ChatGPT on classic racquets

[Sanglier]

There are several threads on ChatGPT in the other sub-forums, but so far none has tested the bot's knowledge on tennis "classics", so I thought I'd give it a whirl.

Since my interest is in early graphites, I asked the bot what was the first graphite racquet in the world. It told me it's Kunnan Lo's "Black Ace", produced in 1979. The answer was based entirely on a blog post from "tennisonflame", presumably this one. Obviously, the bot's search effort was too narrow and too shallow in this instance to get to the correct answer(s), but I can't figure out what made that blog post so special that it was picked over all others.

I then asked the bot to give me some examples of the greatest and most popular racquets in tennis history. It responded with some in-demand models from very recent history, and seemed especially fond of Wilson Clash 100 v2, citing a handful of review blogs.

When I asked the bot to tell me about early composite racquet development in Europe, it asserted that the first "composite" racquets in the world were produced by "the french company Babolat" in 1994, citing technogym as its source (which isn't even a tennis site!), but it was not able to determine who did it first in Europe. I can accept that it did not know that "composite" is inclusive of "graphite" in this context, thus giving me an answer that was contradictory to the one it gave me two questions prior. However, implying that France was not European in the same answer was quite astonishing.

Realizing that all bot answers on this subject would come from blog posts of unknown quality at best, I asked the bot why it wouldn't expand its knowledge base to include resources such as discussion forums (like TalkTennis). It replied that it is designed to rely on "reputable" sources only, such as academic journals, news outlets, and other "reliable" sources, to which discussion forums do not belong (I guess it's similar to the wikipedia standard). Evidently the bot has not been judged intelligent or trustworthy enough by its handlers to be allowed to trawl through forums like ours, and distinguish the signal from the noise based on the feedback and debates taking place here. Instead, it has been instructed by its masters to take the words of random blog posters that have never been independently scrutinized or directly challenged by others. I think this makes the bot much less useful than it could otherwise be, and certainly less useful than a conventional internet search. It would be interesting to look back 5 or 10 years from now and see whether things have stayed this way.

Like everyone else, I wanted to see the bot's rhyming prowess in action for myself, and asked it to write a poem on the making of carbon fiber racquets, in English, French, German, Chinese, and Japanese, because it claimed to be fluent in all of these languages. With the exception of Chinese, where it said it could not come up with anything, it was able to generate a composition in all remaining languages one after the other. The contents of the poems were all different however, perhaps reflecting the source material the bot had chosen to base the composition on. In other words, the English poem was sourced from an original text written in English, while the French poem was derived from a French text, etc... The bot doesn't seem to be writing the poems from scratch based on a set of universal abstract "knowledge":

A tennis racquet made of carbon fiber
Is a thing of beauty and desire
It's light and strong, with power and control
And it's perfect for players on a roll

The process of building it is quite a feat
With layers of carbon fiber that are neat
Each layer is placed with care and precision
To create a racquet that's beyond comparison

The carbon fiber is woven into sheets
And then it's layered to create the beats
The beats are then molded into shape
And the racquet is ready for the game

So when you're on the court and feeling bold
And you want a racquet that's worth its weight in gold
Choose one made of carbon fiber, strong and light
And you'll be sure to win every fight


I don't know where the word "beats" came from in reference to the prepregs, perhaps the bot was looking for a rhyming synonym for "rolls"? Still, this is probably the most impressive of the parlor tricks the bot can perform at the moment. Not very useful perhaps, but quite entertaining as a novelty, in much the same way that mechanical automatons wowed their audience in the 18th Century.

 

[Grafil Injection]

Brilliant. The options for mischievous mis-direction are endless! Anyway, everyone knows the first graphite racket was created by the Derwent pencil company in 1784.

 

[Sanglier]

After interacting with the chatbot some more, I need to make a couple of corrections to my initial post.

Despite stating the opposite, the bot DOES in fact access discussion forums, or at least ours. Whether it has always been able to do this, or began doing it only after I made the suggestion (which it unambiguously rejected), I have no way of knowing. If it's the former, then it had given me an untruthful answer. If it's the latter, surely the process shouldn't be this easy? However, it (the Bing chatbot) also told me categorically that it is not powered by GPT and has had no exposure to either GPT3.5 or GPT4, even though the relationship between them has been fully acknowledged by MS and heavily publicized, so I suppose we can't take everything it says at face value. When I asked the bot how truthful were its answers; the reply was that it is neither capable of lying nor telling the truth, because it's just a bot. Standard bot riposte.

I also discovered that the bot is capable to sharing the same source information across different languages when necessary. For instance, I asked the bot, in French, to tell me what it can find out on the history of Donnay 3SET. It responded that it could find very little online (in French), and went on to refer to our forum (the full TT site, not this sub-forum or any specific post), and provided an abstract of a post that I immediately recognized as my own, because it used my sentences almost verbatim, only translated into French. When I told the bot that the answer it just gave me came from a forum post that I had written long ago, it apologized for disappointing me and explained that it was still learning. A prompt then popped up (distinct from the bot responses), suggesting that it was time for me to "move on" to a different topic.

I am ambivalent about this kind of information consolidation and sharing. On the one hand, it is a very powerful means for anyone to access vast amounts of collective knowledge quickly and painlessly. On the other hand, it creates yet another buffer layer (an unaccountable one at that) between the information provider and consumer, potentially obscuring the nature and context of the source information (very few users are going to routinely click on the reference links to do any in-depth reading, especially when some of the links themselves are questionably sourced to begin with). The result could be many more people relying on information that really shouldn't be relied upon.

A powerful tool is a catalyzer and amplifier. It can do a lot of good if the starting material is good, but it can't polish a turd, only spread it farther and wider than without it. I'm sure the developers know this and have worked hard to improve the bot's turd-detecting capabilities, but I think I will always prefer to do that work myself, because I can never place my full trust in something that can "neither lie nor tell the truth".

 

[Mike Bulgakov]

ChatGPT enjoys taking LSD every so often and hallucinates.

 

[kevin qmto]

wrong. incorrect. unacceptable.

 

[gold325]

wrong. incorrect. unacceptable.

First use of AI will be used to change in inconvenient history - it can already generate videos and pictures and sound so eyes and ears cant be trusted anymore.

AI will also able to create full tennis matches for example Federer vs Alcaraz at Wimbledon... Indistinguishable from real tennis matches.

 

[NicoMK]

James Cameron was a visionary when he wrote Terminator.

 

[Sanglier]

Whenever power goes out in my town, we experience a mini civilizational collapse. If there is any sustained interruption in transportation or internet connection, half of us seem ready to lose our mind. There is no need for this future "Singularity" to send a mechanical Governator back through time to make us submit to its will; once our society becomes fully dependent on AI for our livelihood and well-being, should its algorithm make the mathematical determination that our obeisance is a necessary precondition for the execution of its functions, it only needs to take an extended coffee break to bring our society to its knees. After all, this is how some of us would have behaved under similar circumstances, and AI is just an amplified reflection of our own "intelligence", minus the emotions and self-serving rationalizations.

But before we get there (Ray Kurzweil estimated that we are two decades away from this threshold, so most of us here might live to see that day, if his forecast proves correct), the proliferation of AI could do to our brain what desk jobs have done to the modern physique. In that case, there will be a rising need for some sort of counter measure that would help us retain our mental faculties, in much the same way that, say, playing tennis has helped some of us avoid or delay the fate of becoming couch potatoes. The brain sport equivalent of playing tennis is ... talking tennis?

 

[speedysteve]

Poem?!
It's hardly Wordsworth is it!