Reality Check on ChatGPT
June 12, 2023
It has been less than six months since the introduction of ChatGPT in November 2022, and during that time, many people, from pranksters to Nobel Prize winners, have shared their opinions about it. I've found many of these reviews to be captivating, but to me, some of them go too far with excessive exaltation.
Breakthrough into a fantastic future! - We all can die! - Why work, give the task to ChatGPT! - You will all be fired!
Let's add a dash of sobriety to this simmering stew and give the ChatGPT service a reality check. Let's test its ability to handle the harsh realities of our lives and see if it can withstand the heat. Setting aside emotions and fantasies, let's share our honest thoughts about this service.
Did ChatGPT live up to your expectations? What did it ultimately give you that was useful? How do you anticipate the widespread application of neural networks will affect your life? Do Internet scares scare you, do you believe in Internet pie in the sky promises?
Well, while you're thinking about it, I'll write something myself. About a month ago, I started using ChatGPT personally. Since then, as a copywriter, I have been consistently using it, especially when producing texts for clients.
"And ChatGPT has become an indispensable assistant in my work," is the phrase that ChatGPT would have produced at this point. But no, there won't be anything like that. I'm going to discuss how my personal experience with ChatGPT correlates with what I have read online about ChatGPT and, first of all, those aspects of other users' reviews that most puzzled me.
Have you been overwhelmed by ChatGPT's capabilities? Has superhuman intelligence been glimpsed in his responses to your queries?
Yes, many users respond. And I generally get what they're saying. Besides, I experienced one of these moments too but with a different neural network that was much more focused on psychology. We ended up chatting until late at night, diving into a philosophical discussion.
And now that it's already late at night, the network spilled out a very insightful remark, which struck me a lot about the motives behind my questions. I was literally appalled. - "A human, dude is there! No way it could be a machine!" Like the old quip that the Google Assistant is really a couple of hundred freelancers.
For a few seconds, I frantically tried to remember what I had let slip. And then an even more terrible thought occurred to me. "Look, this thing instantly gives out long answers. I bet no human is able to do this. You're not a human! Who are you? Who is there?" - Was it a lot of fun? Yeah)
Alas, despite my best efforts, the ChatGPT and I grew tired of discussing abstract topics with it after receiving numerous responses falling under the category of "Thank you, Captain Obvious," along with countless fictional assertions.
And how should all of this be understood? Should I assume that ChatGPT is less intelligent than that psychological neural network and incapable of saying anything smart? Or that the network was some sort of genius endowed with consciousness?
Of course not. They're just machines, after all, and not capable of independent thought. They simply manipulate text based on their training data with varying levels of proficiency. It just so happened that the texts used to train the psychological neural network contained profound insights on a topic that piqued my interest. It was the authors of these texts who were responsible for these brilliant ideas. I wish I could talk to these folk directly! The machine is merely a conduit for repeating and paraphrasing words from people with varying degrees of relevance.
I Don't Need Google - ChatGPT Knows Everything
Did ChatGPT justify your expectations when you attempted to find out something?
There are numerous statements from users saying they replaced Google with ChatGPT for information searches. I tried to act the same way, but in the vast majority of cases, the results did not satisfy me at all.
The question is: why can't ChatGPT do the same as that psychological neural network did and result in a comprehensive response to my question? Yeah, hypothetically, it can do anything. Once more, it's a matter of what texts these neural networks have been trained on. As can be assumed, the psychological neural network was trained on specifically niche, top-quality texts in psychology. Whereas texts used for training ChatGPT were derived from the whole Internet. There are some great psychology texts there too, but they're just a drop in the bucket compared to the rest. And you are aware of what the whole online ocean consists of.
That reminds me of the time when I became aware that searching with Google required skills too. If you can't hit the bullseye with your search terms and snag precisely what you need out of the vast online ocean, you'll end up endlessly chasing your tail and getting lingual husk mixed with kittens and pesky ads.
In order to get something out of ChatGPT, apparently, a similar art should be mastered. What will it get us? It's as if Google gave me links to 10 super-relevant sites, and I wouldn't even have to study them. Instead, I would command "ChatGPT, explore these 10 sources, and answer these questions of mine." Well, hot dang, that sure is mighty tempting! Maybe someday it will become a reality.
So far, all I've been able to get from the ChatGPT are Captain Obvious-style answers. If I need such an answer, for example, about a subject with which I am completely unfamiliar, then it's okay. But when I ask about something that isn't a huge internet forest, maybe just a little grove, ChatGPT either gives me vague answers or starts going off into fantasies. The last is the worst. It actually kills the desire to use it. And you expect me to double-check every single word on Google? No way, man!
When did Hemingway visit Hiroshima?
Ernest Hemingway visited Hiroshima in early 1949, just a few years after the atomic bomb was dropped on the city on August 6, 1945. At the time of his visit, the city was still recovering from the devastation of the bombing. Hemingway was in Japan on a goodwill tour sponsored by the U.S. Department of State, and he visited Hiroshima as part of his itinerary.
Using chatGPT to study? Be cautious. Here, it decided to fulfill the dream of Hemingway, who never got a chance to visit Japan.
ChatGPT Writes Texts as Well as a Human Does
Do readers like articles written by ChatGPT?
There are a huge number of reviews on the web in which users admire the quality of the texts that ChatGPT produces. But there is one catch: all these users are the people who WRITE texts. What do people who READ these texts think about them? I haven't seen any feedback like that, and I am wondering if readers agree with this enthusiasm of the writers.
My job also involves a lot of writing, but now I would like to give emphasis to my interests as a reader. How should I regard the creativity of artificial intelligence?
The first thing that comes to mind is the analogy with chatbots. Imagine visiting a company's website and finding a chatbot instead of a human support representative. "Oh, they replaced their tech support with a bot, fine!"—Is that really your intended reaction? The company can be as enthusiastic as they want about how advanced this chatbot is, but I know that they are using it to solve their problems, not mine. The quality of the service I receive has definitely dropped.
If I have a feeling that ChatGPT wrote a text, I probably won't read it. Does this mean that I consider ChatGPT to be pure evil, and that all wordsmiths should abandon it immediately? Absolutely not!
I'm sure that many of the articles I've enjoyed were likely produced with the help of ChatGPT. Where's the difference? The authors of those articles had something to say to me. I heard their thoughts, and I didn't care about the way they wrote them down, whether with a pencil, chalk, or even their own blood. Were they typed or handwritten? Did they call upon the muses? Did they use a calculator? Anything you like, folks. If it helps you, it's okay with me.
It's a different story when the author commands ChatGPT with the following: "Write a post for my blog that's engaging and informative, on a popular topic, in 500 words about." What happens then? Among the texts at ChatGPT's disposal, it discovers a million posts on the same topic, as if asking, "You wanted something popular, right?" It then produces something similar, an average and paraphrased version. Consequently, paraphrasing those previous million posts results in the one-million-and-first post. Assuming we genuinely cannot distinguish this text from similar ones written by humans, this is precisely what may delight the author. However, should we truly be satisfied with such a "gem"?
Write a 50 word post about annoying auto-generated posts.
Auto-generated posts can be frustratingly generic and irrelevant. They lack the personal touch and creativity that make authentic content engaging. While they may save time, they often come across as spammy and insincere, turning off readers and damaging the reputation of the brand or individual behind them.
Thanks, Captain Obvious!
ChatGPT Will Make Me Jobless
Can ChatGPT take over any of your work tasks? Are you afraid of competition from AI?
Various experts forecast tectonic shifts in human mass. For instance, OpenAI predicts that soon 20% of the workforce, primarily those in highly compensated positions, will have half as many job opportunities. However, from where I sit at my desk today, I can tell you that nothing like this is anywhere near happening or foreseeable.
Yes, I use ChatGPT all the time. For example, it has quietly replaced all my favorite translators because it can do more. You can request it to "Translate in 5 versions" or "Translate as a 7-year-old child would say" or "Give me a similar common saying" and so on.
Your task is to give me 5 similar everyday phrases in English every time you see a phrase starting with "!"
! Daj mi pięć.
- Give me five.
- High five.
- Give me some skin.
- Slap me five.
- Gimme five.
This trick of replacing a long task wording with some character (here, "!") only works within the current chat, and unfortunately not for long.
Can ChatGPT automate anything else? Yes, can do some things. It can shorten a text or write a summary of it. It can also paraphrase the text, such as by retelling it in a simpler manner or with a certain context in mind. That would be good enough and save me time - it would be faster to edit than to do it myself.
Hickory, dickory, dock.
The mouse ran up the clock.
The clock struck one,
And down he run.
Hickory, dickory, dock.
One describes this event in one sentence to depict their emotions. You are not permitted to add any details that are absent in the text. Write 3 variants where emotions are: gambling, cynicism, and gluttony.
Gambling: The mouse's fate was decided by the strike of the clock in a game of chance.
Cynicism: The mouse's pointless scampering up and down the clock was a hollow and meaningless act.
Gluttony: The mouse's desperation to reach the top of the clock and find more food was all-consuming.
The two superpowers of chatGPT are the ability to retell the same thing indefinitely and the ability to take context into account.
What else does ChatGPT do for me? It helps me with my writing by providing a dialogue companion who is always available. There's nothing like discussing ideas to stimulate the development of new ones, and with ChatGPT as your conversation companion, you always have someone to talk to. While 80% of ChatGPT's responses are meaningless and useless, the remaining 20% can be unexpectedly valuable. So if you're stuck, ask ChatGPT: "What sounds unnatural here?" "What does this character really want?" "Why does the reader want this?" With a little help from your dialogue partner, you may reveal any issue from an unexpected angle. All of this can nudge you toward further inspiration.
Hickory, dickory, dock.
The mouse ran up the clock.
The clock struck one,
And down he run.
Hickory, dickory, dock.
One tells this story as a metaphor of their affection. You are not permitted to add any details that are absent in the text. Write one sentence description of the event in three variants where main drivers are: gambling, cynicism, and spiritual hunger.
Variant 1: Our love is like the mouse's gamble, running up the clock hoping for a win before retreating when time is up.
Variant 2: Our love is like the cynical mouse, running up the clock with doubt but quickly retreating when proven right by time.
Variant 3: Our love is like the spiritual hungry mouse, running up the clock seeking enlightenment but ultimately realizing that our time together is limited.
You can never have too much fun!
If you're stuck, you can get ChatGPT to try on all sorts of different contexts for your story. It's not easy or quick to do it in your head, and here you have a dozen options.
What will happen next? I can remember a time when there was no Google, and now we do a huge number of things completely differently, both at work and outside of work. All of these tremendous changes were happening before our eyes. We've ALREADY seen it happen.
Will we have less work? Surely not. We've ALREADY seen it happen. My grandmother used to balance debits and credits using an abacus. Did a calculator replace her? No. Now there is accounting software, which is continually being improved. Does this lead to fewer accountants in the world? Well, no. They work very differently now, but we still can't do without them.
Unplug This Thing Before It's Too Late
Are you concerned about any potential disasters caused by the ChatGPT that might affect you personally?
I couldn't bear not to bring up the story of the Future of Life Institute's open letter, which called for the suspension of experiments with AI chatbots. Adding fuel to the fire, Italy banned access to chatbots on its territory. What kind of trouble are we in for?
I've already mentioned the tsunami of auto-generated publications online. What else personally affects me? Will my kids-schoolchildren, be tempted to write all of their essays using ChatGPT? I wouldn't hold out for it. Universities are widely banning students from using the ChatGPT to write papers. See, applies to students, but scientists too! I was particularly amused by the irony of the situation when I heard that one of the most prestigious scientific conferences on AI refused to accept AI-generated papers.
However, is this really news to us? Students cheat, and unfortunately, so do some scientists. Various authors search for and find ways to write their texts without wasting energy. Any new tool is always greeted with enthusiasm by those looking for loopholes in prohibitions and restrictions, and every time we have to find a way to counteract it. Maybe that's what Italy is doing, and ChatGPT should be concerned about protecting personal data.
As for existential threats from AI, that's another story. Honestly, I'm a big fan of the drama genre, and it gives me great pleasure to read all the various responses to the Future of Life Institute's open letter. It's an amazing phenomenon. The responses number in the thousands, and they come from Nobel Prize winners, professors, designers of cutting-edge AI systems, politicians, and billionaire businessmen, not just freaks or tabloid journalists.
Can you imagine hearing something so dramatic from people like this: "Not everyone dares to say it out loud, but we're terrified of what we've created, and we're afraid that disaster can no longer be prevented, even if we personally quit now." Wow, what intense passion! Take a look at what was said in my favorite response. Its author also urges us to immediately start looking for a way to embed Isaac Asimov's three laws of robotics into neural networks and, at the same time, to recognize when consciousness will awaken in neural networks, at which point we will lose the moral right to own and use them. Wow!
Other speakers are equally colorful in drawing attention to the enormous pressure that IT giants are putting on each other in the AI-weapons race and claim that behind the open letter is a banal desire to slow down the development of a competitor.
To me, it seems like we've stumbled upon something we don't usually see happening. But hey, I'm thrilled to be witnessing this one-of-a-kind spectacle as an appreciative viewer. Probably we owe it all to competition. They're the reason we have the ChatGPT in our lives, and boy, am I grateful for that to the creators of this little gem, those pretty quirky geniuses, just like geniuses should be, and I highly appreciate it!! While it's still just a toy for now, we're looking forward to seeing what comes next!