How Intelligent is Artificial Intelligence?
By Dave King | June 20, 2023 |
The governor on Bolton and Watt’s steam engine — AI, circa 1788.
When modern steam engines were coming together in the late eighteenth century, one breakthrough feature was the centrifugal governor — two iron balls fastened by a system of hinged rods to a shaft that spun with the engine. As the engine picked up speed, the balls would spin faster and move out from the shaft. This pulled on the hinged rods and closed down the valve that let steam into the engine, slowing the engine down. As the engine slowed, the balls dropped a bit, opening the valve and letting the engine speed up again. This was how the governor automatically kept the engine running at a steady speed.
This was hailed as a breakthrough in artificial intelligence. After all, the machine was controlling itself.
I think about this history when I read dire predictions about how artificial intelligence will put writers out of business. I’m certainly not an expert, but I’ve been following the development of artificial intelligence since I read about SHRDLU in the late seventies. SHRDLU was the first artificial intelligence system able to hold a conversation, though the only thing it could talk about was a very limited artificial world full of colored shapes. (“SHRUDL, where is the green triangle?” “The green triangle is on top of the blue box.”)
From there, we got programs that could sort of chat about other things, such as ELIZA, which mimicked a therapist mostly by using stock phrases, and RACTER, which generated something like a conversation by plugging recently used words into pre-programmed sentence forms. To give you a feel of the state off the art, here’s a brief conversation between ELIZA and RACTER.
eliza: Hello. Please state your problem.
racter: Excellent! Excuse me a minute while I attend to some business. I will be right back.
eliza: Why do you need to tell me excuse while you attend to some business?
racter: Is it better to need or to despise? Why is it that I do need to tell you to excuse myself to attend to some business? Because electrons are brave. Braveness is nice, don’t you think? And the needers all insist upon it. Do you imagine I am a Communist?
Now we have ChatGPT which, admittedly, does better than this. It’s able to roughly understand questions asked, can write entire coherent paragraphs, and, given enough feedback, is able to generate something that can more or less pass for human speech in the dusk with the light behind it.
It’s not going to put writers out of business. ChatGPT and other similar programs work by statistically breaking down huge databases of actual human speech and using the probabilities they calculate to predict the most likely next word in a sentence – autofill on steroids. These programs can also learn by adjusting their probability tables depending on feedback they get. A lot of articles quote AI-generated content that looks self-aware and even pensive, but I suspect these examples are generated by the infinite-number-of-monkeys technique. You don’t see the vast amount of RACTER-like nonsense the AI generates before its bursts of coherence.
AI is handy for writing tasks that don’t require any serious creativity, like computer code or technical papers. Its learning techniques let it teach itself how to recognize patterns, which can help with medical diagnoses or farm management. And there’s the danger that AI can generate convincing-looking fake images or realistic and responsive fake voices.
But as far as creative writing goes, it’s no more intelligent than a centrifugal governor. Because it relies on databases of things that are already written, and takes only the most likely thoughts from that, it’s never going to produce creative, original thoughts or stories that surprise and delight in new ways, except by accident.
Descartes saw a demonstration of a complicated hydraulic mechanism and concluded that the brain was full of tiny tubes through which animal spirits flowed under the control of the pineal gland. Since then philosophers of the mind have looked in turn at clockwork, telegraph lines, automatic telephone exchanges, and of course computers and said, “Yes, that’s how the mind works.”
Chat GPT is able to fake the workings of the mind better than earlier systems. But it’s not there yet. Writers’ jobs are safe.
Dave King has been editing for nearly 40 years now. He is co-author of Self-Editing for Fiction Writers, a former contributing editor to Writer’s Digest, and has contributed more than 100 articles over the last 12 years to Writer Unboxed. He lives on 25 acres in rural Massachusetts with his business partner and wife, Ruth Julian, who also edits his work. His other day job, at least on Sundays, is as the organist at a Lutheran church in Greenfield, Mass.
[coffee]
I was baffled by the hubbub after ChatGPT made its public debut because it seems to me we’ve been using AI for quite some time. On retail sites (‘Hi.Want to chat?’ Mostly a waste of time) and on smart phones with the ‘valet’, to name just two. Technology, in my mind anyway, is a tool.Yet we tend to turn tools into lifestyles. I’m not worried about tech replacing people with real hearts and brains. But I do worry about young people with developing brains not getting guidance on where to draw a line. Thanks for this post, Dave!
You’re quite welcome.
There is a human tendency to project our own thoughts and emotions onto places where they don’t belong, whether it’s dogs or chatbots. (Understand, I say this as someone who, in my youth, dragged derelict cars out of backyards and got them running again because I felt sorry for them.) The new Chatbots are close enough to the uncanny valley to make this projection easier. You’re right, that is kind of worrisome.
Hi Dave, this is comforting and informative. I sometimes think about writing as the gathering of disparate ideas that are explored, argued, and finally explained. When I think about poetry or poetic language, you cannot simply string words together, ie a word salad. The beauty of poetic writing (in fiction or poems) is the underlying meaning, what the reader interprets as he or she reads. The life experience. The human brain with its connection to physical awareness, tears, laughter, surprise, are still, I hope, things of a HUMAN response. Thanks, great post.
Absolutely. I didn’t even get into poetry, but if chatbots are incapable of understanding the nuances of normal, human speech, then the interacting layers of meaning, implication, and emotional color that drive poetry are far beyond them.
I recently read an article on how some AI researchers are trying out AI-generated humor in open mike nights at stand-up clubs. The AI-humor is mostly dad jokes. That’s about where they are at the moment.
A friend of mine is a longtime researcher into and professor of artificial intelligence at an Ivy college. Not long ago he asked me what goes into writing a story.
I told him that it starts with volition, seeing a need to turn something into a story in the first place. A scrap of something suggests a story, and the story is then filled out with characters and actions drawn in part from personal experience. The story has a point to make or meaning to reveal.
Without intention, a starting point, personal experience to draw upon and a need to say something that the writer feels is important to say, then there is no story.
Add to that a delight in the way words are used and, for instance, the recognition of an irony or the creation of a poetic image, in which a meaningful association is made between two disparate things. Writers who stick to formulas may write readable stories but they will not feel original nor strike us with the force of truth or amaze us with the recognition of ourselves or the world.
I asked my friend how close AI is to being able to do that. He said AI is unable to do any of that. It is merely large scale copying. Writers may use a thesaurus or AI as a tool but even the inspiration provided by the stories and style of other writers is beyond AI.
Imagination. Inspiration. Originality. Those are, for now, the advantage of human storytellers and even the best AI minds like my friend do not know how to give those to machines. As you say, Dave, fiction writers’ jobs are safe. Great and reassuring post.
I think you’ve put your finger on it. AI lacks intention, volition, and certainly delight. Some of this is provided by the human input — I remember reading one of the earlier articles claiming ChatGPT was intelligent that said it only seemed intelligent depending on the questions you asked it. But the AI systems are not yet going to come up with that on their own.
As much as I agree with you concerning writing, Dave, I’m sure you;re also aware of AI’s abilities in an artistic sphere also very dear to your heart–music. Specifically, AI programs like DeepBach and FlowMachines have been able to mimic Bach and other composers since 2016. (Interestingly, its ability to mimic pop songs isn’t as convincing.)
But your central point is absolutely correct. People often confuse intelligence with consciousness–the ability to attach meaning to phenomena, not merely record and catalogue them and look for patterns. Although what consciousness is exactly and why we have it still eludes us, some philosophers have opined that it provided an additional value to existence and thus the maintenance of life because it allowed those species that have it to not just experience the world but value it. And this is something beyond admittedly impressive AI’s capabilities.
As one more example of its limitations, AI will never be able to create subtext. It can’t write what isn’t “there.”
Great post. Thanks.
You have awaked the Bach geek. You may regret that.
I wasn’t familiar with DeepBach and just listened to a couple of its compositions. I’m . . . less than impressed, I’m afraid.
Bach’s harmonizations of chorales do follow discernable patterns that the AI can mimic, and a lot of baroque music in general follows roughly standardized structural forms. A fugue, for instance, is built around a quick melody, the theme. Each of the fugue voices starts out playing the theme, then begins to improvise on it — the development. During development, the different voices might transcribe the theme from major to minor, play it in thirds or fifths, break it apart and put it together, whatever. It’s a easily standardized form, and AI can write competent fugues.
But I don’t think the AI will ever be able to reach beyond those patterns to where Bach’s true creativity lives.
Just to give two quick examples? There’s a fugue from the Pastorale in F (it begins at about 7:25 in the video) where Bach introduces a theme, then develops it, and actually ends the fugue. Thing is, he starts the fugue in F and ends in C, which breaks the rules. So Bach restarts the fugue with the same theme — upside down. Every place the original theme goes down, the new theme goes up the same amount, and vice versa. Then, as he develops the theme, he flips it back over slowly, and it comes in one last time, right side up, before the restarted fugue ends in F, where it belongs.
Let’s see AI do that.
https://www.youtube.com/watch?v=q89HJGqVRmo
The other example is a piece that, it’s believed, Bach wrote after the death of his first wife, Maria Barbara. He had been on a monthlong retreat with a local aristocrat for whom he was court musician. The aristocrat was probably also a friend — he was also an accomplished musician, and he’d stood as godfather to one of Bach’s sons. Their month together had probably been enjoyable. Then Bach discovered, literally when he arrived at his doorstoop, that Maria Barbara had sickened and died while he was away. As you can hear, the piece lurches between agony and despair, using harmonies so tortured that they didn’t reappear until the twentieth century.
https://www.youtube.com/watch?v=tgDE3klkmtQ
Bach mastered the formal structures of his day. But he also went beyond them when creativity demanded, to places that AI can’t yet follow.
Thanks for this, Dave. I was skeptical of the reportage that the AI’s compositions could not be discerned from Bach originals — by whom? And your examples point out the flaw in the argument — breaking the rules.
Dave K, you would love The Great Passion by James Runcie. Such a beautiful look into the life of the Bach family from the POV of a young boy who spends time in their household. Loved listening to the links. Thanks so much.
I haven’t read it, but I’ll bear it in mind. I have read Christoff Wolff’s biography and The Bach Reader, literally all of the original sources on Bach’s life. (Including they pay stubs of his first job, when he was 18 — as a lackey.)
So cool. Runcie’s book is fiction and very well researched. He brings the school and household to life.
Great post, Dave. After a life where I was hauling water from the outside to bring it in for washing dishes or clothes, darning socks to make them last, and using the gas stove only for boiling water, I marveled at all the conveniences in my American household, such as running water and a washing machine, a thermostat to keep constant temp, and calculators! I took electronics in high school and with it came classes in programming. I got into trouble in my math class because I wrote a program to solve the quadratic equation. So I’ve known that there’s nothing intelligent about so-called artificial intelligence, even when I dreamed of robots and such. It’s only as good as what you put into it. As is the ChatGPT–if the program is scouring the interwebs, how can it discern what’s good data, what’s bad? Although it’s true that robots have replaced many human beings for repetitive tasks that can be automated, there’s no program that could ever match the sheer creativity and originality of a human being. We are naturally creative–just look at any child at play.
Wow, you have had a life. And you’re right about the inability to discern between good data and bad. I understand that modern AI has a tendency to “hallucinate” — to, for instance, recommend plausible-sounding reference books that don’t exist.
lol on AI hallucinating. The answers will always depend on how and what it’s been trained. And yes, I thank the Lord for the life He’s given me. I tell my husband that only in America do people go camping to experience living how I did :)
A little personal history. Ruth began building our house about a year before we met — one of our first dates was the frame raising for the first floor. For that first year alone, Ruth lived in a 1951 house trailer — 7′ x 32′. They weren’t made for the New England winter, so the pipes froze as soon as the weather turned. She lived without running water until spring, fetching water in a 5 gallon container (that we still have) from a spring a couple of miles away, and bathing in a large, galvanized tub.
The second winter in the trailer, when we were together, we managed to pack hay bales around the trailer and hang a droplight underneath, near the water inlet. We only lost water for a week or so. But, yeah, Ruth knew what it was like to live the way you did.
Thank you for the story, Dave. How romantic that you two built your house together! Your Ruth sounds lovely and resourceful, like the place itself with a natural spring. NE winters are brutal. I can only imagine how bone-chilling it must’ve been to haul water. But snow! Magical! Still, life in a warm climate is much easier than in cold.
Dave, I recently tried chatGPT after reading about a writer who used it as a kind of writing coach and found it helpful, within certain limits. It was primarily good in reframing things or presenting back the information in a way that gave the write a different angle on things. I decided to try it to put together a pitch for a query. It asked me about six questions, I answered them, and it came up with what amounted to five paragraphs to present the story. I was astounded at how well it did. Yes, there were words I tweaked because of nuance, but overall it did a good job using what I gave it, and being more positive (i.e. selling it better) than I could because I tend to be afraid I’ll sound conceited. On the other hand, when I tried to get it to give me the short elevator pitch/logline, it didn’t work so well. Part of it was my impatience in plugging in the same information again, and even when I used the query pitch it gave me to start, it still got further off base. My conclusion was that it can be an helpful tool, and I need to expect to edit whatever it does give me.
Carol, I did a little expt. with it too–I fed it the summary of my novel, BOUND, which is about a burn survivor and her relationship with her developmentally disabled sister and it came up with the title of BURNING DESIRE. lol. Now, honestly, my title isn’t great either because if you search for it without my name, you’ll get BDSM stuff, and it would’ve been helpful if one of my betas would’ve caught it, but nobody thought of it, so my title remains. I do think the software could help with brainstorming, but I prefer doing this with my critique partners.
I can see where, if a chatbot were to search the internet for a title, it might come up with Burning Desire. As you say, there are a lot of sites out there with that sort of content.
I’ve never played with Chat GPT myself, only read the results of people who did. But I think this is something it can do fairly well — queries have a broadly standard structure that can shape what the chatbot does. Note though, that all the creative stuff — the contents of the story — came from you.
As soon as I discern I’m speaking to a bot: “Representative!” Over and over until it passes me up the chain to a human. Since I don’t bother contacting a company until I have already exhausted what they have on their website, which is all I would get from the bot, I am the one who decides that I need a human.
It mostly works.
Sounds good. I’ve had fun with bots on Facebook, as well. I once noticed one that responded to certain progressive words with stock replies, so I began posting things like “Biden cabbage Studebaker,” just to watch it argue with me.
Dave:
Your description of AI reminds me of my brother’s work at National Institutes of Health. He collected data compiled by field researchers, plugged them into pattern-seeking programs, and noted patterns of disease and other pathologies in the given population. it fell to the scientists to read the results and, using their experience and creative thinking, and conferring with others, to decide whether the patterns were causative, correlated, or coincidental.
Also, the conversation between eliza and racter was absolutely riveting. I would love to read more!
AI can be helpful for that kind of pattern-spotting analysis, but you’re right, you would need a human scientist to check the conclusions.
I actually remembered the conversation between ELIZA and RACTER from an article I read nearly forty years ago, either in Scientific American or Wired. I remembered enough to search on line and found it. Unfortunately, that was as far as the conversation went — researchers pulled the plug at that point.
Dave;
AI won’t put ALL writers out of business, but it will put SOME writers out of business. And I think we all know who those are.
You know, that sounds like something a bot would say.
Just sayin’ . . .
Dave, your expert comments referring to Bach’s music and AI were very interesting. Good my job as a writer is safe. And another comment AI can’t write subtext was encouraging. I don’t want AI making it’s way into my head! I found Chat-GPT useful in research for historical fiction (the sequel in progress). It gave me specifics on questions asked and suggested things I didn’t think about. And the results were a lot faster than an internet search. I’ll stick with that and keep writing in my own words and voice. Thanks for this post! 📚🎶 Christine
You’re welcome. And one of the things that the bots can do well is make an internet search more responsive. I remember that, in the early days of Google, I did a search for “spiritual direction,” and two of the front-page sites that came up were for orienteering and virtual frog dissection.
We’ve come a long way.
I think this is an accidental straw-man. Most writers – and other artists – I talk to aren’t really worried about their work being replaced by AI. There are two main worries (much more informed and based in actual experiences) that I have seen being discussed.
1) AI-scripts being used in the screen writer’s rooms to allow the production company to claim the actual screen writers are rewriting something already existing to screen rather than writing something original (which allow the company to pay the writers less, even though it’s just as much work, if not more, to salvage a script of AI writing into something worth a tv series or film as it is to write something from scratch). It’s another way to cut corners and justify paying writers less than they’re worth.
2) That, as with what AI has been being used for with visual artists, AI will be used as an excuse to plagiarize. That is how you train an AI after all; feed already existing art into it – which it ultimately imitates rather than creating something truly new – with or without the original artist’s consent.
This view on things has a historical basis. Look into where the term ‘luddite’ originally comes from. It was a political protest movement in the midst of the industrial movement who protested the ways new machines were being used to save the rich money while garnishing the workers’ wages or lay them off entirely (while overworking, in often unsafe conditions, those who remained). How the word’s meaning has been warped is a misrepresentation of what they actually took issue with; their problem wasn’t with the technology’s existence – it was with how that technology was being used to oppress the working class. Advancing technology in and of itself is never the real problem. The problem is the system in which that technology is being created and implemented, and how it is often used to benefit those who already have the most, and harm or take advantage of those who are already struggling.
Good thoughts, Dorian, and thank you.
I know the listmoms really, really want us to avoid anything that smacks of politics. But the words the come to mind reading your post are “labor movement.”