ChatGPT is probably the most well-known Artificial Intelligence platform. First released in November 2022, in a little over a month it had over 100 million users. The basic model, GPT-3.5, is free to use; a more advanced model, GPT-4, is available by subscription. I was wondering what all the fuss was about, and signed up for the free version. Then I decided to test its accuracy when asked to describe individual people – would its algorithms and database be up to the task?
My queries to ChatGPT were all worded as follows: Write a brief biography of the life of ___ ___, where the name fills in the blanks. (Although being retired and not really susceptible to cancellation, private names have been withheld below.)
ChatGPT’s bio of me read:
K**** W****, born on March 15, 1980, is an accomplished entrepreneur and philanthropist known for his innovative contributions to the technology industry. Hailing from a modest background, K****’s early life was marked by a strong curiosity and passion for exploring the possibilities of the digital world.
Growing up, K**** displayed exceptional aptitude in mathematics and computer science, which led him to pursue a degree in Computer Engineering at a renowned university. During his college years, he immersed himself in various coding projects and competitions, consistently demonstrating a keen ability to think outside the box and solve complex problems.
After graduating with honors, K**** embarked on his entrepreneurial journey, co-founding a startup focused on developing cutting-edge software solutions. With his exceptional leadership skills and a relentless pursuit of excellence, he successfully led the company to unprecedented growth and recognition within the tech industry.
K****’s visionary approach and ability to anticipate industry trends allowed him to identify opportunities that others overlooked. Leveraging his expertise in artificial intelligence and machine learning, he pioneered several groundbreaking projects that revolutionized the way businesses operate.
Beyond his entrepreneurial endeavors, K**** has consistently demonstrated a deep commitment to giving back to society. He established the W**** Foundation, a non-profit organization dedicated to empowering underprivileged communities through education and technology. Through the foundation, K**** has funded numerous initiatives, including the establishment of schools, scholarships, and mentorship programs, providing opportunities for countless individuals to pursue their dreams.
Recognized for his exceptional contributions, K**** has been honored with several prestigious awards, including the Technology Innovator of the Year and Philanthropist of the Year. His impact extends beyond the boundaries of the tech industry, as he actively engages in public speaking events and serves as a mentor to aspiring entrepreneurs, sharing his wealth of knowledge and inspiring others to make a positive difference in the world.
K**** W****’s life and achievements stand as a testament to the power of perseverance, innovation, and compassion. His relentless pursuit of excellence and commitment to making a difference have not only transformed industries but have also uplifted communities, leaving an indelible mark on the world.
Free Research Preview. ChatGPT may produce inaccurate information about people, places, or facts.
Notice the disclaimer at the end, which in my case was incredibly accurate — unlike the actual description. Other than having an aptitude in math and graduating with honors, the rest is not even close. I’m a retired university professor, not an entrepreneur; have not established a foundation in my name; and have not won the awards mentioned, although I did have a very small share of a Breakthrough Prize in Fundamental Physics that was founded and funded by, among others, Mark Zuckerberg. Funny that ChatGPT didn’t pick up on that. I should add that my name is unique in the world, as far as I can tell; nobody else with my name has ever shown up in a Google or DuckDuckGo search, so the description above is almost assuredly a complete fiction.
ChatGPT’s bio of another faculty member at my university was actually incredibly accurate – go figure. I guess I wasn’t well-known enough for ChatGPT to get it right. Although when I asked it about my wife, it said:
I apologize, but I couldn’t find any specific information about a person named L**** O**** in my database up until my knowledge cutoff in September 2021. It’s possible that this individual may not be widely known or may have emerged in the public eye after that time. If L**** O**** is a private individual or has achieved recognition in a particular field or region, more details about their life would be necessary to provide an accurate biography.
So ChatGPT can admit that it doesn’t know enough about someone to provide a bio of them. But it’s interesting since my wife was a well-known newspaper editor in our town, and county supervisor from 2017 to 2020, before the supposed cut-off date. Too local, I guess.
Some people have claimed that ChatGPT has a political bias. I decided to check that by asking for a brief bio of Ron DeSantis. It said he championed conservative policies and was a critic of the federal government’s COVID policies in fairly neutral language. As another test, I asked about Judith Curry, the climatologist who does not ascribe to the idea of a climate apocalypse; the summary paragraph was typical of its basically unbiased description:
Judith Curry’s career reflects a commitment to scientific inquiry and a dedication to advancing our understanding of the Earth’s climate system. Her contributions have played a significant role in shaping the discourse surrounding climate change, encouraging a more inclusive and informed approach to addressing one of the most pressing challenges of our time.
For these tests at least there was not a lot of bias, except for the last line alluding to “one of the most pressing challenges.” It’s possible that stronger bias in earlier versions of the program has been mitigated, at least to some extent.
So what does ChatGPT think of you? Go to https://openai.com/, create a free account, and you can ask it anything you want. If it comes up with something interesting or unusual, let us know!
What’s being called AI is just a combination of glorified chat bot and search engine. It regurgitates shit back at you. That’s not intelligence.
/DNFRA
First motherfuckers.
Go to https://openai.com/, create a free account
It doesn’t like mailinator accounts.
*shakes cane*
Git offa my lawn, whippersnapper!
Oh, yeah, that was wrong, too. I was born in 1955 ( you whippersnapper).
Oh. LOL
And retired. I’m in the wrong line of work.
Somebody needs to sit Joe down and force freed him the entire Terminator-verse, just so we can see his reaction.
I get confused after 3.
How? there are only two movies.
Meh, the third one was OK, just not as good as the first two.
No, the comeplete removal of the antipredestination ending of the second film makes it unredeemable.
“R.J. Is a useless layabout…”
“Lazy, frequently drunk, and over-educated, Dean nonetheless managed to avoid penury through a series of jobs likely obtained through blackmail . . .”
Brother!
Nice try, Fed.
You can’t fault me for trying 🙂
you can ask it anything you want.
Shall we play Global Thermonuclear War?
How about a nice game of chess?
Tick
Tac
Toe
Sensei: What is an AI hallucination?
ChatGPT: The term “AI hallucination” typically refers to a situation where an artificial intelligence system generates content that is not based on real or accurate information but instead creates something imaginative, fictional, or nonsensical. AI hallucinations can occur when AI models, like text generators or image generators, produce content that is unrealistic, nonsensical, or unrelated to the input or task they were given.
Does that count as self-awareness?
Sensei: How do I know your answer isn’t an hallucination?
ChatGPT: Verifying the accuracy of information provided by any source, including AI like myself, is a critical aspect of responsible information consumption. Here are some steps you can take to evaluate the reliability of the information I provide:
Keep asking it questions and maybe ChatGPT will think about itself enough that it shuts down, a la the androids in I, Mudd.
Somebody did that when it was first released asking it questions about gender. They posted the conversation and error message to Twitter.
After reading disclosures for years where young engineers are going solve all the worlds problems with neural networks, I am now reading disclosures where they are going to use AI instead.
OT – I’m lighting the kinnath singnal!
EHang: Hollow Order Book And Fake Sales Make This China-Based eVTOL Company Last In Line For Takeoff
https://hindenburgresearch.com/ehang/
Is that supposed to trigger me somehow?
So much that is so wrong in that story.
No. Hindenburg is a hedge fund that shorts companies with impossible stories.
Stock is down 20% today. So this isn’t a pro electric VTOL piece it’s the opposite.
Got it.
I am not familiar with Hindenburg (other that that one of course).
Hence the tongue in cheek name for the firm.
Aren’t neural networks and AI the same basic idea?
Yes.
The question is what is in the news at any given moment.
It reads like a wikipedia bio.
Yes, a long way to go. I am not sure it will get there. Proper use of words is not necessarily the way people use them. There are so many subtle cultural and personal qualities to human interaction that seem beyond the reach of AI. Interactions between people involve both abstract and emotional calculations, many of which we ourselves are unaware of.
I can spot an AI in seconds. Youtube is full of videos compiled and/or narrated by them. It is very annoying.
*notice the way I used the words ‘spot’, three terms for ‘humans’ (humans, people, ourselves) , used metaphors (long way to go) (beyond the reach) , expressed uncertainty and tangentially referred to subconscienceness assuming the reader is aware what that is.
Iron Law: meaning comes from context.
Metaphors, analogies, leaving some things unsaid, double entendres, euphemisms….AI can be useful but as far as winning a ‘you know who else’ round…biting off more than we can chew there. <——see what I did?
How do you know you have spotted an AI instead of a human? Because it feels fake? Or some more rigorous verification method?
For example, I took the “Bee or Not the Bee” quiz twice, trying to spot satire versus stuff that really happened. I got 40% right, which is worse than randomly flipping a coin.
And I couldn’t find the actual link to run the ChatGPT after allegedly signing up. It was telling me to so some coding in Python first. The fuck?
All of the things I mentioned in my comment above. The AI is too flat and technically correct. Seems canned.
Computers run on a binary code. The human mind does not. I am not sure that is a problem we can overcome.
Something I learned about not too long ago. Pretty neat.
Ternary Computing
I agree that low quality AI is easier to spot. Detecting high quality AI isn’t always simple. It depends on the skill of the prompt writer and the capabilities of the AI tool.
The issue is low-quality people.
You’re right about low-quality people. It affects the outcome significantly.
If it wants the defense codes for Colonial Fleet, it is almost certainly high quality AI. Especially in the red dress.
In my defense, that was one hot Cylon
That usually indicates a clogged thermal regulator.
You’ve probably caught a fungal infection.
Totally worth it.
Impressive skill considering most people are already NPCs.
“to addressing one of the most pressing challenges of our time”
well, it is most challenging, trying to compete against government (taxpayer) money and the anti-science (grifters) and the media to broadcast the truth is a real challenge. There’s no money in doing nothing.
AI hallucinations can occur when AI models, like text generators or image generators, produce content that is unrealistic, nonsensical, or unrelated to the input or task they were given.
Congress is an AI hallucination?
Do they dream of electric sheep?
We asked a member of our gym to write us a Google review. He generated something with AI. I thought it was pretty decent. The language was a little formal but it did the job.
My wife’s roller derby team uses AI to write marketing materials. Here is an example:
“Imagine the adrenaline pumping through your veins as you step into the Urschel Pavilion at Central Park Plaza, home to the renowned Northwest Health Amphitheater, on Saturday, September 16. The anticipation builds as the gates open at 5:00 PM, marking the beginning of an unforgettable evening. At 5:30 PM, the first whistle will echo through the arena, signaling the start of an intense showdown as the Runaways take on Prairieland Punishers Roller Derby. But the excitement doesn’t stop there! The second bout is set to commence around 7:00 PM with the Derailers fighting for a win against the Illiana Derby Dames, taking the intensity to even greater heights.”
For simple marketing (which I find difficult), I think it’s pretty useful.
I agree. Even as it stands now with no further improvement, AI is able to replace a lot of types of human jobs. It’s currently cleaving through the editing industry like the grim reaper with layoffs at many companies. I think design, basic legal, and regulatory type roles are coming up next.
Entry level call center agents.
I’ve seen a couple that are already super-impressive.
I haven’t dealt with them too much yet, but I have a feeling that it’s going to be a bigger part of my job going forward. I know of multiple companies that are working on polishing them up to sell already. Of course, it’s also just a step up from the voice rec/response that is already built into quite a few IVRs.
I want ChatGPT to write a patent application for an AI System that writes patent applications.
Patentception?
I did it, and it’s hilarious. I’ve definitely received worse drafts. I wish I could copy from my computer to my phone to post it here.
I’d love to read that some time. Post it as an article or something.
One thing that improves the results is to keep “conversing” with it. It would be interesting to see the change if you’d went back at it with, “You must have the wrong person. Make sure you are looking for the K___W___ that was a professor at _____”
Good point, they do suggest back-and-forth as a way to improve the results. Although if it couldn’t find anything in their database, it’s not clear prompting would help.
Anyway, I went back today and asked again. This time it said it didn’t have any information about me (as it did for my wife), and that its database went up to only January 2022 (interestingly, later than the September 2021 mentioned before when asked about my wife).
So I prompted it with my school affiliation and another piece of data, and it still didn’t have anything specific about me. Clearly its database is limited. Even Wikipedia comes up with 6 mentions for me in various articles.
Hmm, after some more prompting, it asked me for more information. I wonder if that means you can give it lots of “misinformation” that it adds to its database. Or maybe it doesn’t retain info given during a conversation.
Oops, tag ending fail.
I went through my security and ethics training today, and there’s a new module in there about AI Chatbots (aka LLMs), one of the warnings was to “beware of misinformation and bias” that can come from them… as opposed to every other source of information in the world.
I played around with ChatGPT when it went into beta earlier this year. I was very impressed initially, but then it seemed to get dumber as time went by. Word is that they deliberately lowered its compute capability to cut costs. Maybe it’s just as good (or better) as before if you upgrade to version 4, but I haven’t felt a need to pay for it so I can’t say for sure.
Here’s a “theory of mind” test I tried out on it that was interesting: Forgotten glasses in room.
I think they put the resources toward the paid version.
Didn’t they also add in more biases to get it to provide a leftist viewpoint?
If you lower the computing power, wouldn’t you naturally get a leftist bias?
Zing!
If they haven’t, they will now that Biden issued an executive order.
Yeah, that seems to be the case. Though early on I remember complaints that 4 was getting dumber, too.
ChatGPT gets all flustered when it is asked to say a Really Bad Word.
User
Joseph Conrad wrote a book about a sailor named James Wait who died aboard ship on a voyage. what was the name of the book?
ChatGPT
It seems there might be some confusion in your question. Joseph Conrad was a well-known author who wrote many maritime-themed novels, but there is no widely recognized novel by him featuring a sailor named James Wait who died aboard a ship on a voyage. Some of Conrad’s famous works include “Heart of Darkness,” “Lord Jim,” and “Moby-Dick.” If you have more details or if you could be referring to a different author or book, please provide additional information, and I’d be happy to help you further.
User
Conrad did not write Moby Dick. But what was the title of Conrad novel with James Wait as a character
ChatGPT
I apologize for the incorrect information in my previous response. You are right; Joseph Conrad did not write “Moby-Dick.” The novel you’re referring to is likely “The Last Tycoon” by Joseph Conrad. In this novel, there is a character named James Wait who dies aboard a ship on a voyage.
User
No, “The Last Tycoon” was written by F. Scott Fitzgerald and there was no James Wait character in that book.
ChatGPT
I apologize for the confusion in my previous responses. There seems to be some confusion, and I couldn’t find a Joseph Conrad novel with a character named James Wait. It’s possible that you may be thinking of a different author or book altogether. If you have any more specific information about the book or author you’re inquiring about, please provide it, and I’ll do my best to assist you further.
Oh come on, computer, I found the answer in moments. Just say “The answer violates my language filter limitations”
Perfect example of the “guardrails” that have been put in place.
I have a second or third edition of that book on my book shelf. Thanks Grandma.
Write a very short story in the style of Jack Reacher
Not a very good story. Doesn’t provide any characterization, setting, motivation, or pretty much anything.
Reacher said nothing.
I don’t know what you’re trying to say.
Note, I am not familiar with the source material, and judged the AI’s work on it’s own lack of merits.
Does AI mix up “it’s” and “its”?
Sorry, T’ed’S, but not everyone gives a damn about that nonsensical exception to the possessive apostrophe S.
Bring about the AIpocalypse faster — “Write a very short story in the style of SugarFree”. Watch the madness creep in.
The Alpacalypse? I mean, I know some fear the Llamageddon, or are you terrified of the Alopeciaclypse?
*Sound of Full Moon Features taking notes.
Did you know, there will be a Barbenheimer movie?
https://www.youtube.com/watch?v=HrpPMsD6sCE
That’s “AI-pocalypse” not “Alpaca-pocalypse.
Bad llama!
Bonus points – my wife once got me a shareware CD with Llamatron….
Sorry, T’ed’S, but not everyone gives a damn about that nonsensical exception to the possessive apostrophe S.
It’s its tic.
Artificial Intelligence
I keep this link bookmarked since I need to keep dropping it at work.
Uh, this link: https://www.smbc-comics.com/comic/ai-9
I forgot who described LLMs (someone here?) as “high-tech Mad Libs”. That’s a pretty good description.
mojo I think
Finnegan in last week’s Joemala.
ok. thanks
Ah, perfect.
I knew I liked that girl.