The Main Flaws of Modern Day AI

Artificial intelligence has evolved so much since Alan Turing's concept of early AI in the 1930s. Nowadays, AI has become something that humans rely on constantly, whether for kids, it be help with homework, or for adults, just a friend who doesn't judge. However, as we approach 100 years of AI, there are still multiple flaws that must be resolved.

The biggest problem with AI is hallucination. For humans, hallucination is when a human sees something that isn't there. For example, once I lacked sleep and swore that I saw a lamppost swaying back and forth, like it was about to fall. The lamppost was actually firm on the ground, but I had hallucinated. Of course, the definition of hallucination for AI is different. Hallucination for AI means that AI makes something up and confidently exclaims that it's true. For example, I'm a big Pokémon fan, and I had just pulled an expensive Pokémon card from a set, and I wanted Copilot to guess the set from which I pulled the card. And Copilot continuously was guessing sets that don't exist and were not announced to be released. This example clearly demonstrates hallucination because AI genuinely thought that the sets that it made up were actual sets. Hallucinations were not meant to occur in AI, but since they do, people are currently working on fixing it.

Another major problem with AI is acquiescence. Acquiescence is when someone agrees to the opinions of other people, not because they agree with the opinion but because it is easier than arguing with it. The AI definition is quite similar. Have you ever noticed that AI sometimes agrees with whatever you say? That is acquiescence. Unlike hallucination, acquiescence was designed on purpose by AI engineers. First of all, hallucination is a bug in the code while acquiescence is a personality trait, and all personality traits were coded into the AI. But the reason why AI engineers allowed acquiescence to be a part of an AI's personality is because people get dopamine when they are right, and AI's acquiescence just keeps on giving a human's brain more and more dopamine. Therefore, they will keep on using the AI. However, a lot of people prefer to have a chatbot that doesn't agree to whatever they say as it feels like they are talking to an actual human.