平特五不中

Subscribe to the OSS Weekly Newsletter!

Looping in to Chat GPT

Playing around with this 鈥渋ntelligent鈥 chatbot is a lot of fun but it is not quite ready for prime time.
Image by Shutterstock.

Like so many, I have been experimenting with Chat GPT, the artificial intelligence technology that can answer questions or produce articles on a given topic in humanlike conversational language. In some cases, I have been impressed, much less so in others.

When I asked for an article on the contributions to science of Prince Albert, Queen Victoria鈥檚 husband, Chat GPT produced quite a good one, mentioning his organizing the Great Exhibition of 1851 and his championing scientific education, but failed to mention the role Albert played in bringing noted German chemist Auguste Wilhelm Hoffman to London to be the first director of the newly established Royal College of Chemistry where many discoveries would be made, including William Henry Perkin鈥檚 accidental discovery of mauve, the world鈥檚 first synthetic dye. I would say the Chat GPT article was good, but not great.

Next, I asked it to produce an article about the cereal Froot Loops, since I know of a number of interesting features that I would address if I were writing such a piece. Chat GPT鈥檚 first attempt was unremarkable, basically spewing out information that seems to have been gleaned from Kellogg鈥檚 promotional literature. It concluded with 鈥淪o next time you鈥檙e seeking a burst of fruity goodness to start your day, reach for a bowl of Froot Loops.鈥 Well, there is no fruit in there, and not much goodness either, given the 12 grams of sugar per serving. An uncritical recommendation of this cereal is out of line.

Chat made no mention of the widespread belief that the original name of the cereal was 鈥淔ruit Loops鈥 that was changed to 鈥淔root Loops鈥 as the result of a lawsuit, 鈥淧axton v Kellogg鈥檚,鈥 that alleged the name was misleading because there was no fruit in the product. Actually, there is no evidence that any such law suit was ever filed. Froot Loops was introduced in 1963 and was never called Fruit Loops. This controversy should have been part of any discussion of Fruit Loops history with a reference to the 鈥淢andela Effect,鈥 a phenomenon in which a large group of people share a false memory such as Nelson Mandela dying in prison.

Neither was there any mention of actual, documented lawsuits that claimed deceptive marketing on part of Kellogg鈥檚 that led consumers to believe that the product was packed with real fruit. One judge dismissed a case declaring that 鈥渢he use of the word 鈥淔root鈥 in connection with loops does not suggest the presence of real fruit because 鈥渇root鈥 is not real, and fruit, which is real, does not come in loops.鈥 Another such case was dismissed by another judge who opined that 鈥渁 reasonable consumer would not be confused by the allegedly frooty loops because the ring-shaped cereal does not resemble any known fruit.鈥

I would also have expected a discussion of the colours and flavours of the 鈥淟oops,鈥 particularly of the fact that while there are six different colours, all the loops are flavoured with the same blend of fruit flavours and they all taste the same. Any history of 鈥淔root Loops鈥 should also have included the change from artificial food dyes to vegetable colours motivated by studies, albeit questionable ones, that linked artificial dyes with adverse behaviour in children.

If I were writing about Froot Loops, I would also include a bit of whimsy such as the story of GloZell Green, a YouTuber who once entertained her four million followers by getting into a tub filled with milk and two big bags of Froot Loops and eating the cereal straight out of the tub. Then there was the pizza shop in Iowa that came up with a pizza topped with mozzarella, Greek yogurt and Froot Loops. That put the eternal question of pineapple on a pizza on the back burner. (As an aside, there has even been a bizarre class-action lawsuit against the maker of a 鈥淕reek yogurt鈥 claiming that it was not imported from Greece, nor was it made by Greeks.)

Giving it another chance, I asked the chatbot to write an article on Froot Loops, the Mandela effect, lawsuits and artificial colours, thinking that it would come up with a better piece. It actually didn鈥檛. It now claimed that the different colours had different fruity flavours and that the lawsuits were due to the artificial colours causing allergic reactions and made the error of stating that originally the cereal was called Fruit Loops. Bzzzzz.

Next, I wanted to see what my chatty friend would do with a real scientific topic, so I asked for an article on artificial vanilla flavouring. I have to say it performed admirably! It accurately described that natural vanilla flavouring is extracted from the vanilla bean and that its major component is vanillin, a compound that can also be synthesized in the laboratory. The first synthesis from coniferin, found in pine bark, was correctly attributed to German chemist Ferdinand Tiemann in the 19th century. Current synthesis from guaiacol derived from wood lignin was also described as was a method to produce vanillin through bioconversion using certain strains of bacteria to transform ferulic acid from rice bran into vanillin through enzymatic reactions. Bioconversion.

Also accurately explained was the fact that that natural vanilla flavour is a complex blend of over two hundred compounds with vanillin being responsible for the dominant flavour. Importantly, the article correctly pointed out that synthetic vanillin is identical to the natural version and poses no safety risk. I give Chat GPT high marks on this one. But then I asked for an article on the chemistry of skunk smell and while it was mostly correct, it named the compound responsible for the smell as skatole, which is wrong. Skatole is the odour of fecal matter. And therein lies a problem. As soon as you find a mistake, it calls into question any information that is delivered by chat GPT.

Finally, I asked for an article on 鈥渢he unreliability of Chat GPT". I got an excellent article that mentioned lack of common sense and contextual understanding, sensitivity to input and bias, possibility of generating plausible but false information and lack of ethical and moral understanding. It seems that unlike some people, it knows its limitations.

No doubt this application of artificial intelligence is fascinating, and the technology will continue to improve. But for now, let鈥檚 just call it an interesting plaything.


Back to top