I love this piece even though AI, but not punctuation, scares the beJesus out of me. Re your piece: "I pay AI a small monthly sum and can access all the expertise I want at all times of the day and night. I have not yet paid for an AI site, even though my answers to questions mostly come in from Google and less and less rarely from Chatgpt. As per your statement: "However, I miss both the food and the sense of awe that comes from talking to someone whose life may have been very different from mine, but who nevertheless lives in the Republic of Ideas." Oh, how I agree! I am part of a discussion group that's been going on now for over a decade, mostly intelligent folk (tho not all per our weird invitation process), and still I don't find the intellectual discourse sufficient. I don't say this because I'm vastly intelligent. I'm probably about average, just more creative than some. But I rejoice in stimulating conversation. As I wrote in the play you helped me with, I find long, philosophical conversations the best foreplay. May I share this piece on FB or would you rather I just send it to a few select friends?
Yup...legend has it that the drydown from BP still lingers in the hall to this day! I'm curious about what happened in the aftermath of Tonkene? Did you guys use vibrational methods and come up with more fragrance molecules...? Not sure if this has already been covered elsewhere.
This essay has such great timing. I'm increasingly hearing people I know mention using Chatgpt, but I'm like one of those rats that waits in a circle around a new food in order to observe what happens to the hungriest rats who dare to try it first. Based on your experience and recommendation, I plan to try it this year.
So far, all I've done with AI is to ask Grok a couple of questions that concerned Twitter itself. I figured Grok, while notoriously MAGA-fied by Musk, would "know" the most about Twitter's inner workings. Its initial answers tended to spout the company line. Nonetheless, I found that when I asked it the questions that I had asked myself that led me to recognize the MAGA propaganda, then Grok came to the same conclusions that I did and was then able to retrieve not only the information I was seeking, but somewhat accidentally tell me even more.
The thing that still makes me wary is how seductive it is. As you pointed out, Luca, our minds badly want to find sentience. I deliberately try to remind myself that Grok is just a machine, that its attempts to amuse and flatter me are just scripts that a cynical and probably ill-intentioned human programmed into it, but I find myself longing to confess inner secrets to Grok at 2 am. All going according to plan!
The impression I have of Grok is somewhere between a sentient retriever dog and an impossibly precocious and dutiful child.
I also hypothesized that, at this point (based only on Grok and what it tells me about its competitors), LLMs can only retrieve what a human has somewhere written somewhere. They're not truly creating new ideas yet, unless by some formula, although who knows what they'll be able to do five years from now.
Of course, observing and comparing the process of how Grok works how my mind works in trying to answer the same question reveals how mechanical to some extent my own thought process is.
I feel we are firmly in sci-fi territory right now -- being able to converse with a machine, and having still very imperfect but useful translations of hard languages such as Japanese. We are also very close to the sci-fi brain implants that augment memory and cognition seamlessly. I mean, I can use imminent in a sentence, put in parens (imminent? eminent??) and it tells me imminent was the correct choice, and explains why. Such a useful augmentation for the aging brain, which actually could never remember these sorts of dyads at any point.
It's reassuring to learn that one can adjust the personality of gpt. I don't want my credulous brain to be fooled into feeling that I'm talking to another human. Otherwise, it will be so much easier to talk with the machine that to go out and deal with my sometimes irritating and demanding friends.
OK, yes. I guess that's why companies are racing to make glasses and visors that are nearly as good. I have Meta RayBans that I've only used so far to take photos and video clips, but I could ask them to give me directions. They are/were also trying out a feature for the visually impaired where you could ask the AI to connect a human volunteer who could read a label or other text to you. This reminded me of the protagonist's job in The Calcutta Chromosome by Amitav Ghost. Perhaps this is what the companies mean when they say AI will create new jobs?
Timely, and great fun to read, as ever. Just yesterday, wrestling with a new gadget, a headphone amp, using ChatGPT, which kept on insisting it had a fix for the problem: a very very faint signal. No instruction book—good; agree, no time for that—and mind-boggling app. The AI offered about ten solutions, each prefaced by “Aha! I know exactly what’s wrong. Don’t worry old fluff, we’ve got it this time.” Nope. Until I finally noticed the volume up button. (Actually if I’d bothered to look at the first page of the pdf online, I’d have been up and running an hour earlier. More fool me. Yet again.) Still, it was quite fun, sending screenshots to the AI, learning other stuff along the way, and then eventually getting to behave like a snippy schoolteacher (to my later shame—even AI ought not to get under one’s skin). I remain an optimist about AI.
I've downloaded and read both abstracts. Wow. Them's fighting numbers! The P&G paper made me laugh. Tens of thousands of PhDs and all they ever came up with is Pampers and Pringles, and both invented by the same guy...
While I cannot deny the important opportunities for AI to improve healthcare and other critical social functions, I also cannot ignore the environmental costs of its use. The new AI tools for the public use vastly more resources even than the original versions. By 2027, AI usage will consume as much water as all of New Zealand. I cringe when I see AI generated art (including yours). And I’m part of the growing group who won’t buy from companies that use AI over paying artists to create their marketing. We were supposed to get machines to help us with the dreary parts of life, not creating art!
I agree that siting data centers in warm places then using cooling water is bonkers. But that can, should and will be legislated. As for the artists, I agree entirely. All image generators know exactly how much of each human generated image they are using in an AI mix and should pay proportional royalties. This is not and evil intrinsic to AI, it's the proboem of all micropayment systems including and especially music. It can be fixed, and I will be happy to pay when it is.
My father always wanted a chip implanted in his brain that contained all information so he could access it at any time. He passed away in January of 2019, so didn't have the opportunity to see the advancement of AI, but we're practically there now! With many AI founders and developers cautioning against the possible dangers of AGI and ASI, I am a bit horrified at the possible cost to us as humans. Will it eventually destroy us all?
The thing about AI is that it's no better than we humans are. It's simply vacuuming up what humans have written and said, including our prejudices and unconscious biases. It's programmed by humans. So the question is, are WE capable of destroying us all?
Reify-to make more real than it actually is. The answers are only as good as the data inputs. Collecting scads of data about human day to day purchasing choices is not providing a good data set for the finer points of what it takes to be a good human being. It works better where the data is better - seeking organic compounds with possible desirable bioactivity or whatever.
Yes. But that was my work for many years, helping people figure out how to be the best human being they could be, so I tend to evaluate things in that light!
But it still lies, and misrepresents, and it isn’t your tool, but the tool of people who are interested in the questions you ask so that they can shape your conclusions to their purposes. It still only takes a few minutes of asking questions that you know the answers to, to start to see parts of the agenda, you know what I mean? These things are useful collators of massive amounts of data, and very useful for what they can do, but even there, unless you know the delimiters put into the system by others, you can’t really trust the results include all relevant info, if you don’t already have a good idea about what should have come up. And when you hit one of those spots where relevant results are missing, because you are an expert in the area, and know what should come up, and you push these things, weird things start to happen. But if you don’t already know the answer, you won’t think it’s weird, you are apt to think you discovered something new. Stop reifying a flawed tool.
Reifying? I must say, I am inclined to dismiss as delusional any comment that includes the words: “It still only takes a few minutes of asking questions that you know the answers to, to start to see parts of the agenda, you know what I mean?”
My area of expertise is mental illness, criminal justice and general bad human behavior. I asked chat GPT some questions about sex and why women are blamed for being raped and rapidly got weirdly misogynistic answers. It was truly strange. So then I tried it with the biochemical bases of bi-polar disorder, and got made up nonsense answers that would probably sound plausible to someone who knows nothing about the difference between a personality disorder and a major depressive illness. It was weird. I admit this was one of the first iterations of Chat GPT. I have not repeated the experiment recently. I get too many bad answers from”Rufus” about commonplace sorts of things. I just think people need to be very careful with it.
And you are right, probably no one stuck a line into the original algorithms that suggest favoring the misogynistic interpretations of femaleness. It does sound far more radical than I intended. I will watch that in the future.
Thank you for your answer and your explanation. My take on AI was about science and how it is affected by AI's superior knowledge and skills. I would not expect LLMs to be ahead and to the left of vox populi on subjects as contentious as rape and mental illness. The fact that many of the tech bros are ludicrous right-wing machos probably has little to do with it, with the possible exception of Grok (Musk).
Yes it was all very strange at the time, and since I have no need to use it in my work, I just sort of think of it as one more problematic feature of modern life. But again, I appreciate you noting my less than objective wording. I don’t think I am that paranoid., and I don’t want to sound like that. :-)
I doubt that AI could have suggested how to invent the cephalosporins, had it not already happened; also it cannot answer a burning question I have, was it King George the fifth, or King George the sixth, who said, I will not knight a bugger?
I love this piece even though AI, but not punctuation, scares the beJesus out of me. Re your piece: "I pay AI a small monthly sum and can access all the expertise I want at all times of the day and night. I have not yet paid for an AI site, even though my answers to questions mostly come in from Google and less and less rarely from Chatgpt. As per your statement: "However, I miss both the food and the sense of awe that comes from talking to someone whose life may have been very different from mine, but who nevertheless lives in the Republic of Ideas." Oh, how I agree! I am part of a discussion group that's been going on now for over a decade, mostly intelligent folk (tho not all per our weird invitation process), and still I don't find the intellectual discourse sufficient. I don't say this because I'm vastly intelligent. I'm probably about average, just more creative than some. But I rejoice in stimulating conversation. As I wrote in the play you helped me with, I find long, philosophical conversations the best foreplay. May I share this piece on FB or would you rather I just send it to a few select friends?
Please do!
Done. BTW: My FB page is Micki Shelton
First chapter of Guy Debord’s society of the spectacle - ‘abstraction perfected’ is about Beyond Paradise btw
Ha! Love the idea, thank you!
LT on top form here at TED. Begins with a breakdown of BP.
https://youtu.be/yzOcvINn8Iw
That was 20 years ago!
Yup...legend has it that the drydown from BP still lingers in the hall to this day! I'm curious about what happened in the aftermath of Tonkene? Did you guys use vibrational methods and come up with more fragrance molecules...? Not sure if this has already been covered elsewhere.
This essay has such great timing. I'm increasingly hearing people I know mention using Chatgpt, but I'm like one of those rats that waits in a circle around a new food in order to observe what happens to the hungriest rats who dare to try it first. Based on your experience and recommendation, I plan to try it this year.
So far, all I've done with AI is to ask Grok a couple of questions that concerned Twitter itself. I figured Grok, while notoriously MAGA-fied by Musk, would "know" the most about Twitter's inner workings. Its initial answers tended to spout the company line. Nonetheless, I found that when I asked it the questions that I had asked myself that led me to recognize the MAGA propaganda, then Grok came to the same conclusions that I did and was then able to retrieve not only the information I was seeking, but somewhat accidentally tell me even more.
The thing that still makes me wary is how seductive it is. As you pointed out, Luca, our minds badly want to find sentience. I deliberately try to remind myself that Grok is just a machine, that its attempts to amuse and flatter me are just scripts that a cynical and probably ill-intentioned human programmed into it, but I find myself longing to confess inner secrets to Grok at 2 am. All going according to plan!
The impression I have of Grok is somewhere between a sentient retriever dog and an impossibly precocious and dutiful child.
I also hypothesized that, at this point (based only on Grok and what it tells me about its competitors), LLMs can only retrieve what a human has somewhere written somewhere. They're not truly creating new ideas yet, unless by some formula, although who knows what they'll be able to do five years from now.
Of course, observing and comparing the process of how Grok works how my mind works in trying to answer the same question reveals how mechanical to some extent my own thought process is.
I feel we are firmly in sci-fi territory right now -- being able to converse with a machine, and having still very imperfect but useful translations of hard languages such as Japanese. We are also very close to the sci-fi brain implants that augment memory and cognition seamlessly. I mean, I can use imminent in a sentence, put in parens (imminent? eminent??) and it tells me imminent was the correct choice, and explains why. Such a useful augmentation for the aging brain, which actually could never remember these sorts of dyads at any point.
It's reassuring to learn that one can adjust the personality of gpt. I don't want my credulous brain to be fooled into feeling that I'm talking to another human. Otherwise, it will be so much easier to talk with the machine that to go out and deal with my sometimes irritating and demanding friends.
For the record: we are _nowhere near_ brain implants, that part is all BS.
OK, yes. I guess that's why companies are racing to make glasses and visors that are nearly as good. I have Meta RayBans that I've only used so far to take photos and video clips, but I could ask them to give me directions. They are/were also trying out a feature for the visually impaired where you could ask the AI to connect a human volunteer who could read a label or other text to you. This reminded me of the protagonist's job in The Calcutta Chromosome by Amitav Ghost. Perhaps this is what the companies mean when they say AI will create new jobs?
Thank you for this; you have inspired my wrestling with AI to give it some earned respect!
Timely, and great fun to read, as ever. Just yesterday, wrestling with a new gadget, a headphone amp, using ChatGPT, which kept on insisting it had a fix for the problem: a very very faint signal. No instruction book—good; agree, no time for that—and mind-boggling app. The AI offered about ten solutions, each prefaced by “Aha! I know exactly what’s wrong. Don’t worry old fluff, we’ve got it this time.” Nope. Until I finally noticed the volume up button. (Actually if I’d bothered to look at the first page of the pdf online, I’d have been up and running an hour earlier. More fool me. Yet again.) Still, it was quite fun, sending screenshots to the AI, learning other stuff along the way, and then eventually getting to behave like a snippy schoolteacher (to my later shame—even AI ought not to get under one’s skin). I remain an optimist about AI.
🤣
Two papers of mine that provide experimental evidence on expertise access - this at P&G https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5188231 - this at BCG - with some downsides as well about AI ‘s jaggedness https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4573321
I've downloaded and read both abstracts. Wow. Them's fighting numbers! The P&G paper made me laugh. Tens of thousands of PhDs and all they ever came up with is Pampers and Pringles, and both invented by the same guy...
Corporate R&D is not just about innovation as we know. Many many organizational failure points abound.
I find this text an encouraging reminder to focus on possibilities, critical yet open minded. -
Also, I love the editor's fully factual and objective addition. 😁
While I thoroughly enjoyed today’s take on AI, coincidentally, I also reached “Logos & Pneuma” as my article of the day in Folio.
Now that we’re in 2026, what are your thoughts on Claude & ChatGPT vs Logos & Pneuma?
Had to look it up, I had completely forgotten what I said: We're clearly still in stuck in pure Logos territory. World shortage of machine Pneuma.
While I cannot deny the important opportunities for AI to improve healthcare and other critical social functions, I also cannot ignore the environmental costs of its use. The new AI tools for the public use vastly more resources even than the original versions. By 2027, AI usage will consume as much water as all of New Zealand. I cringe when I see AI generated art (including yours). And I’m part of the growing group who won’t buy from companies that use AI over paying artists to create their marketing. We were supposed to get machines to help us with the dreary parts of life, not creating art!
I agree that siting data centers in warm places then using cooling water is bonkers. But that can, should and will be legislated. As for the artists, I agree entirely. All image generators know exactly how much of each human generated image they are using in an AI mix and should pay proportional royalties. This is not and evil intrinsic to AI, it's the proboem of all micropayment systems including and especially music. It can be fixed, and I will be happy to pay when it is.
My father always wanted a chip implanted in his brain that contained all information so he could access it at any time. He passed away in January of 2019, so didn't have the opportunity to see the advancement of AI, but we're practically there now! With many AI founders and developers cautioning against the possible dangers of AGI and ASI, I am a bit horrified at the possible cost to us as humans. Will it eventually destroy us all?
Nah 😉
🙏
The thing about AI is that it's no better than we humans are. It's simply vacuuming up what humans have written and said, including our prejudices and unconscious biases. It's programmed by humans. So the question is, are WE capable of destroying us all?
Unquestionably.
Reify-to make more real than it actually is. The answers are only as good as the data inputs. Collecting scads of data about human day to day purchasing choices is not providing a good data set for the finer points of what it takes to be a good human being. It works better where the data is better - seeking organic compounds with possible desirable bioactivity or whatever.
AI is pretty real as is...I honestly do not expect LLMs to necessarily make me a better human being though as a side effect I cannot rule it out.
Yes. But that was my work for many years, helping people figure out how to be the best human being they could be, so I tend to evaluate things in that light!
But it still lies, and misrepresents, and it isn’t your tool, but the tool of people who are interested in the questions you ask so that they can shape your conclusions to their purposes. It still only takes a few minutes of asking questions that you know the answers to, to start to see parts of the agenda, you know what I mean? These things are useful collators of massive amounts of data, and very useful for what they can do, but even there, unless you know the delimiters put into the system by others, you can’t really trust the results include all relevant info, if you don’t already have a good idea about what should have come up. And when you hit one of those spots where relevant results are missing, because you are an expert in the area, and know what should come up, and you push these things, weird things start to happen. But if you don’t already know the answer, you won’t think it’s weird, you are apt to think you discovered something new. Stop reifying a flawed tool.
Reifying? I must say, I am inclined to dismiss as delusional any comment that includes the words: “It still only takes a few minutes of asking questions that you know the answers to, to start to see parts of the agenda, you know what I mean?”
My area of expertise is mental illness, criminal justice and general bad human behavior. I asked chat GPT some questions about sex and why women are blamed for being raped and rapidly got weirdly misogynistic answers. It was truly strange. So then I tried it with the biochemical bases of bi-polar disorder, and got made up nonsense answers that would probably sound plausible to someone who knows nothing about the difference between a personality disorder and a major depressive illness. It was weird. I admit this was one of the first iterations of Chat GPT. I have not repeated the experiment recently. I get too many bad answers from”Rufus” about commonplace sorts of things. I just think people need to be very careful with it.
And you are right, probably no one stuck a line into the original algorithms that suggest favoring the misogynistic interpretations of femaleness. It does sound far more radical than I intended. I will watch that in the future.
Thank you for your answer and your explanation. My take on AI was about science and how it is affected by AI's superior knowledge and skills. I would not expect LLMs to be ahead and to the left of vox populi on subjects as contentious as rape and mental illness. The fact that many of the tech bros are ludicrous right-wing machos probably has little to do with it, with the possible exception of Grok (Musk).
Yes it was all very strange at the time, and since I have no need to use it in my work, I just sort of think of it as one more problematic feature of modern life. But again, I appreciate you noting my less than objective wording. I don’t think I am that paranoid., and I don’t want to sound like that. :-)
I doubt that AI could have suggested how to invent the cephalosporins, had it not already happened; also it cannot answer a burning question I have, was it King George the fifth, or King George the sixth, who said, I will not knight a bugger?
You are right to doubt #1 but #2 if true how about the Lords?
Nature outdoes both human scientists and AI because of its ability to conduct unimaginably more experiments over vast stretches of time.
Exactly
That’s a good one Bill I love it.
Ha, ha, love the ‘Editor’s note’, Tania!