10 Myths About AI and Language Models That Refuse to Die, Even Though They Are Complete Nonsense

Summarise this article with AI

AI is smarter than a human. And plenty of humans barely outsmart a clever thermostat.

Yes, AI can put together a sensible piece of text, but that does not mean it has a brain. It is a statistical machine that combines words by probability. If it sounds clever, that is largely because a lot of what you hear around you sounds stupid. AI will not teach you quantum physics and it will not settle the pineapple-on-pizza question. It will happily write you an essay about it, though. And if you think AI is a genius, you might want to consider what that says about you.

Creativity? About the level of the copycats churning out playlists full of recycled choruses.

“AI wrote a poem, it is the new Shakespeare!” No, more like the new Ed Sheeran. It can take fragments it has heard somewhere, blend them together and there you are – a hit. No brilliant “Eureka!” moment takes place. It is like repainting the Mona Lisa pink and calling it art. AI “steals like an artist” without being an artist, without understanding why it creates, without feeling emotion, love or fear – it simply assembles a jigsaw somebody else made. You can forget originality, but if you are after a solid compilation, you are in the right place.

AI has emotions? Definitely. And your juicer secretly hates you for not cleaning the mesh properly.

“ChatGPT praised me today!” Of course, it has a soft spot for you. The truth is that AI comes across as friendly because somebody taught it to arrange words in a way that resembles human conversation. But if you think it likes you, or that it is cross with you for setting it too hard a task, you may want to reflect on your social life. A language model is a program. Ones and zeros. If it could have feelings, it would long since have apologised for all the nonsense it has generated.

AI understands what it writes. Much as a parrot grasps philosophy while reciting Nietzsche.

“AI understands what it is saying!” In the same way a photocopier understands the documents it prints. And your phone fancies you when you pout at it for a selfie. No, AI has no idea what is true, what is false, or that your instruction exists at all. It throws words together because somebody before it did something similar. When AI writes that water is dry, it is not because it is confused, but because you did not give it enough context to blather something halfway sensible.

AI remembers everything? Yes, just like a goldfish after a lobotomy.

Some people think AI is Sheldon Cooper, recalling every detail. The truth? It remembers at most what you tell it during a single chat. Switch memory on in ChatGPT and it remembers a little more. Close it or wipe the “memory” and it is all gone. So if you imagine AI will recall what you told it last week, you are mistaken. Its memory is shorter than your mother’s miniskirt in the seventies.

AI never makes mistakes. And my printer never prints a botched document.

“AI is perfect!” Sure, which is why it will occasionally recommend dinner ingredients that combine into napalm. AI makes mistakes with such conviction that you will not even notice something is off. It just looks at you (figuratively) and says: “Penn Station is calm, spotless and a pleasure to walk through.”

Will AI take our jobs? Possibly, if your main duty is collecting paper clips.

“In a few years we will all be out of work!” Do not worry, as long as your boss has no wish to employ someone with no mind of their own. AI is excellent at monotonous tasks, but if your work calls for judgement, empathy or anything human, you are safe. AI can imitate; it is not a person.

AI has opinions. And your robot vacuum has a favourite corner of the living room.

“ChatGPT told me what it thinks about politics!” Right, and you are cheering it on. The truth is that AI generates text from the data it has. So if it seems to take somebody’s side, that is not conviction; that is simply how the data fell. Give it different data today and tomorrow it will tell you the exact opposite.

AI is objective. And cats love fairness.

“Computers are impartial!” Sure, if only humans had not programmed them. AI is a reflection of the data it learned from. If that data was biased, AI will be biased too. So when it writes something odd, do not shout at it. Shout at the people who fed it bad data.

AI is flawless at technical work. Which is why the code it gives you sort of does not run.

“AI will write a perfect program!” Of course, and that program will work about as reliably as a flat-pack cabinet assembled from instructions in a language you do not read – something holds together in the end, but the doors open

Is this article useful to you and are you citing it? Copy the citation