It’s time to realize what AI can and cannot do.

The hottest technology right now, a type of artificial intelligence that generates human-like text or images, can sometimes give you that immediate “aha” jolt. But this is often not the case.

I had the experience – maybe you have too – of feeling disappointed when I asked ChatGPT to help me plan my vacation. Recently, I couldn’t remember the name of a pastry and tried to describe it to three different AI chatbots. It did not work. (The pastry was a frangipane tart.)

Such disappointments do not mean that AI is useless. This can be practical. But there’s often a disconnect between the reality of AI and the way companies encourage you to think of their AI as magical brains that know and do everything.

The truth is that AI is fundamentally bad at many tasks. This requires you to learn the right words to get the most out of it. Like all computers, AI will make different mistakes than humans, but it will make them. And the AI ​​imposed on you is sometimes just plain broken.

The lesson is to become familiar with what AI can and cannot do, so as not to be disappointed.

And it’s helpful to see that companies backtrack when AI doesn’t work as well as they promised. They know that AI isn’t magic and you should too. Here are some examples:

• Nine months ago, Amazon announced that it had overhauled its Alexa voice assistant with new AI. But Alexa AI repeatedly botched questions during demonstrations and is still not available on Alexa home devices. (Amazon founder Jeff Bezos owns The Washington Post.)

Some people at Amazon are concerned that the new Alexa won’t be ready for the planned September launch because the voice assistant still gives unpredictable responses, my colleague Caroline O’Donovan recently reported.

Amazon said it has tested the new Alexa with small groups of customers and is working hard to expand it to more people.

(Apple said this month that its AI-powered Siri would start appearing on some newer iPhone models this year. We’ll see how that goes.)

• Microsoft made a big splash last month about a new line of AI-powered personal computers, including time machine functionality for everything you did on your PC. The company later disabled this feature by default after some researchers said it was a security nightmare.

Microsoft also reduced a much-touted AI keyboard button and stopped integrating its chatbot into Windows PCs for tasks like changing the brightness of your screen. (Microsoft said it responded to comments that people didn’t like mixing a chatbot with other tasks such as settings.)

• Last month, Google reduced the prominence of AI-generated answers in web search results, with some answers being nonsensical or dangerous. This was at least the third time Google acknowledged that its AI wasn’t working as well, including a now-suspended AI feature that generated images that defied history, such as a female pope.

The pattern of AI setbacks and frangipani disappointments calls for a rethink of how companies present their technology to you and your expectations.

First, businesses need to be clear that AI can be useful for some things but not everything, especially factual information. AI, for example, should not be used to find out the height of the Eiffel Tower or find out who won the 2020 US presidential election.

Second, when companies publicly demonstrate their AI, they should tell you error rates or back up claims that seem surprising.

Independent researchers recently questioned OpenAI’s claim that its ChatGPT can pass the bar exam with a higher score than almost any test taker and questioned the idea that the chatbot can write software code like human experts. (OpenAI did not respond to my request for comment.)

Finally, it is important to recognize that your imagination can exceed the capabilities of the AI.

In a recent interview with NPR affiliate KCRW, host Madeleine Brand asked me what task she would like AI to do: Could AI tell from your calendar that your mother’s birthday is coming up and order flowers to be delivered to her house?

Sorry, I don’t know of any widely available AI that can accomplish this task today.

You know from your life experience that technology can be amazing, but it is rarely a silver bullet. And neither is AI.