In an August 2, 2024 interview with Lex Fridman, Elon Musk suggested that brain machine interfaces that depend on electrodes implanted in the neocortex will have the capacity to transmit some 1 x 10^6 bits per second, which far surpasses the information throughput of humans at 10 bits per second (on average) or discriminating 1,024 possibilities per second (2^10). Current brain machine interfaces transmit information below 3 bits per second (less than 8 possibilities per second) falling well short of that transmitted by a cochlear implant at 10 bits per second [corrected for percent-correct and redundancy, Tehovnik, Hasanbegović, Chen 2025; but see Mineault et al. 2024, Table 3, for other estimates].
The idea that if one attaches electrodes to every synapse (or information unit) in the human neocortex to bypass the bottleneck of the motor system (at 10 bits per second) then one will generate an unconstrained flow of information of a trillion bits per second (3.2 trillion bits per second based on the number of neocortical synapses and spike rate, Tang et al. 2001; Tehovnik, Hasanbegović, Chen 2025) is to misunderstand how the brain works. This misunderstanding is common amongst engineers who focus exclusively on the parameters of engineering while disregarding biology. Let us take Einstein’s ‘E = mc^2’. To express this equation verbally requires a throughput of about 20-30 bits per second (Reed and Durlach 1998). Now for Einstein to deduce this equation, he spent many years contemplating the details of physics using a neocortex that transmits information internally at a trillion bits per second (and shared with the cerebellum and other subcortical structures). Therefore, if Musk were to attach all his electrodes to every synapse of the neocortex, I predict that he would produce a string of garbage since thinking involves more than just generating signals willy-nilly from a ball of synapses. Under normal circumstances these signals must be sorted by the nervous system to make sense to an organism, a sense that scientists have spent their careers developing, communicating, and refining while engaging their colleagues.
Footnote 1: Musk, E., 2024. Elon Musk: Neuralink and the Future of Humanity’ Lex Fridman, Podcast #438, Aug. 2.
Elon Musk is the most contemptible figure in politics. He's far right all the way. It's really sad when a person with such a wealth is pushing for far-right ideologies to be rolled out. I know that deplorable even met with Bolsonaro, a very deplorable figure from the Brazilian far-right. He's sensationalist, he's dishonest, he's ridiculous. He's not a genius, Feynman is a genius, Penrose is a genius, Lagrange is a genius, Susskind is a genius, Elon Musk is just a capitalist vulture. Not saying he's not smart, he is, but not like the true geniuses. He has a lot of sycophants.
While Musk’s engineering mindset allows him to conceptualize innovative solutions for BMIs, his interpretation of the brain’s data throughput highlights a fundamental misunderstanding of neuroscience. The human brain processes information through complex, multi-layered interactions between neurons, synapses, and neural circuits. These interactions are not solely about raw data transmission but involve dynamic sorting, filtering, and contextual interpretation by various brain regions, including subcortical structures like the cerebellum. This complexity cannot be fully replicated or understood by merely scaling up the number of electrodes interfacing with the neocortex.
For example:
1. Biological Context Matters: The neocortex operates in concert with other brain areas to generate coherent thoughts and behaviors. Attaching electrodes to synapses without considering these interactions would likely result in chaotic outputs, as meaningful cognition depends on the system’s holistic organization.
2. Throughput ≠ Cognition: The claim that increasing data throughput would directly enhance thought or understanding disregards the qualitative aspects of neural activity. For instance, Einstein’s work on relied not on high-speed neural data transfer but on years of structured thinking, experimentation, and conceptual synthesis.
3. Neural Interfaces Are Limited: Current BMIs, as cited, transmit data at rates below cochlear implants (e.g., 3 bits per second versus 10 bits per second). Achieving 1 million bits per second, as Musk proposes, would require overcoming immense technical and biological challenges, including decoding the meaning of signals from trillions of synapses.
Explanation
Musk’s perspective is characteristic of engineers who prioritize the quantitative aspects of systems while overlooking the nuanced biological processes underpinning cognition. Neuroscience emphasizes that the brain’s function is not reducible to its raw data throughput; rather, it involves highly integrated processing that supports learning, memory, decision-making, and emotion. Hence, even if BMIs could theoretically transmit data at trillion-bit speeds, the practical output would not align with human cognition unless these signals are interpreted and integrated meaningfully.
Judgment
Musk’s approach is visionary in its ambition to push the boundaries of BMI technology. However, it reflects a reductionist view of the brain that oversimplifies its function. Neuroscientific insights suggest that meaningful progress in BMIs will require a deeper understanding of how the brain organizes and interprets information, not just an increase in data transfer rates. Therefore, while Musk’s vision for BMIs is inspiring, it needs to be grounded in biological realities to achieve practical and meaningful advancements.
Thank you for your detailed response, Mohammed Saha Alam. Ed Tehovnik.
- Badges
- Science topic
- Similar topics
- Computer Communications (Networks)
- Simulators