Ultraman reveals a shocking prophecy: In 2035, GPT - 8 will cure cancer, and humans will fight a third world war over the explosion of computing power.

As soon as GPT-5 was launched, the entire internet exploded! Netizens roared, "Give me back GPT-4o," and Altman instantly backed down. In his interview, he kept dropping bombshells: By 2035, GPT-8 might be able to cure cancer, and by 2050, the global competition for AGI computing power among all of humanity could directly trigger World War III!

After the release of GPT-5, it unsurprisingly caused another huge stir across the internet.

The cries of "Give me back GPT-4o" swept across the internet like a tsunami!

Even Karpathy agreed that GPT-5 was indeed a bit disappointing.

Finally, amidst the overwhelming scolding, Altman quickly admitted his mistake and said that he would immediately bring back GPT-4o!

Fortunately, for those who miss GPT-4o, as long as you enter the settings and turn on "Show legacy models," you can see the previous models in the drop-down menu.

Although the release of GPT-5 was particularly bumpy, all the employees at OpenAI were beaming with joy.

Because Altman has just decided to give huge bonuses to 1,000 employees in the technology research and engineering teams (about one-third of the company's total staff), with amounts ranging from hundreds of thousands of dollars to millions of dollars!

As for the specific scale of the bonuses, it will depend on the employees' performance, positions, and seniority, and they can be received in cash or stock.

Just after the release of GPT-5 yesterday, Altman immediately accepted an interview with well - known YouTube anchor Cleo Abram.

A few years ago, GPT-5 seemed as distant as a science - fiction novel, but now, we are in the most dangerous global competition in history.

Most people have never seen such a huge technological change in their lifetimes. With hundreds of billions of dollars invested and an incredible amount of human effort... this moment is of great significance.

In this interview, Altman faced many of the most controversial issues currently.

This time, he also dropped a lot of shocking statements:

In 2027, AI will make a major scientific breakthrough.

By 2035, GPT-8 is very likely to cure cancer.

In the industrial revolution explosion in 2050, humans may fight World War III over the scramble for AGI computing power.

However, in his opinion, the probability of AI destroying humanity is 1%.

Moreover, he also believes that today's fresh graduates are the luckiest generation in history. Because there has never been an era where a company founded by one person can be worth over a billion dollars.

The Era of GPT-8: AI Can Cure Cancer

Altman said that he is very confident in AI healthcare in 2035.

Many people have experienced this: They have a life - threatening illness that all doctors can't diagnose, but when they input their symptoms and blood test data into ChatGPT, it can tell users what rare disease it is.

Following the instructions to take medicine, they actually recovered.

Now, GPT-5 has even better performance in health queries. Its answers are more accurate, with fewer hallucinations, and it is more likely to tell you the real cause of the illness and what to do.

Altman said that he believes that by 2035, AI will cure or at least treat many diseases that currently plague humanity.

For example, in the era of GPT-8, we can ask it to "cure a certain type of cancer."

It will first read all existing research and data and come up with some treatment ideas. Then it will tell us, "I need you to find an experimenter to conduct these nine experiments and tell me the results."

After two months of cell culture, when the experimenter sends the results back to GPT-8, it may say, "Well, there's an unexpected discovery. I need to conduct one more experiment."

Then it will tell you, "Synthesize this molecule and test it on mice." If it works, then conduct human trials.

Finally, it will say, "Okay, here's the process for submitting it to the FDA."

For anyone who has lost a loved one to cancer, this is their dream come true.

In 2050, a Ten - Fold Industrial Revolution Explosion: Who Will Be Hurt?

The host asked: If by 2050, we experience a social change like an industrial revolution, but ten times larger in scale and ten times faster in speed, who will be hurt in the process?

Altman said that we are in uncharted waters. This transition period may be very fast, but society won't change as fast as technology.

During this process, many jobs will disappear, many jobs will change significantly, and brand - new positions will also emerge.

In another sense, we simply have no idea how extreme the speed and scope of this change will be.

So, we need a remarkable sense of humility to openly consider new solutions that "can't enter the scope of public discussion."

That is to say, we may need a brand - new social contract. For example, we need to rethink how to share the most important resources in the future.

In Altman's view, the best way is to make AI computing power as abundant and cheap as possible, so much so that it can't be used up. If we can't achieve this, humans may really go to war over computing power in the future.

So the key question is - how to distribute the right to access AGI computing power.

Maybe one day, the "words output" by OpenAI's system every day will exceed that of all of humanity. Now, people already send billions of messages to ChatGPT every day and make decisions based on them.

And a small change made by a researcher to the "way of speaking/personality" of the model can affect a huge number of conversations. This is an extremely great power.

Everything is happening so fast that we must seriously think about what it means to change the model's personality on such a large scale.

What's the Difference Between GPT-5 and GPT-4?

The host asked a sharp question: What can GPT-5 do that GPT-4 can't?

The reason for this question is that not long ago, Altman said, "GPT-4 is the last stupid model we'll use in our lifetime." However, it has already scored higher than 90% of humans in the SAT, LSAT, and GRE.

In response, Altman answered like this: AI systems can do these amazing things, but they can't replicate human abilities in many aspects. Exams like the SAT are actually very limited.

What Excites Altman Most About GPT-5

Altman said that the most exciting thing about GPT-5 for him is that "I can almost ask it any difficult scientific or technical question and get a pretty good answer."

For example, when he was in the eighth grade of junior high school, he had a TI - 83 graphing calculator. After spending a long time, he made a Snake game, which immediately became popular in the whole school.

However, when he used an early version of GPT-5, it made the game in 7 seconds... Then, whether it was adding a new function or changing the interface, it could complete it instantly.

Altman used to worry that this would make kids miss the arduous programming of the "Stone Age."

But now he is completely excited for them: The ability to create software almost instantaneously on demand is one of the most iconic features of the GPT-5 era, which was not available in the GPT-4 era.

The host also asked: You've been using GPT-5 for some time. What task impressed you the most?

Altman said it was programming tasks. "The programming ability of GPT-5 makes me feel that it can almost do anything."

Now, it just can't directly act in the physical world, but it can make the computer complete extremely complex tasks, making software itself a very powerful "control lever."

In 2027, AI Will Have a Major Scientific Breakthrough

The next question is from Stripe CEO Patrick Collison: What comes after GPT-5? When will LLM make a major scientific discovery in the next few years?

Altman said that he bets that by the end of 2027, AI will definitely make an important scientific discovery recognized by most people.

Currently, the only thing lacking is the cognitive ability of the model.

For example, recently, OpenAI's model won a gold medal in the IMO. There are a total of 6 questions in the IMO, which need to be completed in 9 hours.

However, to prove a major new mathematical theorem, it may require top - notch talents to invest 1,000 hours. According to the current growth curve, the model is very likely to reach this level in 2027. The key lies in continuing to expand the model scale.

In 1899, Could AI Discover the General Theory of Relativity?

Next, the host asked a very interesting question.

Suppose we go back to 1899 and give all the physics knowledge at that time to AI, and let it continue to deduce on this basis. When will it propose the general theory of relativity?

If no more physics data is provided, relying only on the existing data, can super - intelligence solve high - energy physics problems? Or do we have to build new particle accelerators?

Altman answered that he guessed that just thinking hard based on the existing data is not enough. We must build new equipment and conduct new experiments. Major breakthroughs depend on these because the real - world process is slow and complex.

In short, if divided by the time scale, today's AI has already surpassed humans in one - minute tasks, but it still has a long way to go to reach the "one - thousand - hour task."

Jensen Huang's Question: What Is the Truth?

The next question is from NVIDIA CEO Jensen Huang:

Facts are What is, and the truth is What it means. The former is objective, and the latter is subjective, depending on perspective, culture, values, beliefs, and background.

An AI can learn and understand facts, but how can it know what the truth is for every person and every background in the world?

Altman said that the fluency of AI in adapting to different cultural contexts and understanding individual differences often surprises him.