No better time than a New Year’s message, so here it is…
I was trying to figure out what to write about… hey, want to hear about the last three years? Should we talk about Covid and Trump and masks and social distancing and Dr. Bonnie?
Probably not. I think everyone is more interested in the future these days… so let’s go in that direction. What does 2023 actually look like?
There was a bit of a societal quantum leap near the end of 2022, and it has everything to do with Artificial Intelligence… so let’s talk about that.
AI research itself seems to operate the same way as society, where it’s stagnant for a while, and suddenly there’s a sort of phase-change quantum leap that takes things to a whole new level. Very suddenly, AI is all around us. It started recently with a few AI-based art creators. Some are very good and some are insanely ridiculous. I’ll let you decide which, with respect to the picture attached to this post. That was AI-generated.
Apparently, my son’s school held an emergency meeting recently with respect to ChatGPT, an AI-based “author” that will write whatever you want, using some very simple prompts. Like:
ChatGPT, please give me a 50-word essay on Shakespeare:
“William Shakespeare was a British playwright and poet who is widely considered to be the greatest writer in the English language and one of the greatest in Western literature. His plays have been translated into every major living language and are performed more often than those of any other playwright. His works include comedies, tragedies, and historical plays, and are known for their complex characters, beautiful language, and insight into human nature.”
That’s 72 words, but jeez – not bad.
ChatGPT, please give me a 50-word essay on Covid:
“Covid-19 is a highly infectious virus that has caused a global pandemic. It has claimed millions of lives and disrupted daily life for billions of people. Governments around the world have implemented various measures to contain the spread of the virus, including lockdowns and mask mandates. Vaccines have been developed and are being distributed, but the pandemic is far from over. It is important for individuals to continue following safety guidelines, such as wearing masks and practicing social distancing, to protect themselves and others.”
That’s 84 words, but, also… not bad at all.
Is this the future? Yes, to a great extent… but the future isn’t quite here yet. One thing I’ve found playing around with this thing, and something many others have discovered, is that it’s very good at broad brushstrokes, but has serious issues with the nitty-gritty. I’ve asked it some complicated questions about things that I know a lot about and that require a lot of “vertical” knowledge. It’s pretty jarring to read something so well-written and so correct that suddenly contains glaring errors. And the issue is that those glaring errors would only be evident to someone who already knows the topic.
What do I see coming in 2023? A ton of misinformation. If we think it’s already bad, these echo chambers of self-fulfilling knowledge, it’s about to get worse because suddenly, anyone and everyone will have the ability to create well-written official-sounding content… which will start with something written by AI, and then be fiddled-around-with to suit the narrative.
For the moment, English teachers have realized that they will catch the cheating students because, inevitably, buried inside the broad brushstrokes will be little flicks of paint that absolutely don’t belong, and ruin the whole thing. But, unfortunately or not, as time goes on, these AI-based algorithms and the knowledge-stores that feed them will only get better.
Many schools no longer teach cursive writing, something a little eyebrow-raising to most Gen-X parents… but also, at some point, they stopped teaching how to use an abacus… and maybe one day, as horrifying as it sounds, the art of writing will be relegated to machines… and what we’ll be learning is how to feed AI the best queries to generate the best content.
Do you remember non-digital research? Heading to the library, looking through the drawers searching for the right Dewy Decimal System cards to point you to the right shelf where you could extract a pile of books and attempt, over a period of hours, to find what you were looking for?
I’m sure some people still do that, but it’s a dying art. Today’s equivalent is learning how to properly Google what you need, know what results are legitimate, and know how to search those results and extract what you want. As someone who knows how to do both things, I assure you… I prefer the latter. Like it or not, and I write this with only a few hours left in 2022, that’s how it works. Tomorrow is always a new day (in this case, also a new year), but that’s how progress, in every sense, works.
ChatGPT, please give me 100-word greeting of good wishes for a happy and healthy 2023:
“Wishing you all the best in the new year! May 2023 bring you good health, happiness, and success in all your endeavors. May you have the energy and strength to chase your dreams and make them a reality. May you find joy in the simple things in life and may all your days be filled with love and laughter. Here’s to a fantastic year ahead!”
Yeah… not bad at all.
Cheers!
Great to hear from you again Horatio… Happy New Year!
Happy new year! Here’s to some excellent memories 💕
Excellent love the chatGPT -never really cared about specifics -Ha Ha
Happy New Year Horatio
Happy New Year! Fingers crossed that we humans figure some things out or the future will be a very scary place indeed but as the saying goes, where there’s life, there’s hope.
Such an eye opening read! Happy New Year Horatio! 🥳🎉🥂
Happy New Year, Hort. Always good to hear what you have to say. As for the picture…. I’m going to say that perhaps you fibbed on your age to the program? 🙂
It’s always a great read from you. Happy New Year to you and your family! Hope to see you next year. All the best.
Happy New Year bro!
Time to train an ML model that can detect chatGPT authored text.
You’re right on the money about how current AI isn’t intelligent, it is just regurgitating what was in its training corpus. It doesn’t *understand* the subjects it writes about. And the danger there is that people who don’t know better will take it at face value. The internet has already greatly devalued the written word, but now AI is coming to suck the remaining value from it. And this is potentially very damaging. The code generating AIs on GitHub and elsewhere is like this — you ask it to write code for you, but how do you know it is doing the right thing? It sure doesn’t, but it appears very confident in its abilities. (The confidence is a human projection, not part of the AI)
It also fools people, even nominal experts in compsci or AI, into believing that AI is far more sophisticated than it really is. To reach true intelligence there are major technology leaps required, for which it’s not even clear that the theoretical foundations have been laid yet. So we may be a long ways off. And this tech we do have has tons of problems, not the least of which is that it *seems* good. That’s really going to hurt when people start trying to apply it to things in the real world which can be life critical (driving cars, controlling weapons, etc).
Happy New Year! #savewestworld https://savewestworld.com/
Happy New Year!!!
GPT authored texts will take away the individuality, ingenuity , style and creativity of any written human text. What to do to preserve all those human qualities?
Brilliant post, Horatio!👍