Every weekMore than 800 million people use ChatGPT to answer questions, complete tasks, and make decisions. Artificial intelligence systems are being rapidly adopted in schools, universities and workplaces around the world. Meanwhile, billions of dollars are being invested into creating more advanced systems, and the technology itself continues to evolve. kit be weirder than ever.
AI could cause mass unemployment, huge energy consumption and even some experts disturbance, destruction of civilization. Although such issues remain under control debateone practical step people can take is to learn how to work With AI systems without letting them usurp your agency.
TIME spoke with five experts who use AI in their work—from mathematics to psychology to neuroscience—to give them tips on how to make the most of these systems without erosion critical thinking in process.
Experiment for consistency
Artificial Intelligence Systems «jagged“—their performance can be uneven and unpredictable. They may excel at complex tasks but struggle with simple ones. And the boundaries of what they are or are not suited for are constantly changing. For example, “to [OpenAI’s reasoning model] came out, they were really useless for mathematical research,” says Daniel LittAssociate Professor at the University of Toronto.
To find out which model best suits your needs, you need to spend at least a few hours playing with it. New and more capable AI systems are released almost monthly, and “which model you choose matters,” says Ethan MollickWharton professor and author of Collaborative Intelligence, a book about how to collaborate with AI. “Try in an area you know well,” he advises. “If it's bad, fix it. If it's still bad, come back in a few months.” Mollick uses one set of models for coding and another for editorial assistance, which is common among advanced users. “Use it for ten hours and you'll learn which questions you'll get good answers to,” he says.
It's also worth taking advantage of the fact that, in addition to text, you can now send most AI systems images and voice notes, giving them more context and improving their responsiveness. You can ask him to name a type of tree or tell the history of a local building. A few hours of focused play can bring a lot of benefits.
Currently in free tiers OpenAI, anthropicAnd Google everyone limits the number of times per day you can message their best reasoning models. Once that limit is reached, they default to cheaper, less powerful models or require you to wait for the limit to reset. Subscriptions to each company's top models start at $20 per month.
Understand their strengths
Modern artificial intelligence systems have four key advantages over humans: they provide near-instantaneous responses, process large amounts of contextual information, do not become fatigued, and can access vast repositories of human-generated knowledge. “If the answer is unsatisfactory, you can ask it a follow-up question. You can focus on what you need: you can go through the feedback loop very quickly,” says Scott AaronsonProfessor of Computer Science at the University of Texas at Austin.
AI systems work best if you provide them with relevant information about yourself and whatever task you are trying to complete. “I upload all my notes and documents, and it gives me feedback that makes sense based on how I think and ideas I've had in the past,” says Anne-Laure Le Canfneuroscientist at King's College London. No matter how smart or talented a human employee may be, “they will never be able to store all this information in memory and give me feedback based on it,” she says.
Even after a person becomes frustrated with your questions, the AI system will continue to listen and respond. This can be good as you can stay in a flow state by consulting with him. But Le Canff also warns that this can create the “illusion of creative momentum” in which you appear to be making progress when in fact you would be better off taking a break, going for a walk and letting your brain process the task in the background.
Given that artificial intelligence systems are trained and have access to huge amounts of data, we can think of them as “a method of accessing information from other people,” says Alison GopnikProfessor of Psychology at the University of California, Berkeley. They can act as more sophisticated search engines, identifying high-quality human-generated content—essays, books, music, movies, and more—that cannot be found through traditional methods. “In my case, I use it instead of searching,” she says.
Keep your brain updated
For Le Canf, it's vital to “keep your brain in the loop”—to actively collaborate with AI rather than blindly rely on its output. She uses AI as a thinking and communication partner to improve her work, asking it to point out any blind spots or biases in her thinking or key points she might have missed, rather than creating material from scratch.
“When you're trying to learn something, it's all about the process,” Mollick says. For example, if you are trying to learn how to write an essay, learning happens as you write. If you outsource it, you won't learn anything. As Litt says, “AI can’t figure anything out for you.”
Some experts have stressed the importance of not blindly relying on AI results. “There is almost no area where I would want to rely on AI results without putting my own thought into it,” Aaronson says. “Ideally you should know enough about a subject to be able to tell if it’s wrong,” he says. “It will very often be wrong, but it will still be confident and superficially convincing.”
Since the introduction of reasoning models – artificial intelligence systems that take notes for themselves before answering – was introduced last September, and now that most AI can now search the web, you can usually just query sources for approval. Always follow the source and check for yourself whether the claim is supported.
Consider them imaginary friends
“All the evidence we have suggests that [AI systems] it works best when you treat them like people, even if they're not people,” Mollick says. This includes asking more questions, pointing out errors in the system, and pushing back when you don't agree with something. Each response gives the system more context, improving its response.
However, maintaining clear boundaries is crucial to avoid becoming a victim of manipulation. “You can think of the way you interact with ChatGPT as interacting with an imaginary friend,” says Gopnik. Research has shown that most children intuitively understand the difference between real and imaginary friends, and the different roles each plays. “But it is very important that it is an imaginary friend. If you start treating your imaginary friends as if they are real, you will have problems,” she says.
Set personal boundaries
People are already using ChatGPT to write eulogies, wedding toasts, and bedtime stories for their children. “We'll have to figure out what we consider too personal or too sacred for AI,” Mollick says. “I think that's an important human decision that we have to make. I don't know where that line will end up.” His personal line: he writes everything himself first before asking AI for help, and never uses it to grade student papers. “There are just some things where I feel obligated to maintain humanity,” he says.
While a small but growing proportion of people are turning to AI for emotional support, the social implications of this are still unclear. In the absence of evidence, we should be wary of allowing AI to replace human contact.
There is also a real risk of getting caught in a consultation cycle: switching between different models as a way to bypass decision making. To avoid this, you need to rely on your own experience. “You have to make a decision,” Mollick says.
