Why Musk believes that the future AI will become the "evil ruler" of mankind

Last Sunday, a very unusual Dota2 game tournament attracted a lot of attention. Dota is a sophisticated real-time strategy game and one of the most popular e-sports projects in the world. One of the teams in the competition has five members, Blitz, Cap, Fogged, Merlini and MoonMeander. Each ranks above 99.95. It can be said that they are one of the best Dota2 players in the world. However, their opponents defeated them in three consecutive games and won the game.A Dota2 game takes about 45 minutes, but the two games ended with 14 points and 21 points.


Their opponent is a team of five neural networks developed by Elon Musk's OpenAI company, collectively known as OpenAI 5. Before the game on Sunday, the neural network had a 180-year Dota competition every day, and gradually approached the game. Its creators chose Dota as the goal of OpenAI 5 training to simulate the incredible changes and complex nature of the real world. Dota is a complex game. If artificial intelligence can handle and interact with the world, not just learning to draw GPS maps or playing chess, Dota is a good way to open video games.

Although this is an impressive technical achievement, Musk emphasized through tweets that this is only the basic technology for the future of artificial intelligence.

The neural interface referred to by Musk was developed by his other startup, Neuralink. The goal of Neuralink is to establish a brain-computer interface (BMI) between human thinking and generalized artificial intelligence. Essentially, an artificial intelligence can think in general, rather than thinking about how best to win a Dota game. By developing an interface between the brain and artificial intelligence. Neuralink hopes to enhance human capabilities and prevent Musk's often mentioned concerns about the future: an immortal, evil, dictatorial artificial intelligence.

In an interview with Y Combinator, Musk explained that he was worried that only one company or a small group of people could control the future of artificial intelligence technology.

"I think this is very dangerous. It can also be stolen by bad guys, such as an evil dictator or country that can send their intelligence agencies to steal it and gain control. I think if you have a very strong Artificial intelligence, I can think of the best alternative, that is, we have realized the democratization of artificial intelligence technology, maybe others can come up with a better method or better results. But I think this will be a very bad A stable situation.

To achieve this democratization, Musk founded OpenAI, whose technology is completely open to the public and used to create OpenAI 5 for "Dota."

His second method of democratizing artificial intelligence is through neural transmission. By connecting human thinking with cloud-based artificial intelligence. This theory holds that we will have the ability to compete with any concentrated, vicious artificial intelligence.



Elon Musk believes that the solution to reduce risk is to be able to implement high-bandwidth interfaces with artificial intelligence. He believes that if we can think with artificial intelligence, it can make artificial intelligence the third layer of our brain, where we can build artificial intelligence for us.

“It sounds creepy, but it has a lot of meaning. If we are all artificial intelligence, not everyone can get control of all the AI ​​systems in the world for monopoly. Maybe someone uses it to do bad things because they are facing a few One billion people of artificial intelligence. It brings us security in a more strange way, even though it gives us more power. It's like you don't want a superman on earth, but if you have a billion Superman, then everything is ok because they check each other."

This kind of human-computer symbiosis may not be as crazy as it sounds. BMI is already present in computer chips implanted in the brain, providing targeted electrical impulses to help treat neurological diseases such as Parkinson's disease or epilepsy. In a recent study, three monkeys were implanted with BMIs that shared their sensory and motor information and sent signals to the robotic arm. Monkeys are able to create a shared neural network and move the robotic arm into three dimensions.



In addition, the interface developed by Neuralink has existed in an original form. “Our email function, our computers, mobile phones, and apps all have a third number of self. We are actually Superman, but we are between the cerebral cortex and our own third-level digital form,” said Musk. The interface is extremely limited. Although our device may give us more information and more power, we still need to use our fingertips to type, our eyes receive letter symbols in our brains Additional computing power to decode these symbols, etc., will greatly reduce our speed."

Like everything that Musk has done, it's a very bold and forward-looking job, but it may not take into account the full impact of BMI, linking human thinking to cloud-based AI. Some researchers believe that artificial intelligence is a biological artificial intelligence, and human symbiosis will help prevent singularities, that is, artificial intelligence becomes as smart as human beings, moving through the essence of movement. If humans can enhance themselves by connecting to artificial intelligence, when artificial intelligence is smarter than humans, it will prevent such things from happening.

However, there are other concerns that are directly related to the specific practices that Musk has taken here. The Morningside Group is an organization of neuroscientists, neurologists, ethicists, and machine intelligence engineers that describes several ethical issues about the intimacy of the human brain and cloud-based artificial intelligence, first and foremost privacy and agree. Consider all the allegations of collecting data around Facebook. Even for future humans, the symbiote system is open source and decentralized, and its personal data will be classified as a cloud computing platform. In this case, how do we change and control personal data?

Agents and identities are another issue that must be considered. This is not to say that those connected to the cloud may know your identity. But you may have completely lost your self-awareness. If everyone can interact with cloud-based intelligence, then personal intelligence may no longer mean anything.

Another question is how this promotion can play its part in society. Of particular concern is artificial intelligence, enhanced humans can be used in wars, and new arms races can begin.

More importantly, the prejudice inherent in our society tends to be adopted in the technology we create. Google showed women a lower-paying job advertisement, and the algorithm used by US law enforcement overwhelmingly predicted that black criminals would commit another crime than white offenders accused of the same crime. Artificial intelligence may be able to objectively avoid these biases, but should not be considered as such.

The problem with all of this is that there is no objective way to understand how artificial intelligence will affect our society. This is a too radical change that we cannot understand. At the same time these changes will happen and we must be prepared. Through OpenAI and Neuralink, Musk seems to avoid the biggest risks by trying to determine the development of artificial intelligence from the beginning.

Follow Me
Link:Tenco

                                                          ——END——

评论

此博客中的热门博文

RoboMaster Ends: Very Cool Robot Design Competition

The sixth generation of Xiao Bing is online. Why did Microsoft spend four years exploring emotional AI?

These raspberry pi products let you feel its charm and fun.