What are some of the criticisms of Elon Musk

Neuralink: brain-computer interface by Elon Musk and criticism of the publication

The media hype was great when Elon Musk announced his new visions in the field of neurotechnology on July 17th, 2019 [I]. With his company “Neuralink” he wants to insert flexible electrodes into the human brain in a minimally invasive manner using a robot and link them to a computer. The electrodes are supposed to record the activity of 1,000 nerve cells at the same time - and to be able to target them in a targeted manner in the future. He is pursuing the seemingly illusory ideas of being able to "import" new languages ​​into the brain at the push of a button or to make advertising and city maps available directly in the brain. The goals of treating diseases of the central nervous system, such as paraplegia, seem more realistic.

On the day after the press conference, Musk published a publication with scientific details on the progress of his company on the preprint server BioRxiv (see primary source), which mainly presents robot and electrode technology. If scientific publications are published on a preprint server, this means that they have not yet been reviewed by independent scientists (peer review). Even after the first wave of reporting is over, it is worth taking a closer look at this type of publication. It raises the question of the scientific nature of such projects. The scientific community, for example, was quick to criticize the practice of publication [II] because, in addition to Musk's name, other scientific authors of the paper remain hidden behind the company name. Another point of criticism is that the results presented do not represent any real progress in the field of the brain-computer interface.

We asked experts to assess the results from a neurotechnology and science policy perspective. In addition, we asked the preprint server BioRxiv directly which criteria and mechanisms are used to ensure a certain scientific quality of the uploaded publications.

Overview

     

  • Prof. Dr. Ulrich Dirnagl, Director of the Experimental Neurology Department, Charité - Universitätsmedizin Berlin
  •  

  • Dr. Philipp Kellmeyer, specialist in neurology, University Medical Center Freiburg, and employee in the research focus Responsible Artificial Intelligence at the Freiburg Institute for Advanced Studies (FRIAS), Albert-Ludwigs University of Freiburg
  •  

  • Dr. John R. Inglis, co-founder of bioRxiv and medRxiv, Cold Spring Harbor Laboratory; executive director and publisher, Cold Spring Harbor Laboratory Press, United States
  •  

Statements

Prof. Dr. Ulrich Dirnagl

Director of the Department of Experimental Neurology, Charité - Universitätsmedizin Berlin

"Summary.
The publication is an advertising brochure from Neuralink. Behind this is an insane amount of money and an impressive collaboration between engineers, programmers, material experts, robotics specialists and others. This has what it takes to advance the field of the brain-machine interface (BMI) incrementally (a bit), i.e. to improve existing BMI systems for controlling the simplest functions through brain activity after intensive training. This could help individual patients, for example after paraplegia - that is of course great. Everything else is dubious hype, science fiction and nothing in the article or our other knowledge about the function of the brain. We have to admit that we practically do not know how the brain works anyway. And the apparatus described will not help us any further. Overall, this episode reminds me very much of the Human Brain Project, which is funded by the EU with over one billion euros. Exactly ten years ago, on July 22nd, 2009, the father of this project, Henry Markram, announced in a TED Talk in Oxford that 'we will be able to simulate the human brain in the computer'. This is the stated aim of the Human Brain Project. Through this simulation we would then understand perception and reality, maybe even our physical reality. He concluded that in ten years (i.e. today) a hologram would hold his TED Talk [1]. Nothing like this has happened in the ten years, except that a lot of money has been spent. "

"The item.
The Neuralink article describes components of a Brain Machine Interface (BMI), i.e. an invasive, i.e. surgically implanted, 'medical device' that enables the brain to control a 'machine' using thoughts after training. This could be a robotic arm or the control of commands on a computer by moving a mouse cursor. "

“The article superficially describes some elements of a BMI generated by the company as a non-functional prototype: electrodes for recording brain activity, a chip for processing and transmitting the signals and an operating theater robot for inserting the electrodes. Everything that is still needed after the signals have been derived is missing, for example the robot that controls the function and so on. So no BMI is described, but parts of it. "

"What's new?
The BMI principle and the components used are not new, but are being developed in a similar form by various groups around the world and have already been used in selected patients with spinal cord injuries. After a long training session, they were able to perform very simple functions, such as grabbing a cup. "

“However, the article describes a number of technical improvements that could potentially improve the functionality of BMIs, for example: new, very thin lead electrodes, the option of deriving a few thousand (instead of a few hundred) electrodes and the surgical robot for Implantation. Nevertheless, the promise that this could improve a BMI (which is plausible) has not been shown (formulation: 'It is plausible to imagine ...'). Much is plausible and we can all imagine a lot. Of course, this is only taken seriously at this point because Elon Musk is behind it. That should guarantee that there has to be something to it. After all, he also builds electric cars ... "

“Is that a scientific article?
Everything beyond this - such as the functions indicated in the article, for example that the brain can be stimulated with it, or as Elon Musk claimed during the presentation that a 'symbiosis of the human brain with artificial intelligence' would be possible with it - remains pure, through nothing substantiated speculation. "

“The article comes in the guise of a scientific publication, but is nothing more than an advertising brochure for the Neuralink company. The specifications and services of the system indicated there are not traceable. At one point the article is referred to as a 'white paper'. Scientifically, the article offers nothing new: Other groups are much more advanced and have already shown applications on humans. "

“In this respect, the acceptance of the article as a preprint at Biorxiv must at least be questioned, the category in which it was published is 'New results'. Basically, however, it can be said that the essence of a preprint is not to be subject to any peer review or editorial review. The only condition is, 'that it concerns a relevant scientific field, the content is unpublished at the time of submission, and all its authors have consented to its deposition' (Biorxiv [2]). The dispute then takes place via comments on Biorxiv, on Pubpeer, through direct contact with authors, or in blogs or on Twitter. This is also happening at the moment, on Pubpeer there are, for example, devastating reviews by Lydia Maniatis [3]. "

“Ethical concerns.
Nevertheless, ethical concerns arise: Elon Musk is the sole author, although it is absolutely clear that he could not have written the article at all. This is a clear violation of the generally and internationally accepted publication ethics (COPE) [4]. The fact that the animal experiments outlined in the article were 'approved' by an in-house review board is very questionable - albeit lege artis in the USA, although unthinkable in Germany. So you decided yourself that the experiments are ethically justifiable. In addition, the article raises unfounded hopes in desperate patients, which is also unethical. "

“The hype.
The hype consciously generated with the article and the presentation by Elon Musk suggests abilities of a BMI that would not be accepted by any serious BMI researcher. The triggering of 'spikes' of individual or groups of neurons, even if they come from several thousand locations, cannot make any statements about thoughts, ideas and the like of a brain. It is exactly the opposite: The principle of a BMI is precisely to train the brain to willingly generate activities of nerve cells that have not occurred there before in order to trigger a specific (previously unknown to the brain) machine reaction . That is why it takes a very long time until even the simplest functions (cursor up, cursor down) work only halfway reliably. When you look at the physical and psychological exertion of patients in operating a BMI, something becomes clear. In a not inconsiderable proportion of patients who have tried something like this, it doesn't work at all. It is unclear whether an increase in the number of electrodes can significantly improve the control capability and is questioned by researchers in the field. "

Dr. Philipp Kellmeyer

Specialist in neurology, University Medical Center Freiburg, and employee in the research focus Responsible Artificial Intelligence at the Freiburg Institute for Advanced Studies (FRIAS), Albert-Ludwigs University of Freiburg

"The announcements published by Neuralink in the preprint - and thus not yet examined in a peer-review process - and the presentation by Tesla boss Elon Musk represent, in my opinion, a summary of interim results of current in-house research."

“The work on electrode technology is not fundamentally new, as various working groups around the world are working on electrode systems with high-density electrode grids. In detail, the materials used, the arrangement of the electrodes and the packaging (hermetic packaging) may differ from other approaches. Wireless signal transmission has also become the standard in the development of newer systems for brain-computer interfacing. As long as there are no independently assessed long-term results on tissue compatibility and non-toxicity in animal experiments - the first application study is planned according to Neuralink 2020 - the results presented here should also be treated with caution. "

"The state of research is inadequately presented because, for example, important research work on flexibly embedded electrode implants is not discussed."

"Insofar as the work follows the standard of good scientific practice, it contributes in the usual sense to the incremental gain in knowledge and progress in the area of ​​brain-computer interfacing. However, since the work presented has largely not been scientifically assessed, let alone validated or reproduced, these quality standards have not yet been met. "

When asked to what extent it is compatible with scientific standards to hide the names of authors and participating scientists behind a business name:
“From my point of view, this does not correspond to the usual rules of good scientific practice. The publications should be made by the authors, naming them, who were significantly involved in the research. "

“From my point of view, the motivation for the publication of the preliminary results is not evident from the documents at this point in time, that is, this may sometimes be driven more by strategic considerations, either economically (Tesla has just again presented poor earnings figures and is economically under enormous pressure to succeed [5]) or with regard to 'intellectual property'(Intellectual property, in the sense of being able to prove that you were the first to come up with this idea; editor's note)instead of publishing for substantive scientific reasons.

“In Germany, freedom of research applies equally to academic research and to private-sector research. From a research ethical point of view, however, in my opinion it is important that the same rules of good scientific practice apply and that ethical standards are adhered to. Unfortunately, in most cases, corporate research does not meet these standards, and the current case of Neuralink is a negative example of this in my view. "

“For research that is primarily motivated by the acquisition of knowledge and progress in knowledge, the possibility of the preprint server is a great opportunity to exchange ideas with colleagues at an early stage (pre-publication peer review) and thus improve the quality of your work. In terms of open science, this is very much to be welcomed. However, if the main motivation is to secure intellectual property and exploitation rights at an early stage, in my opinion the system is being misused for the commercialization of science. "

“So far, most preprint services have only checked the submitted manuscripts very superficially in order to remove gross nonsense. The actual scientific review and validation then still depends on a detailed expert assessment of the work. "

“From my point of view, a transparent publication culture is essential to promote the open exchange of knowledge. In addition to the possibility of preprint publication, this also includes the pre-registration of studies, the open, i.e. non-anonymous review of manuscripts and many other measures that have been required by the open science community for several years. "

Dr. John R. Inglis

Co-founder of bioRxiv and medRxiv, Cold Spring Harbor Laboratory; executive director and publisher, Cold Spring Harbor Laboratory Press, United States

When asked to what extent there are rules for specifying authorship and how the current case of Neuralink is to be assessed:
“Papers from research consortia are rare on bioRxiv. The rules for communicating their work vary from consortium to consortium. There was public criticism that the paper failed to identify the people from the Neuralink consortium who did the work, but there were no comments from the individuals involved. We are examining whether or not a listing of the members of consortia should be a requirement for research from such groups to be submitted to bioRxiv in the future. "

“If you want to use the phrase 'technical filtering' to ask whether we use automated methods to check papers, the answer is no, with the exception of plagiarism detection. We have no plans to introduce such a technology. "

When asked how to understand the following description on the website: “All articles go through a thorough screening process for offensive and / or non-scientific content and for material that could pose a health or biosecurity risk and is checked for plagiarism ":
"It's a two-step process that is carried out by an internal team and an international group of over 120 primary auditors ('bioRxiv-affiliates')."

Information on possible conflicts of interest

All: No information received.

Primary source

Musk E & Neuralink (2019): An integrated brain-machine interface platform with thousands of channels. DOI: 10.1101 / 703801. Preprint publication on the bioRxiv server.
Preprint publication means that this publication has not yet undergone a scientific assessment by independent experts (peer review).

References cited by the experts

[1] Yong E (July 22, 2019): The Human Brain Project Hasn’t Lived Up to Its Promise. The Atlantic.

[2] bioRxiv: Submission Guide.

[3] Pubpeer: An integrated brain-machine interface platform with thousands of channels.Pubpeer is a platform on which scientific results are discussed in the community.

[4] COPE: Authorship and Contributorship.

[5] Siddiqui F (July 24, 2019): Tesla loss shows perils of lower-priced model 3. Washington Post.

References cited by the SMC

[I] YouTube channel CNET: Watch Elon Musk's Neuralink presentation. Published on 07/17/2019.

[II] Pubpeer: An integrated brain-machine interface platform with thousands of channels. Pubpeer is a platform on which scientific results are discussed in the community.

Further sources of research

Choi J-r et al. (2018): Implantable Neural Probes for Brain-Machine Interfaces - Current Developments and Future Prospects. Experimental Neurobiology; 27 (6): 453-471. DOI: 10.5607 / en.2018.27.6.453.