The Next Five Years of Artificial Intelligence
My Answers to a New York Times Survey
Earlier this month, the New York Times surveyed eight leading AI thinkers, entrepreneurs, and researchers about their predictions for the next five years. The group included some of my favorite intellectuals including Yuval Noah Harari, the author of Sapiens and Nexus, Melanie Mitchell, the author of Complexity and Artificial Intelligence, and Gary Marcus, an AI researcher and cognitive scientist.
I enjoyed reading the responses and found myself surprised when I occasionally disagreed with the majority opinion. Each person answered the questions by indicating whether or not they felt AI would make a small, moderate, or large impact before elaborating.
This seemed like a great exercise to evaluate my own feelings on the future of AI.
Do you agree or disagree with my answers? What did I miss?
What is your biggest bet about the future of AI in five years?
We will not discover artificial general intelligence by 2030.
People will lose jobs, whether due to AI replacing humans or because corporations will use it as an excuse to fire them.
The impact of AI will be mixed.
What will AI’s impact be on medicine in the near term?
AI will become more widespread at analyzing visual data for diagnosis, especially in the fields of dermatology, radiology, and pathology. In fact it is already doing this and causing a deskilling in the process.
AI scribes will help improve medical documentation, saving some time for clinicians, while increasing workload in other ways. But it will also help hospitals bill more efficiently and allow insurers to strategically raise premiums based on healthcare usage, medical history, geographic location, etc. Not all impacts will be beneficial.
AI will be more effective in helping nurse practitioners and physician assistants take over primary care, where there is a huge national shortage of clinicians, but it will have a harder time making inroads into hospital medicine, especially in the ICU, beyond initial assessment because critically ill patients depend on manual labor. Let’s see AI turn a patient connected to an ECMO machine and ventilator, and then bathe them.
What will AI’s impact be on programming?
One of the best use cases for AI is that it can help people code faster and more accurately. While AI can help develop robust computer programs, computer science, at its best, is a creative endeavor. AI may be able to write programs more efficiently (and maybe more elegantly) than humans, but it will likely not necessarily find solutions with the same level of innovation as a human in more complicated, unique, or novel software.
If writing with AI does not produce better writers, coding with AI will not produce better programmers. Any large computer program, especially one with essential societal impact (transportation, medicine, security) will have to also be evaluated and edited by humans. But if your exposure to writing computer programs is limited, how will you have the experience to edit software?
Further, in many ways, the Internet is stylistically boring and homogenous. Websites, social media, and apps look largely similar. There’s a navigation bar or button in the header, followed by content, and then a footer. AI will continue to make websites that look similar. But to create something innovative, like the website Pudding.cool or Neal.fun will require human input.
What will AI’s impact be on scientific research?
If a country has a robust scientific research pipeline, I think AI can be extremely beneficial. However, in the United States, where the current administration has sabotaged the research ecosystem by removing funding for grants, firing researchers, or promoting misinformation, the impact will be much more limited. AI needs data to make inferences, suggest actions, or discover patterns. If there are gaps in knowledge, then the output will be flawed.
AI will also convolute scientific research in general because scientific journals are being inundated with false, misleading, or AI-generated scientific papers. It will be significantly harder to extract real information from the noise. Academia will have to figure out a way to validate real scientific papers so that we can actually move forward.
What will AI’s impact be on transportation?
AI will improve transportation on a large scale. The U.S. government has an incredible opportunity to reinvent and reimagine the archaic system that underlies the Federal Aviation Administration. The system is already stretched to its limits. If we want to avoid aviation catastrophes or near misses, AI must be incorporated to help the limited air traffic controllers to do their jobs effectively.
Self-driving cars will continue to expand their reach and scope from beyond local delivery and taxi services to cross-country trucking. People will inevitably lose jobs as a result, but it may also lead to cheaper and faster transportation.
What will AI’s impact be on education?
I think AI is going to be detrimental to education and learning. When a B+ answer is readily available for pretty much any question, what is the point of struggling? Many students are disillusioned with higher education as they face a job market where they are competing with AI and they are burdened with crushing debt. Students just want to get through the obstacles and jump through the hoops so that they can secure their future.
While AI tutors, with seemingly infinite knowledge and patience, may be of great benefit for those who can use them correctly, this requires oversight and discipline. Most learning comes through struggle because getting to the right answer is not necessarily the reason why we study or learn a discipline. Instead, the point of learning is to improve the way we think, reason, practice, and apply information. Using the brain is like exercising a muscle. Without struggle for growth, it will atrophy.
Early signs point to AI harming how students learn. A study from MIT’s Media Lab in 2025 asked students to write SAT essays using ChatGPT, Google’s search engine, or nothing at all. They found that ChatGPT users “had the lowest brain engagement and ‘consistently underperformed at neural, linguistic, and behavioral levels.’”
What will AI’s impact be on mental health?
On an individual basis, AI will be helpful to some people for working out basic grievances and anxieties in their lives. How does one respond to an email from a boss? What are ways one can broach a sensitive issue with a daughter? What are some ideas to focus when studying for an exam?
For other mental health challenges or advanced forms of psychiatric disease, I find it hard to believe that a robot can provide the same level of comfort, assurance, and guidance that a human can. Let’s see how long a roaming robot would last in a psychiatric hospital.
On a larger scale, AI may be detrimental to mental health. There have already been reports on AI-induced psychosis. Further, the sheer loss of agency, ability, employment, and meaning due to AI will only increase the aggregate stress and anxiety felt by humanity.
What will AI’s impact be on art and creativity?
Art and creativity, in my opinion, are inherently human endeavors. A human being must ultimately evaluate whether a creation, whether made by a person or computer, is considered artistic or creative. If one AI system evaluates another AI system’s creation as artistic, can we accept it as a valid judgement? I do not think that the AI’s assessment will become mainstream without first being rated by a human.
If anything, AI will make art and creativity more generic. There will be a regression to the mean. Art will seem polished and flattened but without spirit or life. It will not create a never-before-seen narrative or reinvent painting or develop a new design for fashion. It will only reinforce current practices, perhaps in slightly different ways, but certainly not be groundbreaking.
What’s one misconception about AI that you think is worth dispelling?
AI, as we currently understand large language models, is not thinking. It is accessing a large corpus of information from which it is deriving the statistically, most likely next word or words. If the input is flawed, so will the be the output. AI is great for getting you the average answer, but it won’t lead to a innovative one unless you can provide it new data.
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 AI.”
True
“AI will have led to breakthrough treatment or cure for a major disease.”
True, especially if it involves evaluating large amounts of data like in DNA or protein for a specific answer
“AI will have played a role in a major global security event.”
True. It can be good, like helping prevent a major cybersecurity disaster, or catastrophic, like if bad actors use AI to hack into a hospital or electrical grid or a military base.
“Most Americans will be using AI chatbots at least once a day.”
True, and it won’t necessarily be fun. If you are already annoyed with holding on the phone while waiting to speak with a customer service representative, wait until you are chatting with an endlessly friendly chatbot that is immune to your growing frustration.
Most people define artificial general intelligence as a form of AI that is comparable to human intelligence. How likely is it that we will see AGI in the next 10 years?
Extremely unlikely. AI will likely be able to match or surpass human intelligence and capability in areas with extremely narrow scope, like finding patterns in large data sets or driving a car. But human intelligence does not just exist in an intellectual realm. It is also physical, artistic, and emotional. I don’t see AI playing in the NBA, cooking a novel meal, or comforting a dying patient in any meaningful way in the next ten years.
What’s a technology that had a transformational impact similar to AI?
I believe AI is part of the same technological evolution that led to the Internet, the cell phone, and social media. These technologies changed the way we interacted with data and information and are now ubiquitous. They were quickly released for mass use by profit-seeking companies and now we must contend with their positive and negative effects in society. AI is the next step in the journey. AI depends on the massive amount of information that these prior technologies aggregated. How it will affect society is unknown. But it has the potential to be more bad than good, if left unregulated.
What advice would you give a high school student about how to think about AI and prepare for the future?
This is an unpopular opinion, but learn to code! I will write about this at some point in the next couple of months, but I do not think that vibe coding is enough. Vibe coding is helpful if you want to create small programs to accomplish simple tasks, like sort or search through certain files on your computer. But even now, you would still need to know how to use a code editor, the command line interface, and some basic computer programming principles.
But to do anything meaningful or long-lasting, like develop your logical reasoning, create complex software, or become a tech entrepreneur, you will have to learn to code.
In my opinion, writing code is another level of literacy (more on this in the future). Even if you have AI write code for you, the program must still be assessed and edited. You can only be good at this if you have experience writing and struggling with code yourself. Spending a couple of hours to debug a program will teach you a lot about how your program works and how to improve your code the next time.
There will never be a time where less knowledge or understanding is better than more. That is especially true in computer science, a field that basically undergirds and supports every field and industry on Earth. The rich and powerful would rather that the masses do not have this understanding so that they can continue to be gatekeepers. NVIDIA CEO, Jensen Huang, basically advised that students not learn how to code. I think this is terrible advice. Technology evolves quickly and if any gaps or delays in staying up to date will make it exponentially harder to catch up later on.