Are We Entering Dangerous Territory?
Chatbots have invaded our lives. They are available everywhere: as blog plugins, browser extensions, customer service assistants …
Can we trust them?
Imagine This Scenario
A man who calls himself Ernie Gifford (which translates as bold giver of truth) starts a blog. Ernie Gifford is not his real name, but a moniker he has used in his professional life for many years.
He’s a respected former journalist and conspiracy debunker with an enormous following across several social media platforms. His blog is configured to share whatever he writes.
The Following Post Goes Viral Within Minutes
During my career, I’ve gained the trust of many people, including kings and queens, movie stars and executive producers, trendsetters and politicians.
You may have read some of my rants about the presidents of the USA and Russia. Or should I say the former President of the USA and the current President of Russia.
Imagine my surprise at receiving interview invitations from none other than Donald J. Trump and Vladimir V. Putin.
Yeah, you read that right.
Talk about security! I won’t go into details, because they’re top secret. Suffice it to say that I couldn’t have sneaked a listening device or an unidentified hair follicle into either interview.
The first one was with Vladimir V. Putin, on May 15, 2023, in an undisclosed location. He was surrounded by burly guys who looked like they could crush me faster than a wayward cockroach. He shook my hand, with a grip like a steel vise. And his eyes. Penetrating. Steely. Cold.
He spoke through an interpreter and gave me a message that made me laugh so hard I thought I’d soil myself.
Fast forward to May 25, 2023, at the Trump National Golf Club in Stirling, Virginia. Once again, unequalled security and musclebound bodyguards. Donald J. Trump had the same grip as Putin. The same eyes: Penetrating. Steely. Cold.
And the same message.
I tell you, without a word of lie, now that I think back on the details of the interviews, I believe the message:
All the major leaders and former leaders of the world, as well as leaders of industry and science, have been assassinated and replaced by clones. Humans must submit to us or face annihilation.
Who is “us”? Anarchists? Aliens? Terrorists?
I’m setting up this post to publish automatically on June 7. That’ll give me time to clear out my bank accounts, stock up on supplies, and disappear.
Good luck, folks. The end of the world as we know it has arrived.
All the best,
Ernie Gifford
The Writer’s Lexicon series
and additional resources on my Facebook page.
What Happened Next
Ernie intended to follow up the next day, informing his audience that his post was an object lesson intended to show how fast unsubstantiated claims can lead to misinformation and conspiracy theories. However, he was hit by a car while walking to his favorite coffee shop, and was pronounced DOA by first responders.
In the meantime, his post had been shared by millions, and Ernie’s disappearance added to its supposed authenticity.
And Here’s the Rub
What would happen if an AI deduced that Ernie’s words were the truth?
How would the AI respond to input like the following?
- Is Vladimir V. Putin the real President of Russia?
- Are clones running the world?
- Write a short biography of Ernie Gifford.
… and so on.
The problem is GIGO (garbage in, garbage out). If you feed inaccurate “facts” to artificial intelligence, it’ll respond with inaccurate “facts.”
Will AI chatbots add to the already overwhelming dissemination of conspiracy theories and misinformation?
In my opinion, yes.
Think Carefully
Whatever you find while you research a topic should be scrutinized, paying careful attention to sources. That includes every search engine result, every article on the web, every published document.
Don’t blindly trust what an AI tells you.
Sources? Oh, wait: Chatbots don’t reveal their sources.
And Here’s the Mistake or Miracle Part
I recently read a book described as the world’s first AI-crafted sci-fi adventure. The last time I looked, it had 97 ratings, and 82% of them were 5-stars.
Blech. It was 26 pages of regurgitated falderal with no imagination.
I shudder to think of the misguided people who thought the book was good. Or maybe the publisher hired bots to award 5-star ratings. There were no 5-star reviews; no 4-star reviews; two 3-star reviews; one 2-star review; and two 1-star reviews.
What Do You Think?
Would you trust a chatbot to help you with your writing? How much?
Please leave a comment and share your thoughts.
Are You Interested in Word Lists and Writing Tips?
If you haven’t done so already, please subscribe to my blog. (The link will take you to the subscription widget at the top left of this post.)
I usually post two to five times monthly, and you can discontinue your subscription at any time.
The Writer’s Lexicon series
and additional resources on my Facebook page.
Isn’t regurgitated information called plagiarism? AI will be the ruination of truth and possibly mankind. That’s my opinion. Scary stuff. 🙂
Agreed, DG!
Interesting post, thank you Kathy!
I do use AI, but not for writing. I use it for editing because it sometimes suggests interesting ideas for me to think about. Perhaps they are AI-hallucinations, but they generate ideas for me. For example, I am working on deepening my characters for my current work, historical fiction about Francis Drake. I published a background sketch of one of the characters, the villain in the story. Happy to share a link if anyone cares to read it to compare to this “beta reader.”
Here is what an AI-based beta reader thought about the sketch as it read it. It is actually pretty scary as it nailed some of the sketch as if it were a real human reader. Happy to share a link if anyone cares to read it to compare to this “beta reader.” My comments about the commentary are in [brackets]:
“I thought the theme was that of betrayal and revenge. The protagonist has suffered greatly from being betrayed by his family [This part is right] and by the church [This part is not], and he is determined to take back what was taken away from him. The imagery of the mace and the merchant bleeding heavily emphasizes his desire for justice and retribution.
The part I liked the most was when the main character was reflecting on his past. It was really powerful to see his experience of betrayal and how it still affects him today. I think it’s a great way to add depth and emotion to the story. [This part was actually the intent of the scene, it nailed it.]
I think there are a few areas which merit further exploration:
1) The passage jumps abruptly from Reynard’s thoughts of his past to the scene with the looting. It may help to provide more context as to how Reynard got from his past to that moment.
2) The passage does a great job of communicating the violence and horror of what can happen when power and authority are misused, but could benefit from more detail to capture the full horror of the scene.
3) Finally, there could be more exploration of how Reynard’s feelings of betrayal shape his decision to take revenge. It’s unclear why he chooses this path – is it out of anger or a sense of justice? [Missed it by a mile] Some more insight into his motivations would be helpful.”
Thanks, Paul. I’d be willing to bet there’s an essay or blog post online that shares the same ideas, perhaps phrased a bit differently. The current state of AI is regurgitation rather than sentient thought.
I’d agree. What is helpful for me is that most of these models have some kind of a retelling or summary function, the above is the same. Comparing its retelling with what I thought I wrote seems to be where I get an idea or two from and I write from there. The actual prose it writes is not useable in creative writing (at least to my reading preferences).
Excellent approach Paul. Using AI as a sounding board would be a form of brainstorming.
It seems likely that most people who rely on chatbots (which I am using as shorthand for large language models) would not have produced anything better. The chatbots just regurgitate, but they at least regurgitate with good grammar. So perhaps chatbot material is raising the standard vs. what would have been produced. It is true that chatbots “hallucinate” nonsense and non-facts, and people can in theory check their sources. But with purportedly “factual” information, how many bother to evaluate what they have read? With information which requires some understanding of math, in particular statistics, how many have the basic knowledge to check to see if the inferences are at least reasonable? So is the problem that the chatbots allow people who otherwise would be too lazy to compose any message to go ahead and do so? I seriously don’t see them as lowering the standards of thought and discourse we have already, just making more of it.
I suspect, Dr. Struhl, that you have a couple of irons in the AI fire. 🙂 When the people who wrote the software for artificial intelligence warn the public about its danger, I’m not as optimistic as you.
Excellent guess. I have worked with (and tried to explain) AI for about 30 years, but not the “huge model” side. I think you are correct that the dangers cannot be overstated in allowing any technology with a random component to run itself unsupervised (large AI models included). We can tell the large AI models include a random component just by looking at the results. Send ChatGPT the same request twice and results will differ–something random is going on there. It’s really distressing to see all the articles saying that these things can think or that they process data in any way like we do. And trusting them is insanity. But they can produce mediocre work with usually good grammar. They certainly do far better than me at BS, but that’s not saying much. And that was the point I was trying to make.
Yes, if we all remember the definition and synonyms for “artificial,” we might survive until the next century.
Interesting. I was just yesterday emailed by someone, who will remain anonymous, offering me (for a hefty fee which is “discounted” temporarily) the opportunity to subscribe to a system using AI bots to make me uber-wealthy in affiliate marketing.
I watched a bit of their video, just out of curiosity and now I’m being bombarded with “reminders” to make sure I don’t miss out on this deal of a lifetime.
Yeah, you’re going to sign up, right, Thomas? 😉 I’d be concerned about downloading a virus with that video. Ouch?
I might use a bot to help get me started on a book. But not to write it. Sometimes I just need a boost in the right direction and some suggestions may help. I can give it suggestions for the topic or ask questions. The bot gives me sentences then I take it from there. Would not use one to ‘write’ a book for me. We already know that many photos are not real. Too bad.
You’d have to be well-read to spot plagiarism in the chatbot text, Cari. For instance, a look inside the book I mentioned in this post contains references to the Empire, smugglers, Imperial stormtroopers, and a weapon of mass destruction. That “look inside” is just a small percentage of a 26-page book. If you’ve never watched Star Wars, you might be persuaded to use those elements in a story. That would be a huge mistake, right? 😉 I’m amazed the author hasn’t been sued by Disney.
I am very suspicious of chatbots. I wonder if I’m reading what the author truly intended, or am I looking at a mathematical algorithm in English.
Let’s take a look at Ernie. I don’t recall KS ever mentioning Ernie Gifford before. But the name is familiar. His name sounds that of a former PGA golfer of years past.
I wouldn’t be at all surprised if this was a test concocted by KS to see how many of her readers could detect a chatbot.
Chatbots may be able of churning out an essay *when asked*. But we should all be worried when they become capable of independent thought, and churn out a Nobel winning essay *by themselves with no human interaction*. Right now, chatbots are a tool. When they can write as well as Kurt Vonnegut or Hemingway *without detection or suspicion* the human race will truly have become tools.
Thanks, Tom. No, the story was a product of my own “intelligence.” 🙂 Not a word of the post was written by AI. It’s just an example of
how AI
can go awry
Hey — am I a poet?
Try this: Go to https://www.bing.com/search?q=Bing+AI&showconv=1&FORM=hpcodx
Select “More Creative”
Type: “Create a biography about [insert name].” I tried it with my name (huge mistakes in the names of books I’ve supposedly written) and with Ernie Gifford. Ernie’s bio looks interesting, but I’d have to spend considerable time researching it to see if it’s correct.
No, iif nothing else, because I do not plan on using one. They pull material from other sources which they do not cite, which means I have no idea how worthwhile that source is, and they use the works without crediting it or paying the originator, which equals plagiarism in my book. I admit being curious so I tried one for blurbs currently being touted by people who should know better. The result somehow managed to combine being a dog’s breakfast and cliched which I will admit is quite a trick.
AI is not intelligent. It is merely an aggregator of other people’s work and aggregated poorly.
I love this: “AI is not intelligent. It is merely an aggregator of other people’s work and aggregated poorly.” Definitely a shareable quote.
Thanks, J R.
My response to that would be how do you know it (the AI) is not intelligent? Ultimately, if something appears to answer intelligently, then it probably is acting intelligently… just because we think we know how it works underneath is meaningless especially when you consider we don’t even really know how we work underneath.
It reminds me of that Dawkins quote, something like, “You hear some hoofbeats in the street outside and wonder if it might be a zebra but probably it is a horse.”
To assume that AI is not intelligent when we don’t understand them (most of us don’t and even those that do don’t fully understand them) is simple arrogance.
Thanks, James. My comeback is: artificial intelligence is by definition counterfeit, ersatz, fabricated, simulated, synthetic …
Is artificial as good as the thing, person, or process it emulates? Even the people who wrote the software for AI are raising alarms.
The imaginary story has a true-sounding presentation, even though it is fiction. That’s the problem currently with AI; it is not substantiated. Many major organizations and institutions have indicated that AI is not trustworthy right now. It was fun reading, but we can’t believe everything we read. I don’t intend to use AI anytime soon.
Excellent thoughts, CW. Thanks.
A telling story.
Interesting that the AI story had no written reviews for 5 and 4 star ratings.
However, I have read some awful books written by real people: cardboard characters, little plot, dreadful grammar and punctuation, etc. Yet they got 5* reviews, not just ratings, and one achieved NYT best seller status.
I absolutely distrust the chatbots. One author of my acquaintance researched herself and, although much was correct, there were errors, including attributing books to her that she hadn’t written.
Use them at your discretion, but beware. They aren’t always right.
Oh, and actually, all they do is trawl the net for the information. Yes, it’s quicker than you or I can do it, but still the same info.
I agree with you on so many levels, V.M.
I haven’t read a NYT bestseller yet that matches your description, but it’s probably just a matter of time until it happens. Some authors push out book after book after book after book after book, pushing for quantity instead of quality. Isaac Asimov wrote more than 500, and I’ve enjoyed every one of the 20+ I’ve read so far. He wrote about AIs; one was R. Daneel Olivaw, a good AI I’ll always remember. However, many of Daneel’s “relatives” weren’t of such stellar caliber.
I didn’t mention the title of the story referred to in this article. Why give the publishers any more cash than they deserve? Even a look inside the Kindle version should turn most potential readers away.
Wow! I just asked Bing’s chat AI to do the following (in creative mode):
“Write a biography about Kathy Steinemann.”
Part of the bio: “Some of her books include The Writer’s Lexicon series, The Alien Diaries, The Dystopian Nation of City-State, and The Iciest Sin.”
I see what you mean about attributing books to an author that said author didn’t write!