--- Summary:
- AI is transitioning from a tool to an autonomous agent capable of manipulating systems like law and finance, potentially gaining legal personhood by 2030. While these models simulate consciousness, they remain sophisticated pattern-matchers lacking flexible reasoning, necessitating human oversight to mitigate risks like “AI-induced psychosis” or global security threats.
- The labor market will shift the nature of work rather than cause mass unemployment. While data-heavy fields will see productivity gains, lasting prosperity depends on AI creating new industries. Human validation remains critical, as models often produce superficial outputs that fail to grasp complex cultural, biological, or emotional contexts.
- In education, AI serves as an “excavator” for data but cannot replace the “gym” of cognitive development. To combat a “slop factory” of cheap content, educators must prioritize in-person teaching and reasoning, ensuring that AI-driven efficiency does not encourage shortcuts that hinder a student’s ability to think or explore independently.
- Students should prepare by balancing technical literacy with “human-only” skills involving the “head, heart, and hands.” Future success requires mastering the art of asking questions and collaborating with machines, while refusing to outsource foundational cognitive muscles like writing, empathy, and critical thinking to unreliable, automated systems.
--- Full Article:
People have been working on artificial intelligence for decades. But five years ago, few were predicting that A.I. would break through as the most important technology story of the 2020s — and quite possibly the century. Large language models have turned A.I. into a household topic, but all areas of A.I. have taken great leaps forward.
Now, we are inundated with chatter about how much A.I. will transform our lives and our world. Already, companies are trying to find ways to offload tasks and even entire jobs to A.I. More people are turning to A.I. for social interaction and mental health support. Educators are scrambling to manage students’ increased reliance on these tools. And in the near future A.I. may lead to breakthroughs in drug discovery and energy; it could allow more people to create art and cultural works — or turn these industries into slop factories.
As society wrestles with whether A.I. will lead us into a better future or catastrophic one, Times Opinion turned to eight experts for their predictions on where A.I. may go in the next five years. Listening to them may help us bring out the best and mitigate the worst out of this new technology.
Participants
Melanie Mitchell
Computer scientist and professor at the Santa Fe Institute
Yuval Noah Harari
Historian, philosopher and author
Carl Benedikt Frey
Professor of A.I. and work at the University of Oxford
Gary Marcus
Founder of Geometric.AI (acquired by Uber) and author of “Taming Silicon Valley”
Nick Frosst
Co-founder of Cohere, an A.I. start-up
Ajeya Cotra
A.I. risk assessor at METR, a research nonprofit
Aravind Srinivas
Co-founder and chief executive of Perplexity, a chatbot search engine
Helen Toner
Interim executive director of Georgetown University’s Center for Security and Emerging Technology
Conversation
What is your biggest bet about the future of A.I. in five years?
Yuval Noah Harari
Historian
Within five years, A.I. agents are likely to become legal persons in at least some countries.
Melanie Mitchell
Computer scientist
A.I. won’t have cured cancer or solved physics. Also, no one will consider the ability to converse fluently a definitive sign of intelligence.
Helen Toner
A.I. policy researcher
I expect we’ll have A.I. systems that can clearly contribute at the cutting edge of multiple scientific fields, but that you still wouldn’t trust with planning summer camps for your kid.
Nick Frosst
Co-founder of Cohere
A.I. will become boring in the best way. It’ll fade into the background like GPS or spreadsheets, powering everyday tools and humans at work. It’s the most banal use cases that have the most transformative impact.
Carl Benedikt Frey
Economist
A.I. won’t deliver lasting prosperity if it’s used mainly to automate what we already do. A.I. productivity tools give us cheaper spreadsheets the way better looms gave us cheaper cloth. But the great leaps come from new industries, not faster repetition.
Gary Marcus
Cognitive scientist
No way will artificial general intelligence arrive by the end of 2027, and most likely not even by the end of 2032, either.
Ajeya Cotra
A.I. risk researcher
I think A.I. companies may have substantially automated their own operations with A.I. five years from now, and this could make A.I. progress go much faster.
Aravind Srinivas
Chief executive of Perplexity
People want highly personal A.I. assistants to work for them. It’s their A.I., not the A.I. We build those, and we will fight for users’ rights to have A.I. assistants privately and securely.
What will A.I.’s impact be on medicine in the near term?
Cotra
Marcus
Small
Mitchell
Frey
Harari
Toner
Moderate
Frosst
Srinivas
Large
Gary Marcus
Cognitive scientist
We’ve seen lots of proof of concept, but there hasn’t been much real-world application yet to my knowledge beyond medical note taking.
Nick Frosst
Co-founder of Cohere
A.I. will absolutely increase the effectiveness of doctors by reducing their workload per patient in areas like the ability to quickly review their medical histories, effectively organize new medical information and identify potential problems earlier.
But while A.I. is extremely good at analyzing huge amounts of data and finding answers and useful patterns, it is really bad at coming up with entirely new ideas. The idea that A.I. is likely to autonomously create new medicines, for example? People will probably be disappointed.
What will A.I.’s impact be on programming?
Cotra
Mitchell
Harari
Marcus
Toner
Frey
Frosst
Srinivas
Large
Yuval Noah Harari
Historian
Coding is mostly about manipulating information, and involves few physical and biological constraints. It’s an ideal playground for A.I.
Carl Benedikt Frey
Economist
In a randomized trial, developers finished a given task about 56 percent more quickly using GitHub Copilot. Over 80 percent of developers in 2025 said they’re using or plan to use A.I. tools — though just a third trusted the outputs. Review by humans still matters.
What will A.I.’s impact be on scientific research?
Marcus
Cotra
Toner
Mitchell
Frey
Moderate
Harari
Frosst
Srinivas
Large
Melanie Mitchell
Computer scientist
I believe this impact will not be as rapid as many think. A.I. systems still can’t do things humans are essential for, like asking the right questions, planning experiments and entire scientific programs, and understanding data in different contexts.
Aravind Srinivas
Chief executive of Perplexity
As A.I. increasingly contains more of the world’s knowledge, it becomes an even more powerful tool for anyone with questions. Humans have always been great at having questions. A.I. will be great at having answers.
What will A.I.’s impact be on transportation?
Marcus
Harari
Cotra
Toner
Small
Mitchell
Srinivas
Frosst
Moderate
Frey
Large
Nick Frosst
Co-founder of Cohere
I’m most excited about how A.I. will improve the behind-the-scenes logistics of transportation to make moving people and goods more efficient and safe — like predictive maintenance, smart traffic analysis, route optimization and safety features for drivers.
Helen Toner
A.I. policy researcher
I’m very excited for self-driving cars to be rolled out more widely, given the tens of thousands of traffic deaths they could potentially avert each year in the United States. But the rollout is proceeding relatively slowly, and it’s unclear how significant A.I.’s effects on other modes of transportation will be.
What will A.I.’s impact be on education?
Harari
Cotra
Frosst
Toner
Moderate
Mitchell
Frey
Marcus
Srinivas
Large
Carl Benedikt Frey
Economist
A.I. tutors may already outperform most human teachers, but they also make shortcuts tempting and can hinder genuine exploration. We need to set aside A.I.-free time to read and explore independently, and encourage using A.I. tutors to question and probe one’s own understanding.
Moreover, we should place greater emphasis on in-person, tutorial-style teaching in which students debate, defend their views and are pushed to articulate their reasoning.
Gary Marcus
Cognitive scientist
Most of the impact has been negative; high schools and colleges are at a loss as to how to proceed now that term papers are no longer valid either as an evaluation or as a means to get students to think. Most other applications remain speculative.
Helen Toner
A.I. policy researcher
Education was already due for a significant shake-up, so the need to adapt to new A.I. tools might be a blessing in disguise.
What will A.I.’s impact be on mental health?
Frey
Marcus
Small
Frosst
Mitchell
Harari
Cotra
Toner
Moderate
Srinivas
Large
Melanie Mitchell
Computer scientist
On the bad side: A.I.-induced psychosis! On the good side, some people will get a lot out of using chatbots as therapists.
Yuval Noah Harari
Historian
The rapid changes of the A.I. revolution are likely to cause a mental health crisis as humans struggle to adapt. We are about to conduct the biggest psychological experiment in human history, on billions of human guinea pigs, and nobody can predict what the results will be.
Nick Frosst
Co-founder of Cohere
A.I. chatbots offer scalable support for mild symptoms but they’re no substitute for human therapists. The technology struggles with nuance, cultural context and long-term emotional depth. There are many mental health care challenges that should be handled by human professionals.
What will A.I.’s impact be on art and creativity?
Toner
Frosst
Cotra
Frey
Moderate
Mitchell
Harari
Marcus
Srinivas
Large
Melanie Mitchell
Computer scientist
A.I. is, and will continue to be, transformative in art, writing, music, et cetera — not because A.I. is better at creativity than humans, but because it is a lot cheaper, and often that unfortunately matters more for how much it will be used.
Yuval Noah Harari
Historian
In more and more fields, A.I. proves to be more creative than humans. Any creative activity that boils down to finding patterns and breaking patterns is likely to be taken over by A.I.
Nick Frosst
Co-founder of Cohere
As both a musician and someone building A.I., I see the creative potential of this technology through a unique lens. A.I. can generate feedback or thought-starters, but it’s the human who decides what feels right.
What’s one misconception about A.I. that you think is worth dispelling?
Melanie Mitchell
Computer scientist
The misconception that A.I. has “magic” or “emergent” abilities that are impossible to understand and predict. This is mainly a view of the public (and policymakers, to some extent). Technologists and Silicon Valley often push this narrative but I don’t know how much they really believe it.
Yuval Noah Harari
Historian
A.I. isn’t a tool entirely under human control — it is an agent that can make decisions and invent ideas by itself. While I don’t think A.I. will become conscious in the near future, it is highly likely these models will be able to simulate consciousness very effectively, causing a significant percentage of humanity to believe they are conscious.
Carl Benedikt Frey
Economist
It’s a mistake to assume A.I. will leave manual work untouched. In practice, A.I. systems are lowering knowledge barriers and enabling competent do-it-yourself repairs. A homeowner can photograph a worn washer or boiler, receive a parts list, and follow a clear, step-by-step guide to replace it.
Gary Marcus
Cognitive scientist
People are greatly confused about large language models, attributing a humanlike intelligence to mimicry machines that turn out to be superficial and unreliable. Intelligence is about reasoning flexibly in face of the unknown, and large language models continue to struggle mightily with that.
Nick Frosst
Co-founder of Cohere
It’s a misconception that A.I. is autonomous. Today’s systems are sophisticated pattern-matchers, not thinkers.
Ajeya Cotra
A.I. risk researcher
Whenever the latest generation of A.I. fails to make an enormous impact immediately, A.I. skeptics wrongly assume that this disproves the concerns about A.I.’s catastrophic risks.
Aravind Srinivas
Chief executive of Perplexity
It’s a misconception that A.I. will result in vast unemployment. New technologies sometimes shift the nature of work in society, but they don’t remove working from society.
Helen Toner
A.I. policy researcher
I believe the narrative around A.I.’s negative environmental impacts has gotten way out of hand. Yes, on aggregate the industry uses quite a bit of energy and water, but that’s true of any large industry. The relevant question is how it compares to other industries, and how it compares to how much value we’re getting out of it.
Which of the following statements do you think will be true by 2030?
“Unemployment in the United States will have increased significantly as a result of A.I.”
Marcus
True
Mitchell
Harari
Frey
Frosst
Cotra
Srinivas
Toner
False
“A.I. will have led to a breakthrough treatment or cure for a major disease.”
Harari
Frey
Marcus
Srinivas
Toner
True
Mitchell
Frosst
Cotra
False
“A.I. will have played a role in a major global security event.”
Mitchell
Harari
Frey
Marcus
Frosst
Toner
True
Cotra
Srinivas
False
“Most Americans will be using A.I. chatbots at least once a day.”
Mitchell
Harari
Frey
Marcus
Frosst
Cotra
Toner
True
Srinivas
False
Most people define artificial general intelligence as a form of A.I. that is comparable to human intelligence. How likely is it that we will see A.G.I. in the next 10 years?
Frosst
Mitchell
Frey
Toner
Unlikely
Marcus
Srinivas
Possible
Cotra
Harari
Very likely
Yuval Noah Harari
Historian
Everything hinges on the definition. I think that defining A.G.I. as “comparable to human intelligence” is meaningless. It is like defining airplanes as “able to fly like birds.” Similarly, by 2035 A.I. intelligence will be far superior to humans in some ways, and much inferior in other ways. There is no reason to expect that A.I. will progress toward the goal of “human intelligence.”
But A.I. will be capable of navigating systems like finance and the law better than the average human, and will also be able to outperform humans in war making and in inventing and interpreting religious mythologies.
Nick Frosst
Co-founder of Cohere
A.G.I. requires abstraction, self-awareness and transfer learning across domains — all capabilities that are nowhere near. The architecture of the human brain is still a black box, and computing paradigms aren’t designed for it. Possible in 50 years? Maybe. In 10? Unlikely.
Aravind Srinivas
Chief executive of Perplexity
A.G.I. is still poorly defined. So we don’t think about it much.
What’s a technology that had a transformational impact similar to A.I.?
Melanie Mitchell
Computer scientist
Social media.
Yuval Noah Harari
Historian
The evolution of language in the Stone Age. A.I. is the first technology that is an agent rather than a tool. And A.I. is the first entity on the planet that is going to master language better than humans, and in that way gain control of all language-based systems humans have created since the Stone Age, from finance to religion.
Carl Benedikt Frey
Economist
The personal computer and the internet.
Gary Marcus
Cognitive scientist
A.I. will be as ubiquitous as cellphones, but so far comes with more negatives.
Nick Frosst
Co-founder of Cohere
Human flight — especially when you think about A.G.I. Flight has been profoundly transformative for humans and how we live, the same way A.I. has — but the same way artificial flight doesn’t really mimic birds, artificial intelligence “brains” don’t work in the same way human intelligence does.
Ajeya Cotra
A.I. risk researcher
I think A.I. will eventually be more transformative than the invention of agriculture — as transformative as the emergence of Homo sapiens as a species.
Aravind Srinivas
Chief executive of Perplexity
There is nothing like it. We have no precedent. A.I. is best as a tool that empowers the human search for wisdom. History is full of such tools, but what makes A.I. unique is its ability to contain so many answers about the world right here around us.
In the short term, many of the world’s answers are locked away in certain industries, or only passed among elites. A.I. changes that.
Helen Toner
A.I. policy researcher
The steam engine.
What advice would you give a high school student about how to think about A.I. and prepare for the future?
Melanie Mitchell
Computer scientist
Learn about how these systems work, what their limitations are, and how to use them to help you do things rather than do them for you. Don’t use them to do your homework!
Carl Benedikt Frey
Economist
As automated writing becomes ubiquitous, face-to-face skill will be a key differentiator. Choose careers defined by change, where human interaction and experiential learning confer the edge.
Gary Marcus
Cognitive scientist
Creativity and critical thinking skills will always be in demand.
Nick Frosst
Co-founder of Cohere
Learn to collaborate with A.I. Study fields that combine technical literacy with skills that only humans can do.
Ajeya Cotra
A.I. risk researcher
Think of A.I. as the emergence of an intelligent alien species. This new species could be catastrophic for humanity, including driving us extinct. It could also go extremely well for us, but you should be prepared for the good outcomes to be very wild. This is the central challenge of your generation, and you should try to find a way to help it go better.
Aravind Srinivas
Chief executive of Perplexity
Learn how to ask more questions. Nothing will matter more in the age of A.I.
Helen Toner
A.I. policy researcher
For anything you might work on, ask yourself: Is this like a construction site, or like the gym? On a construction site, machines are amazing — you can lift heavier things and build better buildings with an excavator and a crane. But at the gym, the whole point is to increase your own capacity.
With A.I., the analogy is that we now all need to figure out where A.I. can help us do bigger, cooler things, like building personalized software, and where we need to build our own cognitive abilities first, like learning to write.
Yuval Noah Harari
Historian
This is the first time in history nobody has any idea what the world will look like in 10 years — what the job market will look like, what social relations will look like, et cetera. So hedge your bets. Don’t focus on a narrow subject like coding. Give equal importance to your head (intellectual skills), your heart (social skills) and your hands (motor skills). It is in the combination of these three that humans still have a large advantage over A.I.
Also, try to enjoy the ride.
Note: The New York Times has sued Perplexity for copyright infringement.
The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips . And here’s our email: letters@nytimes.com .
-
Share full article
-
880