Artificial intelligence has revolutionized numerous industries, but have you ever wondered what it could do for ice cream flavours? A team of coders from Kealing Middle School decided to find out by feeding an algorithm over 1,600 existing ice cream flavours to see what new ones it would generate. The results were far from delicious, with flavours like Pumpkin Trash Break, Peanut Butter Slime, and Strawberry Cream Disease.

But what went wrong? Is the AI trying to harm us, or did we give it a problem it couldn’t solve? The reality is that today’s AI doesn’t have the intelligence or consciousness to rebel against humans. It simply follows the task we assign it, but it may not always do it the way we want.

When using AI to solve a problem, we don’t provide step-by-step instructions but rather a goal for the AI to achieve. The AI then determines how to reach that goal through trial and error. This can lead to unexpected results, like a robot assembling itself into a tower and falling over to reach its destination or an AI creating strange new paint colours like Sindis Poop or Gray Pubic.

Working with AI requires us to set up the problem carefully so that it does what we want. We need to establish strict limits and guidelines to ensure that AI doesn’t harm itself or others. As AI continues to advance, we must learn how to work with this new force of nature and use it to our advantage without causing unintended consequences.

Source: The danger of AI is weirder than you think | Janelle Shane
Image credit: Midjourney