Super God-Level Top Student

Obviously, for Sam Altman, Huaxia's artificial intelligence technology is something specific.

It's not just Dou Dou, although Dou Dou is the world's most successful AI product and is very well-known. However, as of now, ordinary people can only interact with Dou Dou online and have no access to its source code.

The situation is the same for most people, though the Qunzhi Framework can be accessed, and the approaches of the two are not identical.

For instance, ChatGPT's logic lacks causal coupling, with its core principle being to predict the most likely next sentence or word within a given context by analyzing large amounts of text data.

This learning model based on statistical patterns lacks an understanding of the real world.

At least for now, ChatGPT has not constructed an internal model of the world to understand causal relationships.

One simple example is when someone asks ChatGPT what would happen if ripe apples are not picked. The answer would most likely be correct: they fall to the ground. This doesn't mean ChatGPT understands the causality of gravity, but more likely that it has learned from the article about Newton's discovery of gravity.

This is different from the causal decoupling framework of Group Wisdom. The Qunzhi Framework is actually closer to the World Model Framework, relying on learning causal relationships. Simply put, it continuously learns from results to infer causes, thereby mastering knowledge and deepening understanding of the world.

This is also why, if someone persistently feeds ChatGPT incorrect data, it will keep making mistakes on certain questions.

Of course, this does not mean that the Qunzhi Framework is perfect.

In fact, the Qunzhi Framework can also make mistakes, but these are often due to incorrect understanding of causal relationships.

For example, there once was a joke online where someone asked the Yi connected to the Qunzhi Framework what would happen if ice cubes were put into freshly boiled hot water. Yi answered that the ice would turn into steam because the boiling water could cause the ice to evaporate.

This is a typical causal misunderstanding. After all, according to physical processes, the ice would first melt under heat and then reach boiling point to become steam.

The difference between the two is that, to have ChatGPT give the correct answer, the type of text generated in training data matters. The Qunzhi Framework just needs it to fully understand the concepts of melting and sublimation.

For such simple questions, clearly, the former has an advantage. The former only needs to repeatedly tell ChatGPT that ice must first melt into water, so it can give the correct answer next time.

But the latter involves giving various examples to make the system understand that dry ice turns directly into gas because, at Earth's standard pressure, carbon dioxide has no liquid phase stability zone, thus transitioning directly from solid to gas, but water is different...

However, the pros and cons cannot be compared so simplistically.

Because once the Qunzhi Framework deduces causality and understands the entire process, it will not make the same mistake again. But if someone persistently feeds the former erroneous data, it might still err next time...

For Sam Altman, the good news is that ChatGPT has been fed enough data that, at this stage, it might perform not much worse than the Qunzhi Framework. But the bad news is that there still exists a formidable product like Dou Dou, and there's also the threat from sponsors.

Really, if not for all the regulatory bans, Sam Altman might have succumbed too.

The confidence to emit those two chuckles ultimately stems from the fact that the opposing company is strictly regulated. Forget about investing in the opponent; if Microsoft really tries to flirt with the other side, it must first face the iron fist of capitalism, even if Microsoft has enough motivation.

The current world structure is such that at the international level, when companies reach the stature of Microsoft and Youwei, decisions can no longer be made purely from a commercial perspective, let alone considering Microsoft's public company status.

Even if Satya Nadella dared to make reckless orders, shareholders and the board could quickly teach this Indian descent CEO a lesson.

Despite the sarcastic laugh, Sam Altman clearly did not want to offend the person in front of him too harshly. Taking a deep breath, he said, "Alright, Satya, we will consider your opinions and try to find a mutually acceptable way of cooperation. But we must be clear that OpenAI's technical advancement cannot be constrained by a singular interest relationship."

Satya Nadella's face was like water, but he could only reply lukewarmly, "I hope you'll consider carefully. If things continue like this, OpenAI may very well lose one of its most important former supporters."

They parted on bad terms.

The outcome of the talk was no outcome at all, yet it wasn't fruitless. At least both parties roughly understood each other's stance. Whether a suitable solution can be found is another matter.

Of course, this conversation was confidential.

Other than the involved parties, no third person could hear it.

But neither of them noticed, perhaps unnoticeably, that when Sam Altman stood up to leave, the WiFi signal transmission light in the corner of the office blinked just a bit more frequently than usual.

...

"On behalf of the Xilin Institute of Mathematics, I warmly welcome everyone and thank Professor Qiao for giving me the honor of introducing a new accelerator theory and the related results we verified with supercomputing."