It seems that everyone is talking about artificial intelligence. But if you're feeling overwhelmed or unsure of what people are saying, don't worry, I can help you.
I asked some of the best AI journalists in the industry to share their best advice on how to talk about AI with confidence. My colleagues and I are obsessed with the technology every day, listening to AI experts and then translating what they say into clear, relevant text, including important context. So, we know a little bit about what we're talking about.
Here are seven things to keep in mind when talking about artificial intelligence.
Don't worry, it sounds stupid
Nitasha Tiku, a science and culture correspondent for The Washington Post, said: "Although the tech industry insists that large language models will change the world, it is not very good at explaining itself clearly. If you're confused, you're not alone. She added that the conversation about AI is rife with jargon, which is counterproductive. "Hallucination" is a special term for artificial intelligence systems to make things fictional; And "prompt engineers" are just those who know how to talk to AI to get what they want.
Tiku recommends watching YouTube explanations about AI concepts and models. She said: "Skip AI influencers and choose more low-profile hosts. If you're looking for something concise, IBM Technology is for you. Can help demystify how AI works. ”
No matter how you talk about AI, there will be complaints. Will Douglas Heaven, senior editor of AI at MIT Technology Review, said: "Sometimes it feels like the world of AI has split into many fan circles, and everyone is talking about it, sticking to their own definitions and beliefs, but not wanting others to understand. Figure out what AI means to you and stick with it. ”
Describe specifically what kind of artificial intelligence you are talking about
Karen Hao, China technology and society correspondent for The Wall Street Journal, said: "In public discourse, 'artificial intelligence' is often seen as a thing, but it is actually a collection of a hundred different things. ”。
Hao says it's helpful to distinguish which feature of AI you're talking about, so you can have more specific conversations: Are you talking about natural language processing and language models, or computer vision? Or a different application such as chatbots or cancer detection?
Billy Perrigo, a special correspondent for Time magazine, said talking about "artificial intelligence" as a single thing masks the reality of the technology.
"There are different models that can do different things, react differently to the same prompts, and each model has its own biases," he said. ”
Keep it real
"The two most important questions for new AI products and tools are simple: What does it do? How is it done? James Vincent, senior editor at The Verge, said.
Currently, there is a trend in the AI community to talk about the long-term risks and potential of AI. Vincent added that it's easy to be distracted by hypothetical scenarios and imagine what the technology might do in the future, but discussions about AI are often more pragmatic and focused on reality rather than hypothetical.
The tech industry also has a tendency to exaggerate the capabilities of its products. Douglas Haven said: "Be skeptical, skeptical. ”
This is especially important when talking about AGI or artificial intelligence in general, which is often used to refer to software that is as smart as a human (whatever that means in itself).
He added: "If something sounds like bad science fiction, maybe that's it. ”
Adjust your expectations
Language models that power AI chatbots like ChatGPT often "create illusions" out of nothing. Madhumita Murgia, an AI editor at the Financial Times, said it might be annoying and surprising to people, but that's how the models work.
It's important to remember that language models aren't search engines built to find and give "right" answers, and they don't have unlimited knowledge. Murja adds that given your problems and all the training they've undergone, they're just prediction systems that generate the most likely words.
Don't anthropomorphize
AI chatbots capture the public imagination because they generate text that looks like something written by a human and gives users the illusion of interacting with something other than a computer program. But in reality they are computer programs.
Chloe Xiang, a reporter for Motherboard, said it's important not to anthropomorphize the technology or attribute human traits to it. "Don't give it a (gender) pronoun, or say it feels, thinks, believes, etc."
Doing so only reinforces the misconception that AI systems are more capable and perceptive than they are.
I find this really error-prone because our language hasn't kept up with the methods of describing what AI systems are doing. When in doubt, I replace "artificial intelligence" with "computer program". Suddenly, you find out that a computer program is really silly for someone to divorce his wife!
It's all about power
Khari Johnson, a veteran contributor to Wired, said that while hype and bad things may dominate the headlines, it's crucial to think about the role of power when you're talking about AI.
Johnson said: "Power is the key to making the raw materials for artificial intelligence, such as computing and data; is key to questioning the ethical use of AI; It is also key to understanding who has the ability to earn an advanced degree in computer science and who leads the AI model design process. ”
Hao agrees. It's also helpful to remember that AI development is very political, involving a lot of money and competing research organizations, she says: "Sometimes the conversation around AI is not so much about technology as about people. ”
Don't think of it like a robot
Don't portray AI as a scary robot or an omniscient machine. Sharon Goldman, senior writer at VentureBeat, said: "Remember, AI is essentially computer programming by humans that combines large data sets with massive amounts of computing power and intelligent algorithms. ”
Support: Ren