Neuralink

From SI410
Jump to: navigation, search
Neuralink Logo - open source image

Neuralink is a company founded by Elon Musk that aims to develop implantable brain-machine interfaces and has given rise to a host of ethical debates. The company's business model centers itself around brain machine interfaces (BMIs), with its first product being called the Link. [1] The Link will be a communication center for the 1024 threads that will be embedded in the user's brain. The company advertises the LINK as a conduit that will provide a seamless connection between the minds of its patients and electronic devices such as computers, phones, etc. The company's products are currently being market towards the physically and psychologically disabled, with hopes that the LINK can greatly increase their quality of life. The company hopes to expand its products and target the general market so that in the future, BMIs will be the norm rather than the exception. As of now the company has not tested its product in human patients but it has shown a proof of concept device in monkeys, which allowed them to play rudimentary video games such as Pong with their minds. Although its contribution towards the fields of neuroscience and engineering have garnered much praise, an equal amount of attention that has been directed toward the novelty of the company and the ethics surrounding its product.

History

Neuralink was originally founded in 2016, and publicly introduced in 2017 but did not receive significant attention until its CEO, Elon Musk, announced its existence during a livestream presentation. [2] Before the company could be founded however, its founders had to purchase its name, NeuraLink, fromm a pair of BMI researchers who had copyrighted the name back in 2013. [3] The pair first met in 2011, Mohseni, a biomedical engineer, and Nudo, a neuroscientist, had been prototyping a BMI in rats that would restore communication between damaged brain regions. [4] The pair eventually sold the rights to the name in 2015 to Elon after being unable to hurdle many of the obstacles that stand in front of BMI development. The main hurdle that the pair faced was the lack of profitability in their business model. The engineering complexity and bureaucratic red tape that surrounds a medical device that will be implanted in the brain meant many investors were reluctant to invest without a solid proof of concept. In addition, even if a promising product was produced, there is a fairly limited pool of patients that could benefit from these devices. The pair were optimistic in Musk's endeavor however, as they believe his wealth of capital and plan to make the device applicable to the general public will overcome the hurdles that they and their predecessors could not. The core of Neuralink's team consist of industry experts in the fields of neuroscience and engineering. Notable members of this team include Flip Sabes and Tim Gardner who both left tenured positions at UC San Francisco and Boston University respectively to be apart of Neuralink. [5] As of now Neuralink hopes to start testing its product in humans in 2022, 2 years later than its original goal of 2020. [6]

Technology

Background on BMIs

BMIs use mathematical transformations to collect and interpret multi neuron signals from the motor cortex and the frontal and parietal lobes. There have been multiple important breakthroughs in BMI technology. One of the first is parallel recording instead of serial recording. Serial recording would involve taking recordings from the same neuron multiple times to detect a discernable pattern that could be correlated with a certain stimulus. Parallel recording would instead look at multiple neurons at a single time. Another break through was the idea of distributed coding. Distributed coding is the idea that the single neuron is not the key functional unit but it is the collection of multiple neurons in a population that encode information. When BMIs were first being developed the consensus in the neuroscience community was that the information from a single neuron had enough predictive power to describe physiological phenomena. Finally, modern computational models are used to simultaneously extract various motor parameters (such as arm position and velocity, or hand gripping force) in real time from the extracellular activity of frontal and parietal cortical neurons. These models are often first trained to predict motor movements by observing modulations in the neuronal ensemble activity of animals as they perform certain tasks. [7]

Listed below is a rough timeline of crucial developments in neuroscience and engineering that eventually lead to the development of BMIs. [8]
~1802: Thomas Young proposes population coding
~1950s: First multi electrode recording experiments
~1980s: Modern approach of sampling extracellular activity from neurons
~1990s: Expansion of electrophysiology and imaging methods
~2010s: Introduction of companies specific to brain machine interfaces

Innovations

Neuralink's website list multiple areas of innovation that separate its product from the current BMIs that are on the market. The first area that Neuralink focuses on is the application of its product compared to other BMIs. Neuralink explains how most of the popular BMIs, such as deep brain stimulation, are used for recording or stimulating the brain only. The Link, Neuralink's flagship product, will be able to record and interpret neuronal signals as well as communicate with them. The second area of innovation that was stated by Neuralink was the number and size of electrodes. Due to the mechanisms of current BMIs, the devices themselves require relatively few electrodes, with the electrodes themselves being fairly large. Neuralink states how its device will have 1024 electrodes, and each one will be extremely small yet flexible. [9]

Another innovation of Neuralink is centered around the electrodes themselves. Classic electrodes are made of metals so they are rigid enough to allow easy penetration into the brain. However, this rigid nature makes the electrodes stand out as a foreign insert within the brain and thus elicit quick and heavy immune responses. These immune responses can hinder the ability of the electronics to pick up clear neuronal signals. The limited flexibility also limits the number of neuronal populations that can be reached and recorded from as the probes can not by pass important brain structures. Neuralink attempts to solve this issue by creating flexible probes that are made of flexible and biocompatible materials with good conducting properties. However, their long term efficacy in biological environments is still not clear. [10]


Besides the BMI itself another area of innovation in Neuralink's design is how they plan to implant the device. Due to the extremely small scale of the electrodes, each one being 5 microns wide (a human hair is 70 microns wide), the company has developed a surgical robot that will help assist surgeons during implantation. [11] The company hopes that in the near future the robot will be able to completely automate the procedure. Neuralink's scientist and engineers created the technology that allows the robot to function however they outsourced the aesthetic design and user interface of the robot to a third party company, Woke Studio. [12] The company focused heavily on making the robot have an "anthroprmorphic characteristic". The principle behind the design was to minimize the invasiveness of the procedure by having the robot that was doing the procedure be aesthetically pleasing. [13] The robot is split into 3 different sections, the head, body, and base. The head holds the head of the patient and holds all the surgical tools needed for the procedure along with a host of sensors for mapping the head and brain. [14] The body holds all of the mechanical machinery that allows the robot to move and the base holds the computational hardware. The robot currently has a autoinsertion mode and can insert up to 6 probes a minute, however, the surgeon can intervene at any point and make manual adjustments. [15]

Neuralink surgical robot - open source image
Needle of Neuralink surgical robot - open source image

Neuralink vs DBS

One of the most popular BMIs used today is deep brain stimulation (DBS). Its efficacy in treating neurological disorders such as Parkinson's and essential tremor has made it a mainstay in clinically used BMIs. [16] However, DBS is a technology that is already two decades old and has not seen much innovation. [17] DBS uses 4 electrodes, with each one being 1.27 mm in diameter .[18] This limits the efficacy of DBS by offering low spatial resolution, the electrodes can not stimulate very precise regions of brain, and by limiting the amount of endemic neural signaling that can be recorded. [19]This latter limitation means DBS can not be individualized to each patient's unique symptoms and pathology. Future discussion of how DBS should evolve draws various parallels with the current Neuralink design. A need for better spatial resolution means having smaller and more plentiful electrode. More electrodes also means that more data can be recorded from the patient and thus their treatments can be catered to their specific needs.

Another defining characteristic of DBS is the implantable pulse generator, the device that powers the DBS electrodes inserted into the patient's brain. This implantable pulse generator is regularly inserted under the collar bone of the patient, similar to where a pacemaker would be placed. The device is two inches in diameter and one half inch thick and often produces a notable bump in the skin. Since the device is a battery it will eventually run out of power. Thus patients are given the option of a non rechargeable battery that last 3-5 years or a rechargeable one. The rechargeable battery is charged by holding a device over the site and must be charged several times a week.[20] The surgery that implants the DBS electrodes and implantable pulse generator requires 6 steps and takes around 4 hours.[21]The size and surgery of deep brain stimulation are areas that Neuralink's technology has innovated. Compared to the two by one half inch implantable pulse generator, Neuralink's device is 23 millimeters by 8 millimeters and will sit in the skull so there will be no visible protrusion. The company also says that the device can be implanted into a patient under one hour without general anesthesia and that the patient would be able to leave the the hospital on the same day. [22]

Issues

Technological

One of the first issues surrounding Neuralink's technology is the materials that are being used to make the threads that will be inserted into the brain. In order to improve the electrophysiological properties of the miniscule threads and have better recording features they were coated with poly-ethylenedioxythiophene doped with polystyrene sulfonate. The issue with this polymer is since it is so novel there is limited research on its biocompatibility and what type of immune response it could generate once inside the brain. [23] Another issue is centered around the robot that is used for implanting the device. Although the robot has impressive speed, its first 19 surgeries had a 87.1 percent success rate with a standard deviation of 12.6% [24]. Considering that these surgeries were not done on human patients but on mammals with much less complex neural anatomy, the accuracy of the robot may be brought into question. The invasiveness of the procedure means that patients will unlikely be comfortable with anything less than absolute accuracy during the procedure. Neuralink has also been criticized for its recent demos that involved its device being implanted into animals such as monkeys and pigs. Among the critics are tenured neuroscientists that express doubt over the novelty of the demonstration. Andrew Jackson, a professor at Newcastle University remarked how the 1024 recording electrodes was not necessarily impressive for today's standards, but did say how the ability to relay their information wirelessly was impressive. [25] The professor went on to remark how the largest limitation that brain machine interface technology has encountered is not recording information from the brain but finding ways to interpret it. He remarked how the Neuralink demo displayed encoding methods and technologies that have been displayed before. He also expressed doubt to the more exotic claims that were presented by Neuralink, such as the ability to write information to the brain and replay memories. He says that there are fundamental limitations around electrical stimulation and how it can convey information to the brain. Overall Dr. Jackson stated that neuralink demonstrated, "good engineering but mediocre neuroscience". [26]

Ethical

One of the most pressing ethical concerns surrounding Neuralink centers around its claim that its technology could one day manipulate an individual's mind and potentially greatly increase their mental performance. An article published by Onezero discussed how this type of ability would potentially make patients become extremely reliant on the device and thus give Neuralink unprecedented power over them. The article drew ethical parallels to how current laws handle other implantable medical devices such as implantable cardioverter defibrillators. [27] The parallel centered around how the implantable cardioverter defibrillators often have a feature called remote monitoring. The devices are crucial for the patients that they are implanted in as they regulate atypical beatings of the heart that would otherwise be fatal. Inorder to accurately perform this task the defibrillators track a large amount of data about the patient's heart. This data is relayed from the defibrillator to a bedside monitor that the patients use and this monitor sends the information back to the manufacturer to be evaluated. This data is often sent to the patient's doctor if requested but the patients themselves are prohibited from viewing their own data. The reasons behind this revolves around liability and regulatory issues but it also lies in the fact that doctors often do not want this type of information to be known to their patients. [28] The technology that Neuralink is promising exacerbates this issue as its device would theoretically be capable of collecting data about an individual's memories and cognitive ability. Issues still arise even if Neuralink's more ambitious promise's fall through. The privatization of a medical device that could become essential for the quality of life of many patients has been the center of many ethical debates. A parallel can be drawn to the price of insulin and how expensive it has become. An article by Vox states that the main reason for its high price in America is due to America's free market approach to the pharmaceutical market. [29] Since the brain machine interface market is still in its infancy, Neuralink is poised to dominate the market in the near future and may have little to no incentive to produce its product at a reasonable price.

Another ethical concern centers around issues of morality. An article by the University of Sydney law school broadly talks about the implications of all future neurotechnologies on criminal law. Some of the questions that they propose include: [30] If people start committing crimes by way of brain-computer interface, what is the criminal act? What if there was a brain implant that could detect the neural patterns associated with impulsive aggression, then issue a warning to the person that they may be on the cusp of a violent outburst, or even automatically act on their brain to calm them down?

The second question had been pondered by Dr Allan McCay, a Sydney University Law School professor. He discusses how stimulation by Neuralink's device or other BMIs in general could be used to alter personalities, whether purposefully or accidently. He talks about how many people may get BMI's such as Neuralink's to handle temper and impulsivity, and thus their actions may be so manipulated by these machines that it may not be reasonable to hold them liable to any of their actions. [31] He warns that the company has to navigate a fine line during its design to ensure that the device can not cause an action to occur due to a fleeting thought. There must be an effortful and conscious decision that leads to an action. [32]

There have already been attempts to insure that the modern legal system does not become overwhelmed by the introduction of BMIs into the general public. A group of academics encompassing fields such as neuroscience, ethics and medicine have called for the development of ethical guidelines that should be used to guide development and deployment of BMIs and other human enhancing technologies. The group is led by Columbia University neuroscientist Rafael Yuste and University of Washington bioethicist Sara Goering and published a paper in the acclaimed scientific journal, Nature. [33]

The paper outlined four concerns that the authors had for future BMI technology.

The first one is privacy and consent. The authors suggest that data gathered by neurotechnologies have incredible predictive power and that its users should be opted out from sharing their neural data by default. If users do decide to opt in, the sharing of their data should be heavily regulated as enough neural data from opt in users could be used to draw conclusions on opt out users.[34]

The second is agency and identity. The paper discusses how invasive procedures such as deep brain stimulation have already led to patients feeling a profound change in self and identity. The group suggests that BMIs could cause its patients to act on impulses that would normally be repressed or fleeting through "auto-complete" functions. In order to combat this issue the authors suggest the formation of an international committee that outlines the basic human rights of agency and identity. [35]

The third is augmentation. The paper predicts that the major augmentations to mental capacity will make BMIs a symbol of high social class in the future. Thus they suggest that limitations need to be placed on how significant these augmentations can be. They also advocate for regulating development of BMIs for military use as an "arms race" of neural augmentation may develop. [36]

The final concern is bias. The paper demonstrates how bias has already made its way into machine learning through targeted advertisement that often marginalizes certain groups. They warn that BMIs may become biased towards certain demographics and thus further increase societal discrepancies. [37]

References

  1. ↑ Studio, P. (n.d.). Approach. Neuralink. Retrieved February 11, 2022, from https://neuralink.com/approach/
  2. ↑ Barton, L. (2021, September 19). Tracing the history of Neuralink. Medium. Retrieved February 11, 2022, from https://medium.com/@lauren.barton/tracing-the-history-of-neuralink-b93354e9a504
  3. ↑ Regalado, A. (2020, April 2). Meet the guys who sold "Neuralink" to elon musk without even realizing it. MIT Technology Review. Retrieved February 11, 2022, from https://www.technologyreview.com/2017/04/04/152788/meet-the-guys-who-sold-neuralink-to-elon-musk-without-even-realizing-it/
  4. ↑ Regalado, A. (2020, April 2). Meet the guys who sold "Neuralink" to elon musk without even realizing it. MIT Technology Review. Retrieved February 11, 2022, from https://www.technologyreview.com/2017/04/04/152788/meet-the-guys-who-sold-neuralink-to-elon-musk-without-even-realizing-it/
  5. ↑ Masunaga, S. (2017, November 20). A quick guide to elon musk's new brain-implant company, Neuralink. Los Angeles Times. Retrieved February 5, 2022, from https://www.latimes.com/business/technology/la-fi-tn-elon-musk-neuralink-20170421-htmlstory.html
  6. ↑ Hamilton, I. A. (2021, December 7). Elon Musk's Neuralink wants to embed microchips in people's skulls and get robots to perform brain surgery. Business Insider. Retrieved February 5, 2022, from https://www.businessinsider.com/neuralink-elon-musk-microchips-brains-ai-2021-2#elon-musk-said-the-company-hopes-to-start-implanting-its-chips-in-humans-in-2022-two-years-later-than-hed-originally-hoped-8
  7. ↑ Nicolelis, M. A., & Lebedev, M. A. (2009). Principles of neural ensemble physiology underlying the operation of brain–machine interfaces. Nature Reviews Neuroscience, 10(7), 530–540. https://doi.org/10.1038/nrn2653
  8. ↑ Chapin, J. K., Moxon, K. A., Markowitz, R. S., & Nicolelis, M. A. (1999). Real-time control of a robot arm using simultaneously recorded neurons in the motor cortex. Nature Neuroscience, 2(7), 664–670. https://doi.org/10.1038/10223
  9. ↑ Studio, P. (n.d.). Approach. Neuralink. Retrieved February 5, 2022, from https://neuralink.com/approach/
  10. ↑ Musk, E. (2019). An integrated brain-machine interface platform with thousands of channels. JMIR Publications, 21(10). https://doi.org/10.1101/703801
  11. ↑ Alonzo, I. (2020, August 29). Elon Musk: Neuralink's chip will sew as many as 1,024 thin electrodes in the brain. Tech Times. Retrieved February 7, 2022, from https://www.techtimes.com/articles/252124/20200828/neuralink-machine-will-sew-many-1-024-impossibly-thin-5.htm
  12. ↑ Etherington, D. (2020, August 28). Take a closer look at Elon Musk's Neuralink Surgical Robot. TechCrunch. Retrieved February 7, 2022, from https://techcrunch.com/2020/08/28/take-a-closer-look-at-elon-musks-neuralink-surgical-robot/
  13. ↑ Etherington, D. (2020, August 28). Take a closer look at Elon Musk's Neuralink Surgical Robot. TechCrunch. Retrieved February 7, 2022, from https://techcrunch.com/2020/08/28/take-a-closer-look-at-elon-musks-neuralink-surgical-robot/
  14. ↑ Etherington, D. (2020, August 28). Take a closer look at Elon Musk's Neuralink Surgical Robot. TechCrunch. Retrieved February 7, 2022, from https://techcrunch.com/2020/08/28/take-a-closer-look-at-elon-musks-neuralink-surgical-robot/
  15. ↑ Musk, E. (2019). An integrated brain-machine interface platform with thousands of channels. JMIR Publications, 21(10). https://doi.org/10.1101/703801
  16. ↑ Cagnan, H., Denison, T., McIntyre, C., & Brown, P. (2019). Emerging technologies for improved deep brain stimulation. Nature Biotechnology, 37(9), 1024–1033. https://doi.org/10.1038/s41587-019-0244-6
  17. ↑ Gardner, J. (2013). A history of deep brain stimulation: Technological Innovation and the role of Clinical Assessment Tools. Social Studies of Science, 43(5), 707–728. https://doi.org/10.1177/0306312713483678
  18. ↑ Cagnan, H., Denison, T., McIntyre, C., & Brown, P. (2019). Emerging technologies for improved deep brain stimulation. Nature Biotechnology, 37(9), 1024–1033. https://doi.org/10.1038/s41587-019-0244-6
  19. ↑ Cagnan, H., Denison, T., McIntyre, C., & Brown, P. (2019). Emerging technologies for improved deep brain stimulation. Nature Biotechnology, 37(9), 1024–1033. https://doi.org/10.1038/s41587-019-0244-6
  20. ↑ University of Columbia. (2021, July 21). Deep Brain stimulation. Columbia Neurosurgery in New York City. Retrieved February 7, 2022, from https://www.neurosurgery.columbia.edu/patient-care/treatments/deep-brain-stimulation
  21. ↑ Mayfield. (2021, January 12). DBS. Deep Brain Stimulation (DBS) for Parkinson's & Essential Tremor | Mayfield Brain & Spine, Cincinnati. Retrieved February 11, 2022, from https://mayfieldclinic.com/pe-dbs.htm#:~:text=What%20happens%20during%20surgery%3F,lasts%203%20to%204%20hours.
  22. ↑ Hitti, N. (2020, September 2). Elon Musk unveils updated Neuralink brain implant design and surgical robot. dezeen. Retrieved February 11, 2022, from https://www.dezeen.com/2020/09/02/neuralink-elon-musk-brain-implant-technology/
  23. ↑ Musk, E. (2019). An integrated brain-machine interface platform with thousands of channels. JMIR Publications, 21(10). https://doi.org/10.1101/703801
  24. ↑ Musk, E. (2019). An integrated brain-machine interface platform with thousands of channels. JMIR Publications, 21(10). https://doi.org/10.1101/703801
  25. ↑ Science Media Centre. (2020, August 29). Expert reaction to Elon Musk's Neuralink demonstration involving Pigs. Science Media Centre. Retrieved February 11, 2022, from https://www.sciencemediacentre.org/expert-reaction-to-elon-musks-neuralink-demonstration-involving-pigs/
  26. ↑ Science Media Centre. (2020, August 29). Expert reaction to Elon Musk's Neuralink demonstration involving Pigs. Science Media Centre. Retrieved February 11, 2022, from https://www.sciencemediacentre.org/expert-reaction-to-elon-musks-neuralink-demonstration-involving-pigs/
  27. ↑ Maynard, A. (2019, November 7). Neuralink's technology is impressive. is it ethical? Medium. Retrieved February 11, 2022, from https://onezero.medium.com/neuralinks-technology-is-impressive-is-it-ethical-812afb38b19e
  28. ↑ Campos, H. (2015, March 24). I can't access the data generated by my implanted defibrillator. that's absurd. Slate Magazine. Retrieved February 11, 2022, from https://slate.com/technology/2015/03/patients-should-be-allowed-to-access-data-generated-by-implanted-devices.html
  29. ↑ Belluz, J. (2019, April 3). The absurdly high cost of insulin, explained. Vox. Retrieved February 11, 2022, from https://www.vox.com/2019/4/3/18293950/why-is-insulin-so-expensive
  30. ↑ The University of Sydney. (2020, September 8). The potential pitfalls of technologies like Elon Musk's Neuralink. The University of Sydney. Retrieved February 11, 2022, from https://www.sydney.edu.au/law/news-and-events/news/2020/09/08/potential-pitfalls-of-technologies-like-elon-musks-neuralink.html
  31. ↑ Kennedy, S. (2020, September 1). Elon Musk's Neuralink is an ethics maze. InnovationAus. Retrieved February 11, 2022, from https://www.innovationaus.com/elon-musks-neuralink-is-an-ethics-maze/
  32. ↑ Kennedy, S. (2020, September 1). Elon Musk's Neuralink is an ethics maze. InnovationAus. Retrieved February 11, 2022, from https://www.innovationaus.com/elon-musks-neuralink-is-an-ethics-maze/
  33. ↑ Martineau, K. (2018, December 9). Experts call for ethics rules to protect privacy, free will as brain implants and AI merge. Columbia News. Retrieved February 11, 2022, from https://news.columbia.edu/content/experts-call-ethics-rules-protect-privacy-free-will-brain-implants-and-ai-merge
  34. ↑ Yuste, R., & Goering, S. (2017). Four ethical priorities for neurotechnologies and ai. Nature, 551(7679), 159–163. https://doi.org/10.1038/551159a
  35. ↑ Yuste, R., & Goering, S. (2017). Four ethical priorities for neurotechnologies and ai. Nature, 551(7679), 159–163. https://doi.org/10.1038/551159a
  36. ↑ Yuste, R., & Goering, S. (2017). Four ethical priorities for neurotechnologies and ai. Nature, 551(7679), 159–163. https://doi.org/10.1038/551159a
  37. ↑ Yuste, R., & Goering, S. (2017). Four ethical priorities for neurotechnologies and ai. Nature, 551(7679), 159–163. https://doi.org/10.1038/551159a
Retrieved from "http://si410wiki.sites.uofmhosting.net/index.php?title=Neuralink&oldid=109964"