>>11926451>The one actual neuroscientist ITTI'm not the anon you're replying to, but I am the neuroscientist who posted here earlier.
>said he thinks it's not possible (obviously)For the record, I did not say this.
I won't argue that Neuralink's ultimate goal is something akin to 'mind uploading', or that such a goal would be a fantasy, at least within our lifespans or the lifespan of Neuralink. But Neuralink's proximate goal, developing 'brain interfaces', I wouldn't discount as a theoretical impossibility at all.
One major and in my view unjustified assumption in the points you've outlined earlier is that any interaction with a brain needs to occur at the synaptic level. An interface would not necessarily need to be restricted to such a high degree of spatial specificity in order to have any usability. In fact, most interfaces that exist today do not target individual synapses or even neurons but work instead through voltage regulation of the extracellular potential.
Yes, every brain is different, but there do exist rules for the topography of macroscale and mesoscale architecture, such as retinotopy and tonotopy. As such, neurons that subserve particular computational roles tend to cluster in relative proximity to one another, and one would only could interact with population activity en masse, so long as it falls within the spatial bounds dictated by the process of interest. In addition, a lot of recent work indicates that neurons by themselves encode very little relevant information, but encoded features (and in particular abstract ones) emerge at the population level. Therefore, manipulation of features of the population level code are what would be required to begin with.
So just to clarify, I don't think practically applicable BCIs are theoretically impossible. I do, however, think that they are a very long way off, and have not seen anything from Neuralink that significantly works towards this goal beyond what is already in the literature.
