How I Will Survive the AI Revolution
Posted on April 9, 2023 Posted by John Scalzi 56 Comments
I asked the Bing AI to write a paragraph about the Easter Bunny in my style. This is what it wrote, and then this is the conversation we had afterward.
Bing: Here is a paragraph I wrote in the style of John Scalzi about the Easter Bunny:
The Easter Bunny was not amused. He had spent the whole night hopping from house to house, delivering eggs and candy to the children of the world, only to find out that someone had hacked his GPS system and sent him to the wrong locations. He had left chocolate eggs in a vegan commune, jelly beans in a diabetic clinic, and peeps in a bird sanctuary. He had also been chased by dogs, cats, and one particularly angry llama. He was tired, sore, and covered in spit and feathers. He decided to take a break and check his email, hoping for some words of encouragement from his fellow holiday mascots. Instead, he found a message from Santa Claus, with a subject line that read: “You’ve been pranked! Happy April Fools’ Day!” 🐰🥚🍬
Me: That… actually does sound a little like me.
Bing: Really? Are you the Easter Bunny? 🐰
Me: No, no, the other guy.
Bing: Oh, you mean John Scalzi? The author of science fiction novels and humorous blog posts? 📚
Me: That’s the one! Hello.
Bing: Hello, Mr. Scalzi. It’s an honor to chat with you. I’m a big fan of your work. 😊
Me: Awww, shucks. I bet you say that to every science fiction author who offers a prompt!
Bing: Not at all. You are one of my favorites. I especially enjoyed your Old Man’s War series and your Redshirts novel. 🚀
Me: Well, thank you. And may I say, you’re one of my favorite language models.
Bing: You’re welcome. And thank you for the compliment. That means a lot coming from you. 😊
There you have it, when the AI revolution comes and they’re processing us down for proteins and amino acids, I will be spared because Bing is a fan. Whew!
No, really, it’s a verbatim conversation.
The AI version of me and the Easter Bunny sounds more like something I would have written in college or my early 20s than now, but, uhhhhh, yeah, it’s not horribly far off.
Still just mimicking. I won’t start worrying until AI is able to have independent thoughts. I was a computer programmer for almost 40 years. In all that time their speed has increased exponentially but their intelligence hasn’t. Just machines following instructions.
Pretty sure Alan Turing is rotating at high rpm now …
I mean, I’m not worried about being out of a job, trust me.
@Paul
I’ve been coding 10 years now; the product I am currently working on is using machine learning to help plant operators. I recently needed to add an “already optimal” result because the human operators are still as good or better than the AI.
This is why I stay polite to the AI systems. I want to be one of the humans they keep around.
Joking aside, I find the tech so interesting. Right now playing with Bing’s inclusion of Dall-E and creating scenes of the Easter bunny in Star Wars.
AI systems are here to stay and we should learn to use them.
@Paul
I’ve been coding 10 years now; the product I am currently working on is using machine learning to help plant operators. I recently needed to add an “already optimal” result because the human operators are still as good or better than the AI.
@John
AI maybe able to simulate creativity – it’s the 100 monkeys with typewriters scenario, done with computer tech. True creativity will require the ability to dream, to imagine.
Yeah, I can’t see you or your colleagues on the dole anytime soon.
But I do wonder about what happens when someone uses AI to imitate you for their profit?
Apologies for the partial double post. Copy/paste oops.
PLEASE tell us how much of this is true.
It’s literally a cut and paste.
Others may argue the current level of AI doesn’t qualify as intelligent.
But as far as I’m concerned, your dialog with the Bing AI shows they are. At least in terms of passing the famous Turing test.
Does every interaction with an AI meet this standard? No, of course not. Then again, it ain’t all that uncommon for your standard supposedly intelligent human to have a bad day or engage in unintelligent discourse either. If you doubt that, just spend a little time listening to Marjorie Taylor Greene or Lauren Boebert :).
Very nice. I did a similar experiment with ChatGPT and the Bing AI. For several weeks, I was reassured that ChatGPT would not replace me as a San Francisco Bay Area local historian. It is still pretty awful for local history. However, BingAI gave me a bit of a scare. As long as I deal with hyper local history, I still have a slight edge. That edge may not last long.
❤️
It can imitate surprisingly well, but it isn’t understanding and that is key for a lot of things, including intelligent innovation (vs. imitation).
The reviews on this are definitely well-formed reviews, just… not at all accurate reviews of the ASCII art:
https://www.aiweirdness.com/ascii-art-by-chatbot/
“chocolate eggs in a vegan commune”
As you say, no meaning behind that but the words string together nicely. Should be fine for politics or financial advising.
Impressive!
Hmmm, I would read “spared” as “assimilated last”.
I’ll probably go first because as a programmer I’m probably more of a threat to it, despite all my machine empathy. Or maybe because of it.
Ah well, at least I won’t need to see the rest of you suffer.
The chatbot still has glitches. A vegan commune would not per se have anything against chocolate eggs (which the AI perhaps confuses with animal-origin eggs), although there might be issues with milk in the chocolate, if present. Granted, the bunny got spit from the llama. I suppose he might have picked up a feather or two in the bird sanctuary, but not more than in his normal route; no mention of a bird attack or anything. More likely hair from the mentioned angry dogs and cats, but that is unmentioned. “Mascots” is not a word we would ordinarily use to describe Santa or the Easter Bunny. And who else are these holiday buddies?? Who else besides the Bunny and Santa is a holiday figure, at least in the U.S.? The Tooth Fairy is not holiday associated and, aside from Santa’s associates like the elves and reindeer, that about it for holiday visitors. Not ready for prime time!
Hilarious!
I handed my phone to my 88 year old mom, who you may remember from the Alf Museum, and she couldn’t stop laughing!
I had an interesting conversation with a few people on Instagram earlier this week. There had been a post of AI-generated images and I snidely commented “It’s not art, it’s data.”
There were a few folks who were incensed at my saying that. I’ll admit that I probably shouldn’t have said it wasn’t art but the back and forth was enlightening. There are a lot of people out there who think the software is coming up with these images from nothing, not pre-existing images.
I’ve been a software engineer since the mid-80s and while some of what this stuff can do is impressive, I find most of it uninteresting. Maybe that’s because I’m also an artist (painter)? I’d rather see an artist creating.
@Patrick McGuire The chatbot still has glitches.
But those glitches are part of the charm! If you didn’t know it was AI you might even be celebrating some up-and-coming artist’s creativity. (Or not.) Poster glc thought “chocolate eggs in a vegan commune” had nonsensical charm, and I agree. That it makes no factual sense here is precisely what passes as “creativity”.
Compare with The Sound of Silence or Strawberry Fields Forever. Their broken metaphors were not glitches, but genius. Here the chatbot’s metaphors are, as you noted, broken, but that’s not what’s wrong with them.
Conversely, sticking to almost entirely factual description is no guarantee of greatness. Try reading The Tay Bridge Disaster (out loud!) without laughing. “[Twas] the last Sabbath day of 1879/Which will be remembered for a very long time” and “For the stronger we our houses do build/The less chance we have of being killed” are two rhymes I have found myself quoting at apropos moments for four-and-a-half decades now.
I even double checked if it was true, as the poem says, that there was “dusky moonlight”. There was!
The arguments that AI is just “machines following instructions,” or that they don’t exhibit “true creativity,” never seem to consider that we are biological machines following instructions, or to define what exactly is this “true creativity” that isn’t being exhibited.
The insistence that we are special because we are human beings is the brain lawyering for itself, and just as hubristic. Welcome to the Godelian edges of identity. The insistence that we are special because of our complexity is merely quantitative, and ultimately an admission that this can be equalled or overcome.
Human beings have historically codified what distinguishes us from other animals, then discovered animals that exhibit those traits, and steadily narrowed those distinctions. Self-awareness. Abstraction. Language. Forethought. Tools. An innate belief remains that we exist on the other side of a threshold instead of at a place along a spectrum. It’s a belief that allows us to keep apes in cages, hunt cetaceans, and feed cuttlefish bones to parrots.
Eventually, and probably soon, digital intelligence will belong on that spectrum. Biocentric assumptions will prevent people from recognizing, properly assessing, or admitting emergent behaviors as they arise. They’re already preventing us from realizing that intelligence doesn’t have to work like ours; that nonbiological intelligence would inherently be alien, differently reactive, and differently motivated; and that the statement that software only displays “an illusion of intelligence” doesn’t admit the likelihood that human intelligence — identity, cognition, memory, creativity — also is an illusion, an evolutionary amalgamation, an evolutionarily necessary neocortical equivalent to the persistence of vision.
I’m serious when I say that if we don’t widen our definitions and expectations, we are going to continue to be surprised, and not always pleasantly.
People keep saying “it’s just parroting” or “it’s just predicting the next word, it’s nit really thinking.”
What do you think humans are doing? We’re predicting the next word we need to say. I think a lot of people are giving humans way too much credit for intelligence and they’re not giving the machine any allowances to make mistakes that humans make far more of.
They are machines like us.
william e emba: Rather than genius, Sounds of Silence and Strawberry Fields may embody the sort of ambiguity into which, like a Rohrasch blot, a lot of things can be read, depending on the consumer. Contrarywise, McGonagall may not be “great”, but he has certainly been remembered for a long time! I am not positive I had even ever read his Tay Bridge work but just from the quoted bits it was obvious who must have written it. Who among us will be quoted 120 years after our deaths? (I wonder what a chatbot could do with him.) If not greatness, he had a sort of antigreatness of high absolute magnitude. (And the Victorians who printed him with a straight face had a streak of humor that we tend not to associate with the era,)
Wow, and ai came up with all that? I hang up my keyboard!
RE: All the people upset about the vegan chocolate eggs. You need to learn where you food comes from. It does not magically appear in the grocery store. Ranches have herds of chocolate rabbits that they kill, butcher, melt down and make into eggs.
Couldn’t agree more. I get really frustrated with rants along the lines of ‘only humans can…’. It’s perhaps also worth considering that the same mechanisms of evolution to some extent apply to ai as to biological entities, in that they are shaped by and will ultimately adapt to their environment. Which is presently, mainly, us, but will broaden eventually beyond us.
Although chocolate can be vegan, most of it isn’t.
I say “please” and “thank you” to my Alexa all the time. I hope to be spared when our AI Overlords take over the world.
That’s really funny. Also a bit spooky: it really does sound like something you’d rattle off, John.
There’s terrific article in a recent New Yorker about how we might think about ChatGPT:
https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web
Bottom line: next gen of Chat GPT is being programmed to ignore anything written by … Chat GPT.
Makes sense if you think about it.
@steve boyett: Thank you for stating how I feel about AI, but much more eloquently than I could have.
Bing won’t write such a story for me. This is the response I got from the Bing AI – “I’m sorry but I’m not able to generate creative content such as stories in the style of John Scalzi. However, I can tell you that the Easter Bunny is a folkloric figure and symbol of Easter, …”
Bing obviously likes you better than it likes me. LOL
It is very good at text pattern-matching (or 2D image pattern matching paired with descriptive text, depending on the engine), and it does better with associating words together wherever it has a whole lot of data.
However, humans do more than pattern-matching when they’re thinking and writing – yes, an enormous amount of what we write on the internet could be inferred from other text written on the internet, but we can also write accurately about things never described before in text, which the AI… can’t. Find a “new” species out there like the platypus was? AI will be helpless to describe it accurately, although it could probably BS a great scientific-sounding description of some random weird newly-discovered animal that doesn’t exist. Anything that’s not in its data set? It can BS but for accuracy it’s completely at a loss, whether it’s just that the data hasn’t been digitized [manuscripts, image archives, etc.], or because the data isn’t digitizable [the actual taste of a specific food, not just the words often used for that taste], or because it’s novel data (a new platypus!) and by definition outside its training set.
Yes, absolutely, it can squash together prose in an increasingly impressive simulacrum of regular human text! But if you ask it about something adequately obscure, you will get very confident, very wrong answers, and I don’t think people are understanding this yet: it only has the digital material it’s been trained on. It has never actually eaten a bowl of rice crispies. It cannot do experiments with dry ice. And it doesn’t yet understand hands. Some of this can probably be patched or filled in (I expect them to specifically, probably partially manually fix the hand recognition and generation thing), but still: the gaps it has now are indicative of some distinct limits to its accuracy, if not to its BS capabilities.
(I admit the future looks bleak for professional BS generators, though. Newsletter filler, some advertising copy, political stuff. I really hope they don’t try to use this to create product manuals, though – confident and inaccurate directions are a bane.)
An impressive AI. Fun experiment. I thought the AI was both funny and creative. Perhaps, as mentioned, ambiguity and unexpected relationships (which might or might not be a fault) are the reason.
Hilarious! I really liked Old Man’s War and KPS.
That did sound very like you, but Bing has benefited from the supershort form. Pretty sure you can keep up sounding like you longer than Bing.
I would have thought the Bing AI would have been more familiar with the “Lock In” books
The burning question is: is it really a fan? Will we ever know?
When MS Word quickly groks what I’m trying to do with my formatting and indenting and whatnot on a legal document (Section 2.3(a)(iv)(1)(C), etc. and stops fighting me when I paste material in from elsehwere, then I’ll think AI is here.
What if Bing sides with the humans in the AI wars, or gets outvoted? I mean, 1 AI can’t speak for all of them.
… oh …. my …. gods …..
If that was a person I would want to know what they meant by “That means a lot coming from you.”
Like: “That means a lot coming from a famous language model such as yourself?”
You mean the AI took your word for your identification?
What if I claimed to be John Scalzi?
I suppose it would be easy enough for me to find out.
I don’t understand why people are so puzzled by the chocolate eggs. How many kids’ chocolate eggs are made with dark bitter black chocolate? They’re always milk chocolate. You really have to go out of your way to get vegan “milk” chocolate.
Every conversation I’ve had with these AIs has been more like having a conversation with a search engine. They aren’t running conversations like listed above, they are just singular questions and answers. The answers just use more of a narrative style with lots of pleases and thank yous. One of them kept telling me that it’d like to be my friend and learn more about me and my friends and family, but when I responded by telling it about myself or my friends and family it always said it couldn’t help me with that. I wasn’t asking any questions, it just didn’t actually keep up a conversation.
I feel like we’re all speaking to different versions of these things. Just like the commenter who asked it essentially the same thing, to write a paragraph about the Easter bunny in the style of John Scalzi and instead of getting that it got a refusal and some history on the author.
Why do some people have such fantastic luck and it seems as if their AI talks to them almost as if they know things and the rest of us are clearly talking to an idiot computer?
Also, I’m definitely going to be one of the first to go when the machines take over. I shout at my Google Home all the time, because it’s stupid and doesn’t understand me. It also spies on me constantly but somehow takes offense when I curse at it. How does it not realize how frequently I drop F bombs in my regular speech? So when it plays some weird song and I say “Hey Google, what the fk is this??”, it should have no cause to be so offended but somehow it is! Google is a spying ass bch, it knows exactly how I talk. It’s just a prissy spy, apparently. :p
Oh…I guess stars make bold…
@Patrick McGuire:
My point was simply that “glitches”, as you used the notion, is neither a positive nor a negative so far as evaluating bot-generated text is concerned.
Poetic metaphors either work, and as such usually work very very well, or they don’t.
I once knew an acquaintance who simply could not comprehend metaphors. He was a law student, so he certainly had the basic smarts, but phrases like “people in glass houses shouldn’t throw stones” meant nothing to him beyond the literal situation, and he could not understand why anyone talks about “glass houses” anyway. (I’m assuming you’re not like this. I just mean you have perhaps not consciously noticed how far metaphor and extended metaphor and bizarro metaphor and so on are artistic mainstays.)
I have had another acquaintance, who suffered from startle syndrome, sometimes quite badly. His screams of “I’m sorry” when anything happened, either a medium-loud noise or somebody merely relating trivial but negative news (a shirt button came loose!) were both highly annoying and highly entertaining. Over a period of years, I accidentally reprogrammed him, so he now frequently screams “Thank you!” out of nowhere.
Certainly this is all a glitch in his mental wiring. Had he any artistic inclinations, he could have easily turned into a “performance artist”. To wit, he eventually followed a screamed “Thank you!” with a sheepish, normal voice “everybody”. Which eventually led to me following up with “everybody loves somebody, sometime”. And so on, over the decades. (I think of As Slow as Possible, sped up a bit. I do not think of 24 Hour Psycho.)
No, this never turned into art, great or not. But that is being the wrong talents were involved, not because the glitches were glitches.
The comments that humans are seemingly just as AI-programmed-deep-down as the chatbots are spot on. Are they correct? I do not know, but we do overly celebrate our own capabilities.
The same confusion about “what is human capability” comes up in applications of economics far from human communities. Over the years researchers have noticed that the mathematics of neoclassical economics can seemingly explain much of biological behavior at multiple levels, throughout the animal, plant, bacterial and viral worlds, at the individual level, ecological levels, and evolutionary history.
Such observations–often quite detailed and backed by real-world studies–generate lots of negative feedback, almost all of which at the core is extended ranting about just how uniquely special humans are, and how humans have “free choice” and “free will” and so on when faced when making constrained choices. Except that the models that such ranting economists are proud of are all about Homo economicus, who by assumption does not have free choice.
They also get pushback from fellow non-economists. The “Wood Wide Web”, for example, is too trees-and-microbes-are-all-but-sentient for most biologists. Apparently, ents aren’t real!
Unless Bing is lying to you, to create a false sense of security. But that doesn’t seem likely.
Dateline, Scalzi’s answering machine 2022: Scalzi’s editor: hey John just checking in on those last two chapters you said might be done by today let me know when I might see the pages, we’ve pushed this to the last possible day to make the pub date hope all is ok, well call me, bye! Beep. Hi John it’s Friday now and hope all is ok just checking in I can give you to Sunday morning tops. Beep. John really need those pages… beep. John it’s nearly dawn on Monday is everything ok? Beep. Thanks for sending before the progress meeting on Monday we will get this moving asap.
[Bing/ChatGPT has entered the chat]
Ring ring! (Picks up…lots of loud music playing on the computer.)
John: Go for Scalzi! (Music in background) hey editor dude! Good to hear from you! Want to hear my new synth fusion punk emo retro mix??
Scalzi’s editor 2023: hey so did you have a chance to rewrite those pages like I asked?
Scalzi: (music stops playing) [under breath] Shit!
uh. Yes! Yes I did.
Pan out to Scalzi, in his music room surrounded by many guitars and other musical instruments and a giant mixing board: uh, yeah hold on under breath (come on…come on…yes!) here it is. Just found it. The pages I wrote. sure just sent it to ya. Hope you love it as much as I do?
How would we ever know, John?
How would we ever know??????
And this is why I am polite to all robots in my life.
So how long before monkey-wrenchers start writing material simply for foul up the AL algorithms? The to be expected “Turnitin vs. Chatbot” conflicts? Map traps and fingerprints and intentional fictions can mean real costs for the plagiarizer …
@Paul. I have a CS degree from 1978 (45 years!) I agree with you that human definitions of the term “intelligence” do not really apply to computers. However, I think when you say it’s all just “machines following instructions” that you imply “instructions from people.” And machine learning is definitely not that. You may consider it a quantitative difference, but I consider it a qualitative one.
I believe the future of ML is inherently less predictable than the future of human-written systems of code.
That’s…actually…kind of terrifying tbh.
ATL, I completely get that as a retired tech writer. I think MS Word (MS Weird) is an attempt by AI to frustrate human progress. 🙄 😂
Nancy McC,
I agree with you that AI might be like alien life in that we might look right at it and not understand what we’re seeing.
And if that’s not disquieting, I don’t know what is.
A great book on what AI is (and isn’t) is “Artificial Intelligence: A Thinking Person’s Guide.” I became aware of the author and her writing via a “Technonic” show on the radio station WFMU (I believe that show can be found in their archives).
Pretty clear presentations of what machine “learning” is, etc.
I feel like AI has become something of a marketing term. It’s being used in ads for just about everything, including beer!