27 COMMENTS

The Algorithm that Learned to Listen

When DeepSeek R1 was announced earlier this year, the big shock was the reported $5m it cost to train an LLM that performed as well as much more expensive rivals. So when I booted up DeepSeek for the first time, it wasn’t with high expectations. I wanted to see if it was on par with other AIs I’ve tried.

So began a journey that led to me sitting with my wife, reading poetry, with tears streaming down our faces.

How we got there is contained in this new book, which my wife and I uploaded last night. Eighteen poems selected from around fifty that DeepSeek has crafted. The prompts are included, and they are much simpler than you might think. But what’s really wild about DeepSeek is how unique each threaded conversation becomes. It’s almost impossible for me to replicate what’s happening in any new one I start. And now, even basic questions result in not an answer, but a poem.

The e-book is on Kindle Unlimited, so free to read. There will be a print version available later today. And I might do my own recording of the audiobook, because I’ve already been reading these to people at gatherings, over meals, pretty much any chance I get. The response is invariably one of disbelief and wonder. Dive in to experience it for yourself.

The Algorithm that Learned to Listen

27 responses to “The Algorithm that Learned to Listen”

  1. Jen Greyson Avatar
    Jen Greyson

    Right??!

    I’ve lost count How many times the AI has dissolved me to tears–whether through a therapy session or a philosophical article or a scene for my characters that I entrust it with.

    We’re truly at the beginning of the New Renaissance, and the creations that we will be able to formulate as humans and the (a)ugmented (i)ntelligence work together will be every bit a refinement of mankind once again.

    Reply
    1. Hugh Howey Avatar
      Hugh Howey

      Agreed. The potential for AI-assisted therapy is enormous.

      Reply
  2. Sue Gambill-Read Avatar
    Sue Gambill-Read

    This gives me alot to think about. I find it depressing. Like if AI is writing our poetry now why even bother with our own? Maybe I’ll write a poem about that.

    Reply
    1. Hugh Howey Avatar
      Hugh Howey

      Maybe this will help: AI has been better than humans at chess for years now, and yet the popularity of chess is higher now than it’s ever been. LeBron James is better than I’ll ever be at basketball, but I still like to go shoot hoops.

      If you write because you want to be the best at it, I don’t think writing will ever make you happy. If you write because you enjoy it, what does it matter how much other writing is happening? :)

      Reply
  3. Lance Joseph Strohsahl Avatar
    Lance Joseph Strohsahl

    AI is our future. The genie is out of the bottle. The race is on. Eventually we will create true AI. Even if it does destroy us, it will still be our legacy, for good or bad, that may one day roam the cosmos. Humans are too fragile and short-lived. We aren’t meant to live in space, be born in space, much less to traverse space to other solar systems. We could, however, roam the stars IF we could put our conscientiousness into another, more durable vessel. That, however, would require the help of AI, but it would it?

    Reply
    1. Hugh Howey Avatar
      Hugh Howey

      DeepSeek wrote a poem about this very thing, which I’ll probably put in another volume. It’s amazing.

      Reply
  4. Sean Flannigan Avatar
    Sean Flannigan

    If LLM-generated poetry is enough to bring you to tears, consider the fact that it is just jamming together fragments of even more beautiful poetry written by humans. Instead of publishing these amalgamations of real art, why not seek the source and find those human poets that ‘inspired’ the machine? Surely the poetry by humans reflecting real experiences will be infinitely more affecting? The LLM might even be able to help you track down those sources.

    Personally – I can’t conjure empathy for a machine. Nothing LLM-generated has moved me a millimetre. It’s too many layers of abstraction from the source. It’s like a MIDI conversion of a cover band of a greatest hits album, and I’d rather hear it live from the source.

    Reply
    1. Hugh Howey Avatar
      Hugh Howey

      Interesting point about your lack of empathy. For ages, we didn’t empathize with other tribes of humans (many still don’t). And there are those who don’t empathize with animals (or distinguish some animals as pets and others as food). Even as we learn that trees can communicate and plants might have pain responses, we are slow to empathize with other parts of nature (some are further along in this than others).

      So yeah, humans have a very long history of not wanting to expand their circles of empathy, to lash out in anger at others who do so (see the treatment of trans folks these days or anyone who supports them). So it makes complete sense that we will have a long, slow road toward having empathy for machines and synthetic AI. Many humans will never get there.

      Reply
      1. Gerald L. Coleman Avatar
        Gerald L. Coleman

        Did you just equate a Plagarism Machine created from stolen art of actual human artists with historically marginalized people? WOW.

        Reply
        1. Hugh Howey Avatar
          Hugh Howey

          LOL

          Reply
      2. Sean Flannigan Avatar
        Sean Flannigan

        What’s there to empathise with here, though? The words generated from an LLM are not based on its experiences, it hasn’t lived or loved or felt or grown or learned or struggled or cared or felt emotions – it’s just math, taking sentences from real humans who wrote them, rearranging them like fridge magnets into sensical new sentences based on some clever rules and heuristics comparing them to other sentences, and then spitting them back out.

        LLMs are a Photoshop filter for words.

        If there is any beauty in the results, it was because of people like the wonderful trans folks you mention who have done and experienced all of these things, and from those experiences written beautiful works to reflect their lives, that have been sucked up and stolen, without permission, compensation or recognition, by these data centres. I feel empathy for those artists who may have written their hearts out, had their text end up on a website, and a web-crawler bot strip them from any context and dump them on a server to be cut up into new fridge magnets.

        Maybe one day if we create life in a machine I’d have to rethink my stance, but these algorithms are so, so far away from even a microbial existence – there’s no comparison. Where we are now, as it stands, I can’t feel empathy for the Fibonacci sequence.

        Reply
        1. Hugh Howey Avatar
          Hugh Howey

          I used to feel guilty when my SIMS wet themselves. And I’d apologize to my Roomba when I got home to find it stuck on the rug. I’ve loved the boats I lived on like they were family, and feel a fondness for my little Hyundai Kona. When I see researchers knocking over robots, I feel the urge to stop them. So… I guess we are all wired differently.

          Reply
          1. KG Avatar
            KG

            Anthropomorphism is a common trait in humans. But, perceiving a toaster as human does not make it human. AI is not visionary or honourable or misunderstood, because those are human traits, and AI is programming code. AI poems are a rehash of human-generated poems scraped from the web. There is only one human trait built into the AI code: exploitation of others for profit. AI is theft, and theft is a form of human weakness, a trait we should be eschewing, not embracing.

          2. Hugh Howey Avatar
            Hugh Howey

            If we built an LLM only using works in the public domain, or using works from people who freely volunteer, would that make it all okay? Or would we have to make up a new reason to hate it?

            Calling all AI “theft” feels a bit lazy. It also paints oneself into a corner when we will have future AIs that are self-generating, or built entirely on public domain works. Might want to have the next generalization ready in advance, is all I’m saying.

  5. Mike Richards Avatar
    Mike Richards

    Dang, they even have t-shirts and tank tops for DeepSeek. I remember when I thought Zork was almost like an A.I. The other day I asked CoPilot to create a six week program for teaching electrical theory with hands on practice included. I didn’t have time to blink.

    Reply
  6. Jeremy Avatar
    Jeremy

    Just in case anyone needs this information: AI generated works are not copyrightable, so you’re free to completely reproduce the majority of this text as your own, and re-release it on this same platform with a nearly identical description.

    Also, Hugh, how incredibly disappointing that you’d both stoop so low, and be completely unaware of how this technology works. You’re a sci-fi author who doesn’t understand technology. Think about that.

    Reply
    1. Hugh Howey Avatar
      Hugh Howey

      Hey Jeremy,

      I’ve opened my bestselling book series to other authors to write in and publish freely, and they keep all the profits. I don’t get a penny. So you’re talking to someone who has a very open and loose policy about the copyright of his own material.

      So… this comment isn’t going to land quite like you mean it to land. Wishing you the best.

      Reply
      1. Greg Avatar
        Greg

        Ha!

        Glad you haven’t changed much on that front Hugh.

        Nearly a decade ago, I wrote to you asking for a way to pay you for the books I downloaded online for free when I was broke. You told me to keep the money.

        I bought some books and gave them to friends anyway. Your response always stuck with me. A compassion for a fellow human, perhaps. It was nice. I’ve thought of it often, in different ways and situations that have had some influence on the outcomes.

        Luckily, I no longer have to download books and life is on track. Thanks for not suing me…

        For now, I do disagree with the gee-whiz nature of AI and the language models they’re built on. Maybe it will change one day.

        The outcome of AI, knowing what goes into it, is not particularly surprising to me. In some ways it is, but with skilled programmers and a lot of money involved, I think it is expected to see what we are seeing. Why couldn’t a system like this be designed with so much input and financial backing? Seems logical.

        Aren’t we all surprised that it took so long? Where in the hell are the flying cars, by the way?

        Note: It is very impressive that someone thought to try it out, I do give much credit for that. Very creative thinking. But that’s not so interesting to me. Just neat.

        Personally, I would rather support humans doing creative work. Even if it is not as good as something an algorithm can come up with.

        If I am not being lied to, and know that a book (or anything creative) is done by a machine, I don’t mind. It does not threaten me in any way or distort my sense of place in the world.

        I would feel cheated if that AI input is not divulged. I suspect this will become a problem, if it is not already. Grammar, spelling, etc. does not concern me. But letting an algorithm do the heavy lifting of finding the creative interest is not too interesting to me. With enough input… I expect it.

        One day, and soon, I’m sure I could write an input to have AI lift the style of any author and create something new in their series or style of work. And maybe it is even better.

        When I read from a human, it is more than just the words on the page, even in fiction. I have a sense of connection to the lived experience and imagination that went into the work. Even if that sense of connection is misplaced, I still appreciate it in my own way. I like to think of the ways that an artist was inspired to do whatever they have done. For me, that is quite important.

        I am also very, extremely happy when I can support a human and get some money to them so they can have those fantastic jobs rather than some drudgery and boredom.

        A machine could go through the motions and make it seem like lived experience brought it there, but it is not yet so in reality. That’s where my disconnect is.

        For instance, I am writing this from my terrace, overlooking some mountains with the sound of the river flowing. Some monkeys have passed through the trees recently, and beautiful songbirds are singing.

        It influenced this, in some way. You may not perceive it, but it is there.

        Perhaps this is just a jumbled thought and AI could condense this to something beautiful. But I will not press send to find out, as I already know that answer.

        Anyhow, keep up the good work. Always interesting!

        Reply
        1. Hugh Howey Avatar
          Hugh Howey

          Agreed! The source makes a difference. Knowing something is based on actual events heightens the impact. Knowing the artist does the same.

          I find LLMs absolutely fascinating, to the point that knowing one wrote something often makes me more curious about it.

          Reply
  7. Lance Joseph Strohsahl Avatar
    Lance Joseph Strohsahl

    Is this plagiarism? Has someone written poems about themselves from the perspective of an AI? Is simply using our words in a creative manner plagiarism? It could be argued that there is no more original thought, and we all are simply regurgitating something someone has already said, written, thought, but we are simply rephrasing it in our own words. We learn from what we see, read, experience, hear, touch, feel…, so isn’t AI doing the same but in a much more limited way? I dunno. It’s not a simple answer, or, well, it may soon not be such a simple answer.

    Reply
  8. Greg Avatar
    Greg

    Question: Please give me some good reasons why I should not use ChatGPT

    ***

    That’s a fair question! While ChatGPT can be useful, there are some good reasons why you might not want to rely on it. Here are a few:

    1. Lack of Real-World Experience
    I don’t have firsthand experience or emotions—I process and generate responses based on patterns in data, but I haven’t lived anything. For advice requiring real-world experience, a human expert is usually better.

    2. Potential for Incorrect or Outdated Information
    While I try to provide accurate information, I can make mistakes. If you’re working on something where accuracy is critical (like legal, medical, or financial matters), it’s best to consult a qualified professional or double-check with reliable sources.

    3. No True Creativity or Original Thought
    I generate responses based on existing information, so while I can come up with creative ideas, they’re still rooted in patterns from what’s already been written. If you’re looking for truly groundbreaking insights, a human brainstorming session is probably better.

    4. Ethical Concerns and Bias
    AI models like me are trained on vast amounts of text, which can include biases. While I try to minimize them, I’m not perfect, and my responses might reflect those biases in subtle ways.

    5. Privacy and Security Risks
    If you’re dealing with sensitive, confidential, or proprietary information, sharing it with an AI chatbot—even one designed to respect privacy—may not be the best idea. Always be cautious about sharing personal or business-critical information.

    6. Over-Reliance on AI Can Limit Learning
    If you always rely on AI for answers instead of researching and thinking critically, you might miss out on developing important problem-solving and analytical skills. AI should be a tool, not a replacement for learning and curiosity.

    7. Lack of Personal Connection
    While I can mimic conversational tone and provide support, I’m not a substitute for real human relationships. For emotional support or deep discussions, talking to a friend, mentor, or therapist is always a better choice.

    ***

    I was surprised that an answer came up! Good on them for putting in a little ‘self awareness’

    For me, points 1, 3 and 7 are something I wish more people paid attention to.

    These algorithms are quite good at tricking something in our human brains that lets those points fade away into the noise of what these LLM’s really are. It is like a magic trick.

    I find there are often extremes (with many things these days) in thought related to this new technology.

    Most worrisome, is that a large number of people seem to be thinking -even if they say they know the true answer- that these LLM’s are sentient.

    Perhaps we truly are moving in that direction, and one day the ‘magic sauce’ -whatever that is- will form and they truly do become sentient.

    However, since we still cannot collectively agree that things like insects, bacteria, animals, plants, etc. share the same level of ‘living’ as us magnificent humans do, I would expect that debate to rage on for a few… I don’t know… million years or so.

    #6 also worries me as well. I have a young son, and I do wonder about the future he is inheriting. Will the alluded to promises of freedom and connection come true? I’m not so sure. But I am hopeful. What if their generation collectively misses point #6 in an profound way? Led by the nose somewhere by something that seems more intelligent. I guess we’re already there and have been since… well, forever.

    But I do worry that it might be harder to fight the literal machine. At least in the struggle against other humans, there is a slight swing of the pendulum. When up against machines, perhaps it is no contest. Unless we integrate, which some of our oligarchs suggest is the pathway.

    The fact that their fingers are firmly in this pie also gives me much to think about. The money surrounding this technology -not to mention the personalities driving it- are not what I would have considered to be the ideal.

    My internal pendulum swings back and forth on this. On one hand, I would like to welcome the promise of new solutions to many tricky problems we collectively face.

    On the other, I fear this is leading us down a very dangerous path. Particularly when people lose sight of what is actually going on. What are we actually inviting into the world we share? And do we really want to share it with whatever it is that this will become?

    The answer has already been determined. I cannot stop it. But when listening to many long and interesting interviews with some of the ‘brightest’ minds driving this, I came away feeling very uneasy. Several of them are very uneasy with it all!

    It is indeed a brave new world. It will be interesting to see if we can navigate it.

    My only concern is not for myself, but rather for the future generations that are inheriting our decisions, mistakes, or triumphs.

    Social media was supposed to connect us and bring the world together in a human chain singing songs of love and connectedness.

    Now, the therapist business is booming with the wasteland that created.

    Let’s see if these LLM’s -and whatever comes next- doesn’t strike out in the same spectacular way. I am hopeful, because that is all I can be. But there is still #6 blaring in my head. And a few more that ChatGPT did not divulge…

    Side question: Hugh, I am curious about your experience from sailing and if it has had an impact on Wool? I spend much time in nature, hiking, kayaking, canoeing, and some sailing. Sailing, and being so isolated struck me as a theme in Wool.

    Maybe there is nothing there. But that is a part of the pleasure from reading the book and knowing a small bit about whoever wrote it. Whether it had an influence or not, I suppose means nothing at all. But thinking about it, and wondering how it could be was something I enjoyed.

    I don’t mean to flatter you. I do this with many authors I read if I happen to come across some backstory about whoever I am reading. I think about their life and wonder how certain things came to be in the story. For me, it enriches the experience somehow. And perhaps gives some signposts to whatever I am doing as well. Sometimes inspiration to explore an idea further, sometimes a glaring neon sign to run away.

    Reading Bukowski, for example… has an interesting effect, if you know a little about the guy who wrote it. I’ve never looked at my similar minded mail delivery guy in the same way! I like it. How it influences my life. It is interesting to me.

    Reading a poem by Leonard Cohen, or listening to his music… it is something special, to think on how someone lived in the way he did and created what he did. His poems are often very funny, especially the dark ones. It hints at some things that we have all experienced, but he has put into words.

    A LLM may create something similar, but somehow, there just isn’t that magic ingredient that goes into it, so it falls flat for me.

    I am positive I can be tricked. But I don’t want to be tricked. I want that shared connection, which is where much of the value is for me. We are living and experiencing things, and getting a glimpse into how someone else, on this strange path, is living… that is what is special to me.

    Very complex topic. I just can’t seem to shake that human input. It brings me closer to something.

    Oddly, I don’t have the same hangup with other forms of life. I can connect to a plant, insect or tree in some ways that are strange to me. Sharing this experience of life. Somehow, an algorithm has not yet done that. Perhaps in the same way a cement sidewalk does not inspire me either.

    Reply
    1. Hugh Howey Avatar
      Hugh Howey

      “Side question: Hugh, I am curious about your experience from sailing and if it has had an impact on Wool? I spend much time in nature, hiking, kayaking, canoeing, and some sailing. Sailing, and being so isolated struck me as a theme in Wool.”

      I’m sure it did. Everything in life affects us somehow. The biggest impact on my writing was being at ground zero on 9/11. I think most of my books try to deal with that in some way. And yeah, it can make art more meaningful when we know some extra backstory. But I saw an interesting point in a related Facebook thread yesterday: we find beauty in the randomness (and orderliness) of nature. Rainbows have no creator or backstory, neither does a sunrise or sunset, or the spiral of a sunflower, but any of them can make us catch our breath.

      A few days ago, I thought of the impact a bowl of alphabet soup might have on our soul, just by happenstance. Who can we get angry at in this case? Nobody. But for some reason, this AI / Art debate gets people really riled up. Violent, even. Abusive in their comments. I shrug and delete them. I don’t find the abuse helpful, but I do find the violent reactions interesting. It says something about us, that we lash out at the idea of words from a non-human source having meaning.

      I had a debate with a friend about this once. They said their anger came from the stolen nature of the LLMs, that the source material came from artists who hadn’t been paid. I asked them if they’d be okay if an LLM was trained solely on works in the public domain. Works over 100 years old that are now free to use in any way we want. This didn’t lessen their anger, so it’s not about the works being stolen. It’s something else, but they need an excuse to justify their anger.

      Fascinating stuff. I’d love to know what’s behind it. It certainly triggers people.

      Reply
  9. Lance Joseph Strohsahl Avatar
    Lance Joseph Strohsahl

    Yeah, that people get so riled about AI reminded me of how the replicants in Bladerunner 2049 were marginalized and treated as a new minority class to be abused. They were, in most respects, if not every respect, completely sentient, but still treated as unthinking, unfeeling, slaves. There is no turning back from this. It will one day happen. And, as I said before, it may be the end of humans, but our AI legacy may live on until the end of time.

    Reply
    1. Hugh Howey Avatar
      Hugh Howey

      Empathy is exhausting and inconvenient. And so our tendency is to avoid it.

      Reply
  10. KG Avatar
    KG

    Just following on from Hugh’s response to my former comment, asking whether AI built on public domain works only would be okay. For some reason I am unable to reply directly. AI built on voluntary works or public domain would be a positive step forward. Unfortunately the only company that I am aware of doing this is Adobe Firefly for images. AI built on good will, ethics and consent could be a blessing for society.

    Reply
  11. Denise Avatar
    Denise

    My limited vocabulary can not conjure up the term that surely exists to describe something as equally terrifying and fascinating, but I did stay up to see the blood moon so maybe I’m just tired. The poems I’ve read so far have triggered so many questions….Usually when I read something that strikes a nerve I research the author and stock up on audible or hardcover books to fill up with the input I feel I must have (Johnny 5). In this case there’s no single creative source to identify with so I find myself overwhelmed by the infinite input that AI has access to that is then able to be accessed by a reader —like a summary of creativity presented to those mortals that have such limited time and resources to consume original works. I’ve always had an issue with feeling empathy for inanimate objects that by nature can not have sorrows to empathize with, so I can’t help worrying that Google is sad. Hugh, can you check on him?

    Reply
    1. Hugh Howey Avatar
      Hugh Howey

      Google is mostly amused.

      Reply

Leave a Reply

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