Elon Musk has the mind of an investor rather than the mind of a scientist. A case in point is his recent adventure into brain-machine interfaces via Neuralink (https://en.wikipedia.org/wiki/Neuralink). He (and his team of neuroscientists) believe that the way to solve the problems associated with current devices is to plunge more flexible electrodes into the neocortex to trigger a technological advancement, not knowing that once 40 neurons have been sampled in the neocortex the information transfer hits a ceiling (Tehovnik and Chen 2015). Brain-machine interfaces (unlike the human brain) fail to transmit sufficient information to outperform even a rhesus monkey doing an 8-choice object identification task at 95% correctness, namely, anything better than 2.5 bits per second without the assistance of body movements (e.g., Lorach et al. 2023; Martin et al. 2014; Metzger et al. 2023; Nicolelis 2019; Willett et al. 2021; also see Tehovnik, Woods et al. 2013). Once after completing a lecture on how brain-machine interfaces transfer little information, an engineer in the audience (at a university in Texas) claimed that supporting such a device “makes patients feel better” despite the low transfer rate. There are cheaper and less dangerous ways to make patients feel better, for implanted devices can even kill if not secured properly [see: Reply by Tehovnik and Schiller to Schmidt et al's letter, 2010].
Brain-machine interfaces that are reliant on neural discharges from the neocortex have been deemed highly unreliable for controlling artificial devices as well as for controlling muscles (Rokni et al. 2007; Schaeffer and Aksenova 2018). In regard to muscles, this unreliability is related to changes in muscle tension, sensory noise, attention, and mood, and to the enhancement of muscle fatigue (Schaeffer and Aksenova 2018). Rokni et al. (2007, p. 653) have suggested that “...any single behavior can be realized by multiple configurations of synaptic strength” in the neocortex, which might explain the unreliability. The final common pathways to muscles (i.e., the alpha motor neurons) have minimal discharge variability during repeated behavioral executions (Sartori et al. 2017). Hence, the further one's neural recordings (i.e., brain implants) are from the final common pathways (as well as from the sensory receptors, e.g. for the cochlear implant) the more unreliability will be introduced to the preparation. This propensity is magnified as the number of neurons sampled is reduced. Some have toyed with the idea of using Machine Learning to optimize the performance of cortical brain-machine interfaces. The power of such an optimization will be contingent on sample size as well as on having continuous and faithful feedback on the outcome of previous behavioral performances. Indeed, Stephen Hawking was wise not to have a brain-machine interface implanted in his neocortex to facilitate his physics. We are years away from a technological breakthrough on this front and Elon Musk will most likely not be part of the breakthrough since he remains ignorant of the science behind brain-machine interfaces, which should be a warning to all investors trying to avoid a Ponzi scheme.
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Biomolecules
- Peptides