Neuralink plans to develop a product called "Blindsight" to help people who are completely blind regain their vision by direct stimulation of the visual cortex.

Finish丨Liu Jie

Editor丨Cen Feng

Last Saturday, Lex Fridman, a well-known tech blog, invited ·Elon Musk and the rest of the Neuralink team for an in-depth discussion that lasted an astonishing eight and a half hours, covering topics such as Neuralink's ability to expand the human mind, thinking about civilization, the creation of a backup for humanity by colonizing Mars, and the combination of artificial intelligence and robots.

In the conversation, Musk and Lex Fridman talked about his vision for the future of technology and AI applications, which is super technical and covers a wide range of content, and the content of the discussion is rare from depth to height.

As we all know, Musk has always advocated a "first-principles" way of thinking, and has made many disruptive innovations and breakthroughs under the guidance of this principle. In this conversation, Musk also used "first principles" to explain his way of thinking several times.

For example, when talking about Neuralink, Musk mentioned that if you want to achieve a seamless connection between the human mind and the AI system, you need to increase the communication bandwidth between humans and machines, which is also the decisive factor for him to start the Neuralink project.

Also, he suggested trying to delete a step or section, and if you didn't need to re-add at least 10% of the content after deletion, it wasn't enough. This approach can be applied to any engineering or development process, including xAI and Optimus projects.

In addition, Musk uses first-principles thinking when discussing human behavior and history, considering the fundamental motivations of human behavior, such as seeking pleasure and avoiding pain, and how these motivations affect the development and use of technology.

Due to the huge amount of information, this time we only intercepted some of the essence of the future development of AI and summarized it, but even the "provincial summary version" ended up being a long article of nearly 6,000 words.

1

The future of Neuralink

Musk said Neuralink is committed to solving basic neurological problems, such as spinal cord injury or blindness. As in the case of their first two patients, the first priority was to address basic neuronal damage to the spinal cord, neck, or the brain itself.

Musk has outlined a rough roadmap for the future development of Neuralink's technology plan, starting with a basic and relatively simple solution to neuronal damage, and then slowly expanding to treat schizophrenia, epilepsy, and even improve memory, just like the upgrade of the technology tree.

Neuralink works by implanting electrodes that read brain activity and convert it into electrical signals that can be used to control external devices or communicate with AI systems.

Musk plans to significantly increase the number of electrodes in Neuralink in the future, significantly improving its signal processing capabilities. This will allow the transfer rate to increase dramatically from current levels to 100 bits per second, or 1000 bits per second. Even five years later, it could reach 1 megabit, faster than anyone can communicate by typing or speaking.

Musk shared his personal experience of improving the speed of communication, such as improving the efficiency of receiving information by increasing the speed of playback.

Musk believes that in real conversations, in order to facilitate communication, people need to compress the huge content into a few syllables and say it again, and hope that the other party can decompress it into a conceptual structure close to the speaker's mind. As a result, a lot of information is compressed and cannot be fully expressed during communication, and the communication efficiency is very low. He called it a low effective bit rate.

With the increase in the number of electrodes, people may be able to achieve direct communication of consciousness through extremely fast communication speeds, avoiding the large loss of concepts in the process of expression and interpretation of language and writing.

With the development of technology, Neuralink can improve the resolution even beyond the human eye by adjusting the electric field between neurons. In the future, people will be able to see different wavelengths of light, like Geordi La Forge in Star Trek. People can see ultraviolet, infrared, eagle vision, anything they want to see.

2

Application prospects

Musk said that the application prospects of Neuralink are very broad.

At present, Neuralink's main research and development direction is focused on the medical side, and the initial target is to solve basic neurological injury problems, such as helping patients with spinal cord injuries or neck injuries to restore their ability to communicate.

In addition, Neuralink plans to develop a product called "Blindsight" to help people who are completely blind regain their vision by directly stimulating the visual cortex.

Musk said that Neuralink is not only designed to restore basic human functions, but also hopes to provide those with neuronal damage with communication data rates that exceed normal humans.

Neuralink is expected to provide enhanced visual abilities by stimulating the visual cortex to restore vision, and possibly even beyond natural vision.

Of course, Neuralink's application is not limited to the medical part, it covers everything from medical applications to non-medical applications.

Neuralink's future abilities may include controlling the Optimus robot, improving or enhancing memory, treating schizophrenia, and more.

3

Convergence with artificial intelligence

Musk believes that human beings are the source of will or purpose, and Neuralink can enhance the expression and realization of this will. With Neuralink, human intelligence can be combined with the computing power of AI to expand our ability to solve problems and innovate.

With the development of technology, our computing devices such as smartphones and computers have become an extension of our computing power, participating in our daily lives, including the process of finding a sexual partner. Musk went on to make an interesting point that with the convergence of artificial intelligence (AGI) with humans, this computing power will be expanded, potentially further enhancing our ability to pursue pleasure. Therefore, AI may be a limbic system in a benign situation just to please humans, which may mean that AI will become one of the tools for humans to pursue sexual pleasure.

Musk said that Neuralink aims to improve the symbiotic relationship between humans and AI by increasing the communication bandwidth between the human brain and the machine. The long-term goal is to achieve a seamless connection between the human mind and the AI system.

This vision can only be achieved by increasing the communication bandwidth.

If AI communicates at a speed of 10,000 bits per second, while humans communicate at less than 1 bit per second, interaction with AI will become very slow and inefficient. Musk said it was like talking to a tree. One needs to take into account that AI will get bored by waiting for a human to spit out a few words.

Musk said in the discussion that by increasing the rate at which humans output data, we can more effectively integrate with AI systems, thereby improving the synergy between our collective will and AI. If a computer can execute trillions of instructions per second, a second that passes means it can do a trillion things. Once the bitrate is very high, the human experience can change in unimaginable ways.

4

xAI

Musk mentioned that xAI's mission is to understand the universe. xAI aims to reach a level beyond human intelligence and explore the nature of the universe.

Musk believes that to win the game in the AI market, the most powerful training computing power is needed, and the training computing power of AI must improve faster than others.

Musk also revealed that he hopes to release Grok 3 by the end of this year, and is confident that Grok 3 will be made into the world's best large language model and the best AI system.

For Musk, the training and computing of AI models is like the horsepower of an engine, and he must be the best at this. The key lies in how to use training computing resources efficiently, and how to efficiently perform AI inference and application. This part relies heavily on human talent and the unique data obtained.

Musk has his own views on the scraped training data.

Musk believes that most of the leading AI companies have already scraped all Twitter data, so the useful part of the future AI training race is to scrape Twitter data in real time. This is difficult for other companies, but Grok has a real-time advantage in this regard.

Master is confident in AI training data scraping. He said the biggest source of data in the future could come from Optimus robots. Tesla already has millions of vehicles on the road, and there may end up being hundreds of millions, as well as hundreds of millions or even billions of Optimus robots that learn vast amounts of information from the real world. Eventually, Optimus will likely become the largest source of real-world data, as the scale of the real world matches the scale of the real world itself.

5

Optimus humanoid robot

Musk compared the production of cars to the production of the Optimus humanoid robot, believing that the two are similar.

He estimates that there are about 100 million cars in demand worldwide each year, and about 2 billion cars in use worldwide. Since the lifespan of the vehicle is about 20 years, it is possible to produce 100 million vehicles per year in a stable state, maintaining a fleet size of about 2 billion vehicles. For more practical humanoid robots, he guesses that the annual production of humanoid robots could exceed 1 billion units.

Musk also elaborated on the difficulties of mass-producing Optimus, especially the hand design. The "hand" here refers not only to the palm, but to the entire forearm that extends forward from the elbow. In the specific design of Optimus, structures such as tendons on the human arm must be included in order to achieve all the functions that the human palm can do.

Musk said that in the initial attempt, they tried to put the actuator on the hand and ended up with huge palms that looked strange. And these palms don't actually have enough freedom or strength. So they realized that the actuator needed to be placed on the forearm, just like a human being, and the fingers needed to be operated through a narrow passage.

At the moment, Optimus still has difficulty walking in the park, but he is already able to walk on various terrains and pick up various foreign objects.

6

Technical and ethical considerations

Musk affirmed that the team will first focus on addressing security issues and risk mitigation before considering the application of Neuralink to the non-disabled population. It is only after the safety and reliability of the technology has been ensured that it will be considered for its use to enhance the capabilities of normal humans.

Musk believes that the convergence of humans and AI through Neuralink may be a way to ensure AI security, as it can improve the efficiency of human-AI interaction.

Musk drew an analogy between the human brain and a supercomputer center. He believes that most of the human brain tissue is like a cable. The gray matter of the brain is the computer, and the white matter is the cable.

In this metaphor, Lex Fridman proposes the idea that superintelligent systems will one day be built in giant brains like supercomputer centers.

Musk agrees and believes that super-AI must be smarter than all humans combined, and that to ensure that super-AI does not harm humans, it is necessary to stick to the truth and avoid forcing AI to lie or training them to lie.

7

The Fermi Paradox, AI and the Human Civilization Filter

In the conversation, Musk delved into several key issues in the development of human civilization, especially emphasizing that artificial intelligence could become a filter for civilization. He believes that emotional state directly affects the quality of individual decision-making, and the quality of individual decision-making is closely related to the direction of development and survival of civilization. Musk's quest for curiosity-driven understanding to delve deeper into the mysteries of the universe is reflected in several of his projects, including Grok and SpaceX.

Exploring extraterrestrial life and the Fermi paradox, Musk made the point that intelligent life can be extremely rare, because being a multi-planetary species is a huge challenge, which can be seen as a kind of "filter". He argues that despite the proximity of the Moon, Mars is better suited to build self-sufficient human cities because it is more environmentally adaptable and more resilient to potential disasters. Musk believes that the difficulty of reaching Mars is a testament to its environmental adaptability, and that being a multi-planetary species is the key to the survival and reproduction of civilizations.

In addition, Musk and Lex Fridman discussed artificial general intelligence (AGI) as another possible filter. Musk quoted Jeff · Hinton, who believes that the risk of AI potentially destroying humanity is about 10 to 20 percent, while he believes it is actually higher. This shows that mitigating the risks of AI is crucial. He stressed that becoming a multi-planetary species would not only reduce this risk, but would also be an effective way to maintain population numbers and avoid the collapse of civilization.

Musk also mentioned the issue of population, noting that demographic collapse is a recurring problem in the history of civilization. He warned that if the birth rate continues to fall, the population will be reduced to zero, so encouraging childbearing will be essential to delay this collapse. Both he and Friedman believe that inspiring people to create and reproduce is not only for the continuation of civilization, but also to explore the possibilities of artificial intelligence and robotics in the future.

Overall, Musk's view is that whether it's the exploration of extraterrestrial life, the cautious optimism about the development of artificial intelligence, or the pursuit of humanity becoming a multi-planetary species, all of these efforts are trying to ensure that human civilization can continue to evolve and eventually transcend existing boundaries through various filters.

8

Team member discussions

Not only Musk, but other members of the Neuralink team also shared a lot of in-depth and interesting content.

DJ Seo

DJ Seo, co-founder of Neuralink, shares his own journey to become interested in the human brain.

DJ Seo's grandparents suffered from a very severe form of Alzheimer's disease, and their state of illness, loss of memory and thoughts, made DJ Seo begin to notice the power of the brain.

DJ Seo arrived in United States as a teenager and didn't speak English, had difficulty communicating with his peers, and developed a keen interest in science fiction during long solitary periods. Films like Ender's Game, Neuromancer and Snow Crash, as well as The Matrix, have all profoundly influenced DJ Seo's view of technology and life.

As an undergraduate, DJ SEO was fascinated by physics, building physics, and especially physical things with some kind of intelligence. The experience of studying MEMS, or microelectromechanical systems, to build these micro-nanostructures for temperature sensing led DJ Seo to apply for a Ph.D. program at the University of California, Berkeley, and eventually the Berkeley Wireless Research Center.

At the graduate level, DJ Seo collaborated with Prof. Michel Maharbiz on a Smart Band-Aid project, which was the first time that DJ Seo had designed an electrical solution and applied it explicitly to biology.

After sharing the history of BCI and the concepts of biophysics related to BCI, DJ Seo gave a detailed introduction to how Neuralink works, and also shared the technical challenges encountered in the development of Neuralink and the subsequent team innovations.

In terms of future technology upgrades, DJ Seo believes that if you want to upgrade a device that has already been implanted in the brain, you have to cut the wire or pull out the wire, and you can reinsert a different wire with an updated implant pack.

DJ Seo shared that the Neuralink team is working on different needle designs as well as loop engagement types to minimize scarring by researching and upgrading imaging techniques so that implants can be easily removed over time.

DJ Seo said the team is also working in another direction, directly changing the structure of the implant instead of using the current structure, which is a monolithic implant plus a bonded line.

However, the priority was to expand the number of threads so that the team could record more and more neurons. DJ SEO plans to reach 3,000 or even 6,000 channels by the end of this year, and hopefully reach 16,000 by the end of next year.

DJ Seo hopes that in the future, by expanding more Neuralink devices, not only can help people with motor and visual impairments or mental illnesses solve their problems, but also apply it to everyone to help them achieve a different life.

Matthew MacDougall

Matthew MacDougall, Chief Neurosurgeon at Neuralink, also shared his passion for the human brain.

Matthew MacDougall, who loved to think and observe since childhood, naturally developed a keen interest in the most important part of humanity, the brain. Wanting to use his knowledge to make a difference in real life, Matthew MacDougall switched from undergraduate neuroscience to neuromedicine as a Ph.D., eventually deciding to become a neurosurgeon.

Matthew MacDougall also shared the full details of the process of implanting the N1 chip in Neuralink, including the procedure, the technology used, and more.

Contrary to what most people think, Neuralink's implant surgery is a relatively simple and straightforward process that involves making an incavity in the skin on the top of the head, identifying the area, and then a robot avoiding the blood vessels, inserting electrodes precisely into the cerebral cortex, and finally suturing them by humans.

The whole process only takes a few hours. It has extremely low risks compared to general neurosurgery involving the brain and is much smaller than many conventional tumor or aneurysm surgeries.

Matthew MacDougall sees the implantation of Neuralink as a huge leap forward.

Matthew MacDougall shared a recent study by Nicholas · Schiff, Jonathan · Baker and others, who gathered a group of patients with moderate traumatic brain injury and implanted electrodes in the deep nucleus of the brain (called the central median nucleus or near the central median nucleus), which can significantly improve people's attention and concentration when they apply a small amount of electrical current to this part of the brain.

That's one of the reasons why Matthew MacDougall joined the Neuralink team. Neuralink is able to improve people's ability to perform tasks. From a purely utilitarian point of view, this allows people to once again take care of themselves and their families financially, allowing those who are completely dependent or may even need significant resources of care to be completely independent, take care of themselves, and give back to the community.

As a neurosurgeon, Matthew MacDougall has an intuitive understanding of life and death, and he believes that although death is inevitable, human suffering can be alleviated through technology.

Neuralink technology has such potential.

In the future, Neuralink will be able to help those who are suffering from severe pain by providing the ability to control devices such as computers, improving their quality of life. For example, helping a quadriplegic person to regain their freedom and independence.

Bliss Chapman

Bliss Chapman, Head of Brain-Computer Interface Software at Neuralink, first shared his motivation for joining Neuralink, hoping to help those with spinal cord injury or ALS through Neuralink.

Through tiny devices implanted in the brain, Neuralink is able to read and decode neural signals, enabling direct control of external devices. Those who have lost their physical functions due to illness or injury can also regain their ability to communicate and control with the outside world.

The acquisition and processing of neural signals is at the heart of Neuralink's technology. Neuralink's N1 implant has 1,024 electrodes to collect nerve signals from the brain. Once the signal is acquired, it needs to be pre-processed and decoded before being converted into actionable commands. Through machine learning algorithms, Neuralink is able to identify the user's intentions and translate those intentions into actual control commands, such as the interaction of a mobile computer cursor or other digital device.

According to Bliss Chapman, the most critical challenge in this process is to ensure the accuracy and real-time nature of the signal, while reducing latency to provide a smooth user experience.

Bliss Chapman also shared a lot of experience in application user experience (UX) design. He believes that UX design is crucial in BCI technology, influencing how users interact with the technology and how effectively they can use it.

Since Bliss Chapman's initial goal was to help the paralyzed, he and his team initially designed the app to control mouse and keyboard input, rather than the entire process of daily life and communication, considering the current practicality.

The Neuralink app actually converts the wireless brain data stream from the implant into computer control, which is achieved by establishing a mapping from brain activity to HID input to actual hardware.

Because it was difficult to actually observe what the paralyzed person wanted or was trying to do, it was difficult for Bliss Chapman to make sure that the correct mapping was established between the given intended action instructions and the nerve impulses, so Bliss Chapman spent a lot of time researching how to improve the accuracy of the calibration process and interface.

Bliss Chapman particularly emphasized the importance of designing an intuitive, easy-to-use interface that helps improve user learning efficiency and user satisfaction.

Bliss Chapman is very excited about clinical trial operations for the next year or so, and he and his team have chosen to focus their time and resources on user experienceAspect. Bliss Chapman was curious about how the future would handle all the nuances of the user experience, i.e. not being able to feel the different keys under your fingertips, but still need to be able to adjust all the keys simultaneously to achieve the desired effect. And, in the absence of a suitable feedback loop, how do you allow the user to intuitively control the high-dimensional control interface without actually feeling it?

Bliss Chapman believes that as technology advances in the future, Neuralink will continue to expand its range of applications, including helping people with visual and speech impairments, so that more people can regain their independence and freedom.

This article is from Xinzhi self-media and does not represent the views and positions of Business Xinzhi.If there is any suspicion of infringement, please contact the administrator of the Business News Platform.Contact: system@shangyexinzhi.com