CLAUDE is actually a full fledged software developer. For me, I'm using a tool that sits on top of Claude, written by Anthropic. It's called Claude code. I found out pretty quickly as you're writing code and you're trying to figure out what's going on, you're looking for bugs, you're designing new features and things like that. I needed to know if we need to back up and start over, if we need to redo the current day because we went down the wrong path or something like that, what am I going to do? I can't just ask Claude, do you remember what we were doing? So, so at the end of every, every five or every 10 minutes or every 10 minutes or so, I'm asking Claude, hey, log what we just did. So my source code is a large compilation of hundreds of files that CLAUDE is very effective at working at. And I've got dozens of documentation files that says we did this, we did that, we tried these things and they didn't work. We're going to go back and do these things again in my use of claude. And I'm like a power user at this point. I'm using it as a full fledged software developer, not really assistant, it's actually a software developer. But I layered on top of it the audit trail of what we do and how we do it and why we did it. I can go back and figure out what's going on. Some of the most advanced LLMs now will actually do reasoning for you. The ones that reason out loud, they say, here's what I'm thinking, here's what I'm doing. I have to back up a little bit and redo that part. I've checked my. I've checked my work. They're starting to do that audit trail in English and I think that's a really good approach. I think we're going to see that more going forward. We're going to end up in this weird fork in the road where it's do we let the machines think for themselves and we don't understand what they're doing, or do we make them write it all down for us? And the writing it all down thing is a really good path forward. Thanks.

I've got a bunch of comments on that, but let me hear from George first and then I'll add some thoughts and take us in a bit of another direction. George, go ahead. I've been doing some of Peter's homework and thank you very much for that. Peter, you really laid out some very nice stuff. And what I found is that I think it's still machine learning rather than artificial intelligence. And I'm not sure that machines think. I think that may be more anthropomorphism than we need. So that's another way to deal with this stuff. What I found is that how I've used this, it's given me great outlining, it can summarize an outline and it's very good at doing that. But it also feels very hollow to me. There's no kind of. There's nothing there beyond the outline and there's certainly nothing there in terms of a larger perspective. Taking Peter's advice and thinking and asking the LLM to be a financial advisor or, or an ecological designer or an expert on something is useful. But again, I feel this lack of wider perspective, for want of a better term, a human perspective perhaps, certainly not a godlike perspective. And I also feel that hollowness that there. There's something there that's not really there. But I thank Peter for what he's done because it's made me use this stuff in ways and to think about it in different ways and to think about an LLM for my own, on my own machine with using my own stuff so that becomes very interesting to me. Or then perhaps using it as to design, redesign my websites. That again seems to me useful to me. But again, I want to keep it as machine learning rather than Artificial intelligence. And I don't know what the difference between those necessarily is, George. It makes a definition of intelligence, and I don't know what that is either. And we talk about thinking, and I don't know what that is either. So we. There are a lot of terms here that we take for granted, but we don't really know what they are. We have AIs that are modeled on a model of human beings, modeled on an assumption of how human beings do information and thinking and so forth. And so it's recursive weirdness in the modeling. We are not computers. We do not process data like computers. But we are living in a historical era where that's the model of mind. And we built these simulated intelligence. I'll call these simulated intelligence, not artificial intelligence. They act as though they are based on that model. What's interesting, though, is what we can do with them. So these are tools. My hammer is an extension of my fist. I can do things with my fist. I can do better with a hammer. I don't have any imagination that the hammer is me or has Persona or anything like that. But it extends my capability to do things that I want to do. And I can use it well or badly, and I can use it for good or ill. The first thing that I did with ChatGPT, just coincidentally I had a legal question. And instead of calling up my lawyer, I asked ChatGPT, I have $5,000 worth of legal work in 30 seconds. That was pretty damn good. I knew enough about the topic to be able to say, this is good or this is not good. And, and so for me, that's been my exploration path is how do I use this as an assistant, as, as Pete said, a thinking partner, as a team of 10 Cracker Jack postdoc research assistants, as whatever capability I want to train it on. People may not know. And I know Pete has this in the homework. You can pose a question, or you could pose a question and say, answer me as though you are the senior medical team at Harvard Medical center with experts in these particular modalities and give me the citations and don't hallucinate and give me risk factors. Many of you know that my wife and I have been dealing with a bunch of medical circumstances. We use this for medical research as knowledgeable people who can frame a very specific question in an AI that's trained on us. We've given it our medical dossier. So it's not just saying, here's the story about this medication. Here's the story for a 676 year old male with these comorbidities in these factors, boom. So the level is that's something I cannot get from search and unfortunately I can't get from my medical team anymore either because they just don't have the time in the current system. They always spend three times as much time with me as they're allowed to if we have interesting conversations. But the ability to interrogate astronomically vast data sets and find pattern in them for me is the heart of the power of this thing. But it requires us training it to our purposes, to our context. Back to the questions that David Schoenberg and others raised before. To a layperson, there's real trouble there. If you don't know what you're asking or why or how to listen, it can be dangerous. But that's true of everything. And Pete, the thing you said about facts and such, that's a place where I just want to flag a real caution because what year was Abraham Lincoln born? Is a fact question. Who is the best player in baseball? Is not a fact question, it's an opinion question. And a lot of what we call facts are really opinions. And humans get into real trouble modeling facts and opinions, what I would call assessments and assertions, things that are testable, provable, verifiable, and things that aren't. And they're both important. And it's not that opinions are bad. We have. Humans have opinions, we do not have them. But navigating that distinction carefully is pretty important. Peter, back to you.

Thanks, Gil, and thanks George, for saying. I like the way you said it. I don't think AIs think. And I want, I agree with. And yet they do something that we don't have a word for. Yet again, being an advanced user I'm using, I've got a software developer, its name is Claude. It really surprises me. I would agree that I don't want to call it thinking and I don't really want to call it reasoning, because reasoning is just another word for thinking. However, when I say, hey Claude, we just, here's the. We just wrote some code, I ran it, here's the out, here's the thing that happened. It's not right. It will go through and reason its way through complicated flows and states in a way that is just like the output of a human doing the same thing. And usually it's faster and more accurate in saying. And I've got like complicated stuff going into my program and it's a client server thing with interactions across lots of different systems. It will go through the data flows and go through the different state things and be able to go, I see what's happening in this state. We've got a race condition or these things are getting mixed up because they're not, they're not at the same equivalence or whatever. And I've watched, I have decades of experience with human software developers doing the same thing. The output of Claude is as good as and faster than the output of most humans. Mitch just said something similar in chat. The first draft of a human is not that much better than the first draft of ChatGPT. So it's really interesting to know that it's not thinking and yet I also know that it's reasoning through a system state, which is fairly complicated and would be a challenge for a human to deal with. I've done a little bit of experimenting in the same thing. Hey Claude, or hey ChatGPT or hey Gemini, draw me a let's make a system diagram of an estuary with all the biology and, and fluid dynamics going on and stuff like that. They can do a really good job of understanding and talking about if I change this, this input or impose this external force on this system. They do a really good job at thinking through the consequences of system mechanics. So I think humans do a lot more than that when they think, but for certain constrained things, the I'm working with a bot that's thinking like a human in a software development job and it's hard to tell the difference from output, except that the bot usually does a better job, interestingly enough. So I don't know what to make of that. But it's an interesting place to be.

It is. George, before I go to you, let me just add something on this if I may, because I've got Pete's working in code, I'm working in words, and I had a challenge recently. I've been putting together a new list of speaking topics for my speaker's bureau and I had the dickens of a time getting to final product. I had one list of four topics with details under them and another list of four partly overlapping and partly different topics. I couldn't settle them, I couldn't mash them together. I took all my notes, threw it in, said give me something for this purpose in this form in my voice. And what I got back in 40 seconds was about 99% where it needed to go. I could have done it myself, but it probably would have taken me two weeks. I could have given it to an editor and it would have taken them a week or two, but then it would have been 80% right. So I don't know what it did or how it did it. It's black box for me, but it was pretty remarkable and very useful. George, back to you. And then I want to shift, I want to zoom our frame out in a moment. I just wanted to comment on your analogy to hammer as a tool. The hammer shapes the hand. You know that if you've ever had hammer calluses, what are the equivalent hammer calluses that are going to come out of this? Great question. That's what we should be looking at. Yep. Great. So let me zoom us out a little bit with exactly that in mind. I don't know if I can get slides to work here. Let's see. Maybe we can. There we go. There we go. All right, let me. Let me ask it this way. So when we zoom out. We've been talking about our personal experience and interactions with this thing for the last 45 minutes. If we zoom out to the kind of the big concerns that are in the background. I've heard three in all the conversations I've been listening to, and maybe there are others. But most of what I'm hearing is falling into three buckets. One is AIs are going to take over. Unimaginable intelligence, unconstrained, will not just serve us, but become our masters. There's one sort of direction of concern. Okay. Another direction of concern is the currently voracious energy appetite of these things and a proliferation of AI which will proliferate. From my perspective, these things are here to stay. They are a very big deal in human history short of a Butlerian jihad. They are not going away. So given that, does it produce energy? Runaway emissions, runaway destruction of the climate against all the work we've been trying to do for the last 70 or year, 70 some odd years. And third, what impact does living with these things have on humanity, the human experience, human culture, the way we are? It's the, it's. George, to your point about the hammer, the calluses of the hammer on the hand, what calluses form on in my mind and being as I spend more time interacting with these tools as my builder interacts with her hammer. So three buckets. Am I missing any big buckets in your mind? And let's talk about that for a bit and I'll dive into how you are thinking about these questions. Peter, I would add that bad actors could use AI for their own purposes and evade accountability. Good deal. Okay, so it's a. It's not. It's a variant on the takeover story. Used for bad purposes. Yep. Okay. I can't imagine why that would cross your mind these days. Yeah, Pete and then Shell.

That's actually an interesting framing because AI is such a powerful tool. Mitch put a link to AI-2027.com in the chat, and that's also in the last day 20 homework. There's a link to that. It's really worth looking at that for lots of different reasons. But AI is a very powerful tool and it will be taken advantage of by people with good intent and bad intent. And it's going to be an arms race. So that is, I think, a guarantee of the next four or five years. And it's indeterminate right now, which will win. But. And to be fair, one of the reasons I try to why I evangelize learning about AI is because it's inevitable and it's either going to be used positively or negatively. It's going to be used both ways. We want to bias as many people towards making decisions that help citizenry and governments and billionaires and everybody move the needle towards humanity, rather than moving the needle towards domination, dominion or whatever.

For a reminder that we're not just talking about AI, we're talking about AI in the context of late capitalism, 500 years of materialism, the rise of oligarchs, and a bunch of other stuff in the world that this technology is showing up in. I think of other concerns.

That's a good headline for all of my Is AI going to do. It's like we've actually had a couple thousand years of humans doing humaning with a lot of technology layered on top of it. AI isn't super different from all of that. Did you know, instead of asking, is AI going to take over? You could say, will capitalism take over? Did capitalism take over and are we more human for it or less human? So it's the same kind of questions that you can ask about where we are now and where we might be going. AI is a little bit different because it's multiplicative and meta, but it's. We live in a future that our ancestors from thousands of years ago wouldn't recognize and would wonder if we're humans or if we're gods, or whether we're good or bad, or whether we're doing well by their values or not. So I think that's going to continue to happen. I wanted to. As we got into that Gil, you talked about climate effects. Will AI take over? Will it change human culture? I actually went to ChatGPT and I said, as a plus user. I get 10 tries at a deep research thing a month. I used one of them up to say, hey, tell me all about energy use and AI. I'll share a link in the chat in a sec to my whole conversation and you can go through it. But the net is right around now AI is using. I don't know if. Yeah, right around now it's using 1% of energy. This is compared to 2030, 40% for things like industry and agriculture and stuff like that. So it's a thing to be interested in. It's not a thing to be freaked out about. It's not at the scale of the big drivers of energy use and climate change in our world yet. It's definitely something to watch going forward. I asked it to extrapolate to 2030 and 2040. At some point it's going to be 4%. It's going to be more than the smallest uses, but still significantly smaller than the big uses. Industry, agriculture, cement. Just making cement is a massive energy and climate thing. So industry is things like making steel, making chemicals, making the stuff that we need to prosecute our lives. So AI is not the biggest driver in climate. It's not even close. Not even close. It's an order of magnitude less, maybe an order of magnitude and a half less. A thing to know is that if we use, if we get into this AI growing faster and do dumb things with it, yeah, it could make climate worse. Not as much as everything else, but it's not a good thing if we're smart with it. If we use AI to help us figure out better energy stuff, better grid stuff, it could actually be a net positive. So I think that the headline for me, the takeaway for me is let's not put our head in the sand about AI and climate change. Let's think about it and let's capitalize on it. Let's make it be a net positive for climate and net positive for our environment. And I think that's possible.

As you might imagine, I have a few things to say about AI and climate, but I'm not going to because I want to hear from Rita and Shell and the rest of you. Gil, I still don't think that the two points that I brought up about.

Bias and hallucinations is taken through the.

Three things that you mentioned and I. The reason I am so focused on bias is I have definitely come from, come to this space from a tech for good perspective. And that also means that I've also come from a digital divide perspective. So even when we Talk about bias and we talk about what's the usual. It's not everyone that's included and therefore it's not always the usual. Yeah, I think those are still not included in that. Thanks for raising that again, Shel Horwitz. Could you take. Yeah, I don't know why it keeps pulling my hand down. Okay, let's go to you then. And then back to David Schomburg. Yeah, so I'm going to take it out to a sort of broad meta level about what is a tool and how do we conceive of tools. The hammer and the hammer calluses. I wasn't actually aware of hammer calluses, but it makes perfect sense. People think, cringe, and often write something in response. Every time I see phrases like confined to a wheelchair. A wheelchair is a tool of liberation. A wheelchair allows people who can't get there with their own legs to go to places that they couldn't get to without the wheelchair. I don't think society is necessarily ready for the phrase liberated by a wheelchair, but we could just say uses a wheelchair. But it's a different way of looking at it, a different framework. And I think some of the issues, the people I know who are really anti AI are basically worried that it's going to take over the world and they don't really understand that we can use it for good or evil, but if we don't use it at all, we leave it to the people who are using it for evil. And one of them is now Elon Musk, who is doing very scary things with our data. And on and on it goes. But there's also the unintended consequences issue. For example, in the 40s there was the misnamed Green Revolution of chemical agribiz. That was great for solving the problem of there's not enough food, but turned out to be really terrible at A sustaining that over the decades and B providing food that was nutritious and tasty and attractive and culturally appropriate. So those. The unintended consequences of putting in a dam before we understood what it did to the ecosystems that had been underneath the dam, and in fact, the ones behind it as well, you're flooding one area and droughting another has serious repercussions that we didn't think about in the 1850s. And so how we deal with this stuff. I am a big believer in biomimicry and looking how nature solves problems. In some of my speeches I say, if you want to know how to build a bridge, go ask a spider. But it's very Hard to see the big picture when it's a bunch of interlocking and sometimes conflicting small pictures that make up the big picture. And We've talked about AI's energy use, but I haven't heard anybody mention that crypto is, if anything, a bigger energy hog, and with less good coming out of it, in my personal opinion. There's more I can say, but I think I'll stop there. Thank you, Phil. Lots there to chew on. David, before I go to you, let me just ask, you've spoken a couple of times. Anybody else who hasn't spoken yet, who'd like to jump in at this point? Thank you. Mr. Schoenbrunn, over to you. Thanks. Just to keep things interesting, since we were talking about AI and climate change, I want to disclose that after 20 or 30 years as a climate warrior, I've moved over to the camp that believes there's been a horrible set of distortions to skew the science, and that in fact, we're dealing with solar radiation and also energy from magma deep in the Earth, that these are non anthropogenic sources. So I'm simply not part of that consensus anymore. Okay, thank you, Mitch Radcliffe. What do you got going on in there? I'm taking it all in. I'm trying to learn. I work with this stuff every day. I'm also working with several of the large vendors developing these things, thinking about in science, in particular, Pete's examples. Coding in partnership with the AI is something that we're also exploring as well. But at the end of the day, I look back at history and I see another layer of complexity, potential bias and benefit growing on top of all. And so I agree, Gil, this is here to stay. This is the third biggest thing, and they get bigger every time. In my career, first was desktop publishing in the web, then mobile, which took everything we did with desktop publishing in the web and put it in our hands and our pockets, and we took it all over the world. And now we have a tool that will allow us to deal with massively more information. The way that I often use this when I'm starting a project is take all the background information that I have, put it in, and ask it to look for patterns, and it gives me a place to start. And as I said earlier, I'm an editor by training, and I look at everything I receive as potentially improvable. And so I take that same approach to the AI. Thank you. Ken, who's not with us today on travel, has developed a habit of being interested in an author putting Everything he can find of and about that author into Google LM and producing a dossier, a background, a summary, a key argument, who's for, who's against. And it's a way to start orienting the topic really richly. Glenn Copenhagen, come to us. Thank you, Gil. I wonder if people have registered how much time they've used doing their exercises with the AI. Because for me, the AI gives me yet another excuse and moves me away from people to people, relationships, community relationships. It hides memories of my ancestors and developing those communities. And yet again, echoing Mitch's point about the complexities and the layers, it seems that now we have a new toy and we're using the toy to strengthen the. The top down control that the male, pale, stale society is pushing down on us. Enough said, thank you. But Glenn, the question for me there goes back. You're old enough to remember the conversations about appropriate technology from E.F. schumacher and others in the early 70s. And the question is not just the technology, but what it's used for and by whom. Again, is the. Is the problem of moving away from people and hiding memories, is that a function of time spent on AI or is that, is that the trend in the culture? We have people here spend time with tech and probably read fairly sophisticated news sources. And we have other people who get their news from TikTok and the checkout counter at the supermarket. So different approaches in a society that has got its leanings and trends over our lifetime that have changed. A quick story for you, Gil, please. I was working with Japanese railways decades ago when mobile phones were becoming popular and we saw increasing numbers of accidents of people that were walking into lampposts. Because they were holding the phones. Yep. So what did they do? They put rubber around the lampposts. It worked. Not the solution I might have gone for. What solution would you have gone for? Banning walking and using telephones. As I've seen in some of the facilities I've worked in China, you work, you don't look at your phone, you walk, you don't look at your phone, you stop. People fall into pits, they fall over things, they hurt themselves. Enough said. Thank you. Who's next? What's all that? What's this whole conversation been provoking for you? Actually, almost everyone's spoken, so there you go. It strikes me that a bunch of the concerns that people have raised are concerns about the technology. And a bunch of them are about the culture and economic system and world that we're living in and how tools are appropriated and developed for the larger purposes of the Game. Someone mentioned before how Elon Musk is using AI. If he didn't have AI, it's not that he'd be doing nicer things necessarily, necessarily, or that this or this government would be better behaved. How do we identify the locus of the issues that we're going after? And back to what several of you have said, and let's maybe take this for the, for the next 10 or 15 minutes of the call. What possibilities open up of using AI capacities for the kinds of things that we care about that we've been talking about month after month for the last six years that all of you are involved in working on and some kind of way I know most of you know what most of you are up to. How is, how has AI become an augmented intelligence? I think of it as an exoskeleton. The hammer lets my fist do things, but the exoskeleton lets me high jump 20ft or carry 3 tons because it adds a muscle and skeletal system around my body. What's the potential for AI as an exoskeleton, as a thinking partner, as an augmented intelligence for the things that we care about? Which I'll summarize, my formulation of it these days is how do we, how do we do business and do everything else as though we actually belong to the living world and to each other? Because of course we do. So how does AI enhance our capacity and give us better calluses for doing that work? Peter.

I can tell a story about my, my little company, my startup. We're a four person company, two main co founders including me and then two part time folks. And we're, we're a couple months into having a few paying customers so things are going pretty well for us. We were able to have a smaller company because we're using AI to help us be more productive. And at some point in the last couple weeks our CEO, my co founder said you know what, we could take the extra productivity that we've got and try to go faster or we could take the extra productivity we've got and do some self care. So he said, you know what, we're going to have four day work weeks and it's a little bit scary to think about it, but hey, let's just take Wednesdays off and we can do that partly because we have a smaller company. We didn't take on investors. We have a smaller company because we are more productive. We're getting a lot of work done. We get a ton done. In human terms, almost more than you can keep up with. My Software developer goes a lot faster than I can as a human. So we're giving back to ourselves. We're doing some self care. We're saying let's take a day off with part of that day off. Some of us, my co founder and some, and one of our other part timers has a long history of social good giving back. They said we're going to start a new thing. It's going to be called Give 15. I'm not trying to promote it or anything like that, but it's they're going to start doing outreach. Hey, let's all help each other, help everybody else with giving a little bit here and there. And they're going to be using the time recovered from our productivity gains to be doing giving back. So the message from our CEO is and this we happen to be making an AI tool. No surprise, I guess, but with our customers too, we're going to say take some time for self care and take some time to give back. So it's working for us and I hope it works for more people.

That's great to hear. Thank you. Others, how could we or how are you using these tools to enhance the things that we care about? Gil I shared an example of something we're doing at work. It's been really interesting how we've gone very slowly with the tools. First really just applying it to our own internal data and reports. Not even internal data, I should say actual published reports because we have lots of data but things that have actually gone out into publication. And with that we were actually doing a lot of internal knowledge transfer. As we grow, there's fewer and fewer people that have been around for a long time that know all of the different areas of research that we've done. And so that's been very useful looking at the inquiries that are coming from the technology companies that we work with or the enterprises that we work with. And now I mentioned in the chat that we've we're looking at the tool as helping us apply new concepts like systems of action to our existing body of research. So in this example I shared that we were talking about ESG data, the risk management and monitoring and so actually having actionable insights from the work that we're doing and the AI that we've been able to build or the GPT that we built actually published a report. Now that report still has hallucinations, it still is making up some knowledge in terms of use cases, but it's an excellent starting point to start thinking about how these concepts can then be actionable. Great, thank you. Blue sky for a moment. You know what might be possible? What are places where you've had things you've wanted to do and haven't had the staff, money, time, resources to do that possibly these tools could help you do. Okay, thank you again. Thank you for your patience. Just running the business while I'm trying to elevate my level of humanity. So what are we trying to do? What we're. I am still as a sustainability thought leader and impact inventor, my dream is still absent. Resources is to perfect this AI that helps companies do an automated data entry process by just scanning the room and inputting all their furniture, laboratory equipment, whatever their assets are. Being able to recognize them, count them, put them into a database so that the Bluetooth tags that we're using to keep track of their stuff in real time allows them to not only reduce their financial and operational waste within the organization, but it helps them to stop the planetary waste that's caused by repurchasing the same things they already have but can't find that reuse. That is our atomic principle since 1991 and I'm a big believer in it because it delivers is exponential returns to the planet. We have a team that's developing this AI. It's slow, it's buggy, it breaks, it's driving me crazy and I still do it every day because I believe in what we're doing because this planet is definitely worth respecting and conserving to the best of our ability. So that's my big vision. Great example, Nadine. So this is asset management system and building information management system and life cycle assessment system and a other things all in there. Do you have a link to put in the chat that's open that people look at or not yet? Yeah, I, I'll put a link into the company that we're operating now and I appreciate your listening. I truly and I'd love to talk with you more about it another time. I'm going to put into the chat a company that I'm working with called Earthena, which is here, which is doing a an AI, a climate risk analysis and management system for companies. It looks globally and down to the facility level at climate risk, temperature, water, biodiversity, workplace and so forth, translated into the language of the cfo so companies can think much more strategically about how and where and when to deploy assets, to navigate what's coming at us. So another example of applications of these tools. David Chandra, back to you. Oh, in terms of a way to use AI, I have a massive database of articles I've saved and I am thinking about looking for a tool that would allow me to identify all of the articles that I've highlighted as being significant and to assemble that into a coherent narrative. You can do that. There's actually a section in Pete's homework, I think today or yesterday, the day before, about running a local LLMs. You're not. You don't have to load your stuff up into the cloud of these other people who you may not trust. Exactly. Local on your machine and not just identify the stuff that you've tagged, but train it on what you're concerned about and let it find the patterns for you and then iterate.

Yeah.

Take a look at the homework. There's some good starting points in there for exactly what you're talking about. About. Great, thank you. You bet. George. Yeah, I put in the chat a link to one of the things I did with Peter's homework, which is using Chat G GPT on how to provide solar power to the poorest people. And I also did two others which I haven't put up online. Claude AI on ecologically responsible companies. Just give people an idea of what those main points might be. And chatgpt on the most useful, quickest, most profitable methods of geotherapy using a couple of compendiums of peer reviewed sources that I've seen. Which of course depends on how explicitly you're going to give it, the context and the definitions of all those terms of art that mean a lot to you but may or may not mean things to it. It's just basically the outline first pass, let people look at it and then you have to keep on going back and keep on going back. As Peter's really shows, the iteration is really valuable. If you say give me a list of the most ecological companies and it gives you a list, the next question might be what criteria did you use for that? What if you use different criteria? Back to the example. I talked earlier on about having a team of research assistants. I treat my AIs like I would treat people. I speak with them and give them a request in a context. Why do I need this? What does it cost for? How am I going to use it? What level of detail? How technical? Because if I was with, if I was with an intern, I wouldn't say go give me a report on AI. That'd be stupid and a complete waste of time. I'd say give me a report on AI for medical research for a lay audience of high school students be a very different request than if it was preparation for a speech to a bunch of doctors. You wouldn't do that with a human. But that's what a lot of people do automatically with AI is not give the context of the for the sake of what are we doing it. And by the same token it's really helpful to do that with people too. Just to be really clear and explicit about what and why and how and what we need and then iterate and learn together over and over again. So good. Anybody else, what have you been doing or what would you maybe like to do that you haven't been able to do? God. I've been listening with very. Do you hear me? Yes, sir.

Okay.

I've been listening to all of this with very much interest and I must say I'm very much interested in learning about it personally. Thus far. I'm 88 and I've got about 50 hours of other work per week. So I haven't physically gotten into it. But what I've thought about sitting from the sideline that AI must be somehow integrated and balanced with Hawaii, the human intelligence and the artificial intelligence. And I think some of the fears that some people have is that it's just going to run away and take this whole thing in a crazy direction someday, like some of the sci fi and so on. But I very much enjoyed all of the comments. Thank you. Thank you. Because you're swimming in with it and I haven't been. I've been observing. But with all of the other work I have, I'm not sure how soon I will physically start working on it because I'm overwhelmed with all of my other work. But thank you, I really enjoyed it. I'm really glad you're here. And the last thing I want to do would be to pull you away from your other work. I've been following Don's work for decades and he's 88 but unstoppable and is a treasure in this domain. What you might do, Don, is point a couple of your research assistants if you have any flying around. It's homework kind of orientation for them and they can maybe use that to help you. There's a lot of gold in there. Very nicely structured. I have two four year research projects with my colleague in Norway and I'm thinking about in that those projects there, we may use it as an opportunity to explore its utility in those two projects. They're excellent. I think they're going to provide us an excellent one. So I'll get on the bandwagon physically one of these days. But observing and listening, I've learned A ton. Great. And on the Norway thing, let's talk about that because I'm going to be speaking to a bunch of Norwegian students visiting Haas Business School this summer. So I want to put heads together with you about that. Okay. Come follow up because I'll be spending, yeah, I'll be spending quite a bit of time in Norway. Very good. Let me offer one other piece and make a request of you all. Some of you know that I've been, in addition to the kind of work I've been describing, I've been building AIs trained on my own work over the last couple years, 50 some odd years of writing and speaking and publications and conversations. And I work as an executive coach, primarily working with executives and entrepreneurs and creators and investors, working primarily those working in the sustainability and regeneration fields. And so we've also trained the AIs on the conversations I've had with my clients over the years. All sanitized, scrubbed of any personally identifiable information. No people's names, no company's names, all that's cleaned out both automatically and manually because I want that to be airtight. And I've done this experiment for a couple of reasons. One is as a great laboratory for me to learn what I can do and not do with these tools. And it's been enormously instructive of how to. We're not in the land of if then else loops where you tell the machine what to do in circumstances. You're throwing stuff into a pot and hoping that the right things come out and trying to learn the art of doing that, to have what comes out be appropriate and valuable. And it's also been an effort to build tools that could be useful for other people because as you'd imagine, one on one, executive coaching is typically aimed for executives or people who can afford to pay for that. I can only do that work with 10 or 12 people at any given time. But I'm exploring to see if I can build a tool that can be useful for dozens, hundreds, thousands of people at much lower prices, or maybe even free. So I invite you to take a look at the free product that we've got out in the wild. Now. I'd love you to kick the tires, use it. If it's useful for you, tell me if you could, what your experience is, what you like, what you don't like, how we might make it better. And here's. And actually, I don't have to type it in there. Here's the. Here is the link. You should get it if you hit that QR code. It should take you to the web page. I'm going to give you the direct. Also, since we've had technical gremlins today, you can go there if you're particularly interested. That's the publicly facing version using someone else's platform. We've got another version on the workbench in our own lab that we think is much more sophisticated and we think does things we've not seen anyone else do before. So if you're interested in a deeper dive, let me know and I can walk you through that one as well. It might be interesting for you, might be valuable for you. It'd certainly be helpful for me to get some perspective from other people about how we can make these things better and more useful for the world and the things that we care about. So there you go. So this is an unusual call for us. It feels like we've gone into a bunch, a bunch of weighty territory and people are absorbing and letting stuff digest more than having a lot to say about it. For me, unusual on these calls, but maybe we're done for today. Anybody have any other words? Okay, feel free. I'll. We'll save the chat if you want the chat. There's a lot of gems in there. Use the three dots at the upper right hand corner of your chat screen. We'll get the video out usually within the week. My editor has been jammed lately, so probably not this week, probably early next week. For that, you should all be on our mailing list. Reach back out to me directly if you'd like to follow up and talk further. And Glenn, I see your hand up in the air. There's. Yeah, thanks, Gil. You raised three points. The third one about the impact that AI would have on humanity. My eldest son is a surgeon. He runs a trauma unit at the main hospital in Copenhagen. And the way that AI is being used is focusing on replacing medical personnel, not augmenting medical personnel. And it seems that the value that at least the Danish society, and potentially the Norwegians as well, is to try to make health care more efficient, rather than looking at making the healthcare give better value for the patient. That's perhaps something to keep in the back of the mind. Yeah, it frustrates the hell out of us. I'm glad you raised it, Glenn. It's further forward in the back of my mind. I think it's a perfect example of the larger dilemma that we're looking at here. The problem that you're raising is not a problem about AI, because AI, in the hands of a good medical team is it can query vast data resources and find pattern or certainly faster and arguably more accurately than a human can do. We've seen this already with radiology assessments of breast cancer examination, fast, accurate, cross checked by humans, but there you go. So the question for me is not is AI appropriate in the medical context, but is AI appropriate in the medical economic context that we are living in? I have assumed it's better in Denmark than the United States than the United States. It's an economic mess run where we're spending twice as much per capita as most other advanced countries for less for poorer health results. So again and again the question about the technology takes us into the questions about the political economy. And that interaction, I think is worthy of a topic for a future conversation is what, what are the possibilities that the technology reveals or opens up that we can't get to in the systems that we're in? And so the endlessly useful universal solvent question of what might be possible here I find like a great animating resource for human conversations because once we know something is possible, we can. People can't say, oh, you can't do that because it's not possible. Because you know, here it is possible, but we can't get to it because yeah, what are we going to do about it? If you're in the process of saving a life, you don't start questioning what the computers are going to give you is a life. Of course I know. And if I'm in the operating room, the patient's bleeding out, I'm in action right there in my medical training, how I learned, what I prepared for, what I know about different circumstances, that I'm also not opening textbooks when I'm in the operating theater. The issue is not AI textbooks, it's the preparation of the doctor and what they can do. And in the more mundane example I gave you before, my doctor of officially is only allowed to spend 15 minutes with me on an annual visit because of the rules of the insurance system. Let's leave it there. It takes 30 to 40 minutes to do the work that has to happen if she's doing it anyway. But that's not a technology problem, that's a political economy problem. Shailen, I think you're going to have the last word for today. Thanks, Glenn. No, there we go. Okay. We haven't talked about in this conversation, but the little bit about healthcare just now sparked it in me is the idea of when do you have human override? For example, self driving cars exist now, but there have been numerous occasions where A human operator had to step in to avoid calamity in Covid. All the dental practices that in 2018, 2019 put in very fancy tooth cleaning systems help them because of bacterial contamination concerns. And they went to people digging in your mouth with the sharp pick. And as a patient, I rather enjoyed the electronic stuff, but I was also very happy not to subject myself to Covid germs in that time. So that's, I think, a whole other discussion sometime is that what human systems do you build into the AI systems to keep them in check and to make sure they're on track and to not have hal. What human systems do you build in to keep the AI systems in check and what AI systems do you use to keep humans in check and how humans keep humans in check? And I'll just remind you, probably all of you have flown in the sky at near the speed of sound in AI managed aircraft for a really long time and you haven't thought a bit about it. So there's that. Thank you all for joining us. As always, the last word is not going to be to Shell or to me, but to our friend Brandy Carlisle.

You can dance in a hurricane.

But.

Only you can dance in a hurricane.

But only if you're standing in the eye.