Altman reveals: GPT-5 will reconstruct everything, and one person can rival five teams.
The release of GPT-5 may not just be a noisy technological show but a sober list for the future.
Stronger reasoning, more natural multimodality, and smarter collaboration methods are enabling AI to evolve from a "tool" to a "companion."
In a podcast, OpenAI CEO Altman mentioned the redistribution of computing power and capital, the true establishment of startup moats, and even the repositioning of education and family under the impact of AI.
This is not just a flashy declaration but an actionable list for the future.
The host of this podcast, Nikhil Kamath, is also remarkable:
A high - school dropout, he now has a net worth of $2.6 billion and was included in the Forbes list of India's 100 richest people in 2024 along with his brother Nitin.
Harsh Dwivedi, co - founder of MedialApp with over 400,000 users and an honoree on the Forbes 30 Under 30 list, summarized the core content of the podcast:
Altman: Three Suggestions for the AI Era
After experiencing GPT - 5, Altman finds it "extremely painful" to go back to the old models because in almost every aspect, the old models perform much worse.
GPT - 5 is an integrated model, which means users don't need to select a model, and it can automatically adapt to tasks.
GPT - 5 has significantly improved capabilities in terms of fluency, depth, reliability, and multi - step task handling.
Altman said bluntly that compared to GPT - 4, GPT - 5 feels like a doctor in your pocket that you can turn to at any time.
In short, Altman believes that GPT - 5 represents a greater leap, and AI is shaping the future:
With AI tools to boost your abilities, a 25 - year - old can achieve more than any of their peers in history.
Behind all this, the most crucial thing is to have an "AI - native" mindset, that is, to think about how to solve everything with AI tools.
Truly mastering AI and thinking with AI tools will be the greatest leverage of this era. Altman believes it is as important as learning programming when he was young.
For this generation of young people, the most important current skill is to master AI tools proficiently.
Whether you want to start a business, become a programmer, or enter other industries, large models enable an individual to use these AI tools. As long as you have a good idea, you will have the ability to put it into practice.
Altman believes that GPT - 5 makes individual entrepreneurship possible:
Use GPT - 5 to write product software more efficiently, handle after - sales services, formulate marketing and communication plans, review legal documents, etc.
All these things used to require many people and a great deal of professional knowledge to accomplish, but now you can use GPT - 5 to do all of this. It's really amazing.
What exactly should one learn in the AI era?
Altman believes that the most important and specific thing is to learn to truly master the use of new AI tools.
Learning itself is valuable, and learning how to learn is a meta - skill that will benefit people for a lifetime, regardless of what you major in. However, being proficient in using tools is very important.
In today's rapidly changing world, the second general skill to learn is to learn how to adapt and stay resilient.
If you choose to start a business, the most important thing might be to learn to understand what people want.
At the age of 19, Altman dropped out of Stanford University to start a business.
In 2011, he joined the startup accelerator Y Combinator and served as its president from 2014 to 2019.
Before the establishment of OpenAI, a saying by Paul Graham, the founder of Y Combinator, deeply impressed Altman, which is
Make something people want.
This may seem easy, but many entrepreneurs, despite their hard work, still can't find the key.
So, the question is: How can one learn to adapt, change, and learn AI tools more quickly? Is there a path?
Altman believes that "with large models, you can now quickly test and correct mistakes."
Moreover, GPT - 5 is quite good at quickly creating small - scale software. When you encounter small problems in life, you can use it to quickly create a software to solve the problem.
However, this is not the end. In this rapidly changing world, you need to keep an open mind. As Altman often says, "No one knows what will happen next. Thinking you're smart and thinking you know what users really want will only lead to bad decisions."
Having an open mind, curiosity, and being willing to adapt and change your mind based on new data is very important, according to Altman.
Altman's Self - Awareness
In 2009, Paul Graham, the founder of Y Combinator, mentioned five founders to watch out for: Steve Jobs, TJ Rodgers, Larry and Sergey, Paul Buchheit, and Altman.
At that time, Altman was only 19 and just starting out, while Steve Jobs, Larry and Sergey, etc., were already well - known. Why could Altman be compared with these people at that time? What innate skills did Altman have that made him so different?
Altman was like an innocent young man at that time and didn't really know what he was doing. This is not false modesty.
He thinks that "now, some of the things he does are indeed impressive. But at 19, his self - awareness was very uncertain and he was very inconspicuous."
If the world of tomorrow becomes an AI kingdom, Nikhil Kamath believes that Altman will definitely be the "monarch" in a sense.
Niccolò Machiavelli, an Italian Renaissance writer, wrote in The Prince:
A prince may be regarded as kind, loyal, humane, straightforward, and pious, but it is most important to seem to have these qualities.
In recent interviews, Altman has repeatedly said that he is not intimidating and has used words like "pious" and "humble" many times.
The host wondered, "Is this humble image suitable for the world we are living in now or the world you are entering?"
Altman is not sure which world it suits.
When he was 19, he assumed that those running the tech giants really had everything figured out. The room was full of adults with clear divisions of labor, the company was running well, there wasn't much drama, and everything was under the control of adults.
Now, it's Altman's turn to be the adult in the room. He thinks no one has a plan, and no one really makes everything run smoothly. Everyone, or at least he, is groping his way forward.
At least, he just keeps moving forward step by step and focuses on his work.
He believes that "everyone is groping their way as they go. Everyone is learning on the job."
This is not him pretending to be humble. He thinks the world works this way.
One of his often - repeated sayings is: No one knows what will happen next.
Having an open mind, curiosity, and the willingness to adapt to new data and change your mind is extremely important, according to Altman.
Just like how many times OpenAI thought it knew something but was proven wrong by reality. One of OpenAI's advantages is that it can adjust in time when this happens.
Maybe there are other ways to succeed.
For example, bluffing can be effective, but among the founders Altman has observed closely, the best ones are more of the type that learn and adapt quickly.
A New Method for Evaluating the Intelligence Level of Large Models
In the following interview, Altman proposed a new indicator for evaluating the intelligence level of large models, which is different from saying that large models have the level of a doctor during the product launch. The new indicator Altman proposed emphasizes the time dimension more.
Altman said that for tasks that humans only need a few seconds to a few minutes to complete, such as pattern recognition and recalling knowledge, most current large models can perform these tasks quite well.
However, when it comes to asking questions or handling problems that also take humans a long time to deal with, large models are definitely not close to human performance levels.
For example, in mathematics, large language models (LLMs) have advanced from the level of minutes - long intelligence to hours - long intelligence.
A few years ago, large models could solve problems that human experts took a few minutes to complete.
Recently, large models like GPT - 5 have achieved a gold - medal level in the International Mathematical Olympiad.
Each problem in the competition takes about an hour and a half to complete.
Therefore, we can say that the time range within which large models can simulate human thinking has increased from a few minutes to an hour and a half.
Proving an important new mathematical theorem may take a thousand hours. When will AI be able to solve problems that human experts need a thousand hours to solve?
You can make a prediction. Anyway, today's large models are completely unable to do this. So, this is another dimension where AI falls short.
However, there's no need to be overly worried about AI replacing humans.
Altman said, "Although AI may surpass humans intellectually, humans' longing for real interpersonal relationships is deeply ingrained. We will continue to care about real people."
For example, a smarter AI podcast host is probably not going to be more popular than a real human host.
In a world with infinite AI - generated content, the value of being a real person will increase.
Real people do make mistakes. So, to put it jokingly:
In the future, becoming more stupid will be more novel than becoming smarter.
Maybe this is part of our connection with real people.