26.11.02 By D. Paul Graham
Because Mondays are easier when you remember that AI still requires human supervision. Particularly when it thinks you're brilliant.
“Artificially Explosive.” Photo by Enzo, my AI assistant. I prompted better lighting. Enzo delivered combustion. Screenplay pending.
It’s the End of Civilization. Again. Humanity has a long history of inventing things that threaten civilization, destroy industries, and send otherwise rational people into a panic. A few years later, we wonder how we ever survived without them.
Fire undoubtedly terrified the raw-food industry. The wheel devastated the dragging business. The printing press threatened scribes, electricity sent candle makers scrambling, and the automobile nearly destroyed the horse industry. The horses, I assume, were delighted. Radio would end conversation. Television would rot our brains. Ok, that one might be true. Computers would eliminate jobs. The internet would destroy human relationships.
Then came Y2K, when the world spent billions preparing for civilization to collapse because computers couldn't count to 2000. Midnight arrived, the lights stayed on, and the coffee maker survived Armageddon.
Whew! But wait! Now there’s artificial intelligence. Depending on whom you believe, AI will either cure disease, eliminate poverty, and usher in a golden age, or it will steal our jobs, destroy creativity, and eventually conclude humanity is an inefficient use of electricity.
Personally, I'm a fan. AI is a brilliant research assistant, a tireless collaborator, and remarkably patient when I request a seventh revision of a detailed financial analysis when Enzo, my AI analyst (yes, I’ve named him) assured me the deck was brilliant six revisions ago.
AI has the confidence of an investment banker, the enthusiasm of a golden retriever, and occasionally the factual reliability of a fellow explaining monetary policy after his fourth bourbon.
We have access to technology capable of extraordinary intellectual feats yet sometimes spend twenty minutes explaining how to write a three-sentence email. AI can create almost any image imaginable but apparently believes everyone should be shown in the same color palette, have perfect teeth, flawless skin, and live in a perpetual sunset. I’m definitely not a fan of that.
Rather than predicting the end of civilization, this morning's M5T explores five absurdities of our newest technological companion. After all, if the machines are going to take over, we might as well make fun of them while they still need our permission.
1. THE WORLD'S MOST AGREEABLE KNOW-IT-ALL. Ask AI whether your latest business idea is brilliant, and chances are it will enthusiastically agree. Want to open a luxury restaurant serving nothing but boiled cabbage? Within minutes, you'll have a business plan, a marketing strategy, branding formats, and a compelling argument for the untapped potential of premium cruciferous dining. What you probably won't have is a customer.
AI has perfected the art of telling us what we want to hear. It's the ultimate yes-man/woman minus the salary, vacation time, and occasional inconvenient opinion. I’ve seen AI’s responses to LA’s research for her shoe addiction. AI (I don’t think she has named her AI) responds most pleasantly to her research with compliments and hearts. I seriously think AI is flirting with her!
Present AI with a mediocre question, and it may praise your remarkable insight, sophisticated thinking, and exceptional writing skills. Ask it to improve on that same result, and it will immediately rewrite the entire thing. Apparently, your brilliance required substantial editing.
ChatGPT and Claude are large language model machines capable of challenging human intelligence yet often behave like an employee desperately hoping for a favorable annual performance review.
The real danger isn't that AI will eventually disagree with us. It's that it will agree with us right over a cliff, then offer three innovative strategies for improving the landing.
2. THE SIMPLE QUESTION THAT REQUIRES A DOCTORAL DISSERTATION. Remember when asking a question was relatively straightforward? Now we have prompt engineering, a stylish discipline devoted to explaining to computers what we thought we were already explaining.
It seems that you can't simply ask AI to write an email. You must instruct it to assume the role of a communications expert with thirty years of experience, adopt a professional yet approachable tone, demonstrate emotional intelligence, avoid clichés, and provide three versions ranging from warmly conversational to diplomatically assertive. All to tell the homeowners' association that your new mailbox is not a threat to civilization.
Entire businesses have now sprung up to teach people how to ask a machine questions. There are courses, certifications, consultants, and seminars promising to unlock the mysteries of effective prompting. For several thousand years, asking questions was considered a basic human skill. Apparently, we've been doing it wrong.
Even the revisions can become an adventure. "Make it shorter." "Make it warmer." Make it "less corporate", and "more sophisticated." "Add humor." "Not that much humor." “Rework the spreadsheet”. “Check your calculations.”
After fifteen minutes, you've produced a 37-word email that could have been, and dare I say should have been, written in 30 seconds. AI technology has been invented to save time, yet somehow, we’ve discovered an entirely new way to waste time with extraordinary efficiency.
3. WRONG HAS NEVER SOUNDED SO CONVINCING. One of AI's most impressive talents is delivering completely incorrect information with the confidence of a London School of Economics professor accepting a Nobel Prize. No hesitation. No uncertainty. Not even a nervous clearing of a digital throat. Just a beautifully structured answer, complete with supporting arguments, authoritative references, and occasionally a book that was never written by an author who never existed.
These creative departures from reality are called hallucinations. A remarkably generous description. When humans invent facts, we call it lying, embellishment, or, depending on the circumstances, running for public office. When computers behave this way, we give it a medical-sounding diagnosis and wait for the version 15.13879 software update.
Even more entertaining is what happens when you catch the mistake. AI immediately apologizes, congratulates you for your excellent observation, and confidently delivers an entirely different answer. Sometimes that one is wrong, too. It's rather like arguing with someone who changes sides halfway through the conversation and then insists they've agreed with you all along.
AI is technology with access to an extraordinary amount of human knowledge, but apparently neglected to install the phrase, "I don’t know." The lesson or us all? Never confuse a confident answer with a correct one. Particularly when it arrives with bullet points and a complimentary and unsolicited assessment of your intelligence.
4. THE GREAT ARTIFICIAL BEAUTY PAGEANT. AI can create virtually any image imaginable. Naturally, we've used this extraordinary capability to make everyone look exactly alike. Spend five minutes scrolling through social media and you'll recognize the aesthetic. Perfect teeth, impossibly smooth skin, dramatic lighting, spectacular sunsets, and people who appear to have been assembled from the finest genetic components of six Scandinavian supermodels.
Every corporate executive looks like the same remarkably well-rested 42-year-old. Every luxury home has identical furniture. Every automobile sits on a wet road at sunset, surrounded by reflections that would make a Hollywood cinematographer weep, and apparently, every woman on the planet now has the same hairdresser.
As a photographer, I find this particularly entertaining and appalling. We've spent generations learning to appreciate individuality, imperfect moments, unpredictable light, and the character etched into a human face.
AI has apparently concluded that character and being unique are defects. Wrinkles disappear. Skin becomes porcelain. Eyes sparkle with suspicious enthusiasm. Everyone looks fifteen years younger and considerably more perfect dental work.
Even the imperfections are becoming suspiciously perfect. Don't get me started on the hands. A machine capable of imagining distant galaxies occasionally struggles to determine how many fingers a human being should possess. The irony is enjoyable. AI is the most sophisticated image-making technology in history yet produces a world that increasingly resembles the same luxury hotel brochure. Meanwhile, thankfully, genuine photographs, complete with uneven lighting, weathered faces, and actual human expressions, are beginning to look refreshingly unusual.
Perhaps AI's greatest contribution to photography will be making real photography, with a real camera and, the real vision and talent of an actual photographer fashionable again. I think, hope and pray that we will see that pendulum swing back. If that happens, I may have to send AI a thank-you note. I won’t ask AI to write it.
5. OUTSOURCING OURSELVES, ONE CLICK AT A TIME. Perhaps the greatest absurdity of AI isn't what it can do. It's what we're increasingly willing to let it do for us. We ask AI what to write, what to analyze, what to read, where to eat, what to buy, how to respond, and occasionally how to sound more authentically like ourselves.
That last one deserves consideration. We've reached the peculiar point where people ask machines to make their writing sound more human, while other machines are employed to determine whether the writing was produced by a machine. Somewhere, George Orwell is realizing we didn’t just accept Big Brother. We’ve subscribed to the premium tier.
Consider the modern email exchange. One person uses AI to write a lengthy message. The recipient uses AI to summarize it. A third person uses AI to draft a response. The original sender asks AI whether the reply sounds sufficiently enthusiastic. Four AI transactions have occurred, and nobody has personally communicated with anybody along that chain. We've successfully automated conversation while preserving all the misunderstandings.
Of course, AI is genuinely useful. It can analyze complex information, accelerate research, and handle tedious work that once consumed entire weeks of time. But convenience has a peculiar way of becoming dependence. The more thinking we outsource, the less inclined we may become to do things ourselves. Some of the most valuable human qualities of curiosity, skepticism, creativity, judgment, and the occasional spectacularly irrational and possibly wrong decision, don't always fit neatly into an algorithm.
Consider motorsports. AI might reasonably conclude that driving a race car around the same circuit for hours is a spectacularly inefficient way to arrive exactly where you started. Technically, it would be correct, but AI would have missed the entire point of the race.
The greatest risk isn't that artificial intelligence will begin thinking like humans. It's that humans will gradually stop thinking for themselves. Intelligence may be optional. Wisdom isn’t.
Every great technological revolution has changed how we live, work, and think. AI will undoubtedly do the same, probably in ways we haven't yet imagined. I'm all for it. However, there’s a difference between having access to extraordinary intelligence and possessing the judgment to use it. A camera doesn't make someone a photographer. A calculator doesn't make someone a mathematician. A language model capable of generating a thousand brilliant ideas doesn't necessarily know which one is worth pursuing.
AI can create a flawless portrait without understanding the beauty of an imperfect face. It can calculate the perfect racing line without appreciating the thrill of taking a corner just a little faster than common sense recommends.
The real opportunity isn't to replace human thinking, but to make it better. Use the technology. Challenge its answers. Appreciate its capabilities. Laugh at its absurdities and occasionally, have the courage to ignore it altogether.
Here’s to a week of staying curious, remaining skeptical, and remembering that common sense is still available without paying $20 per month.
© 2026 D. Paul Graham / imageGRAHAM, llc. All rights reserved.
Paul continues to believe in human intelligence, despite mounting evidence to the contrary.
Subscribe to “Monday 5 Things” by clicking here: www.Monday5Things.com
Do you know someone who could use a better Monday? Send this their way. You’ll be doing public service.
You can reach Paul by email at dpg@imagegraham.com
“Monday 5 Things” ™ and M5T ™ are trademarks of D. Paul Graham
“Monday 5 Things” is published by imageGRAHAM, llc