back
16 comments
Probably coincidental timing, but today it was reported that one of the last remaining co-founders left Neuralink looking to work on a less invasive approach, for safety reasons.

https://qz.com/neuralink-brain-chip-safety-elon-musk-benjami...

notably he said this... "Dr. Rapaport continued to explain on the podcast that this method has one big drawback. It causes “some amount of brain damage when they’re inserted to the brain.” That, of course, is less than ideal. And Rapaport says he believed it was possible, “to extract information-rich data from the brain without damaging the brain.”
This is highly concerning
> In the weeks following the surgery, a number of threads retracted from the brain, resulting in a net decrease in the number of effective electrodes

1. Was this expected/seen in animal studies?

2. Will it degrade further? Are follow-up surgeries needed?

Yes

Yes Yes

Neuralink is putting the cart before the horse: none of what they are doing here are the true challenges of neural interfaces, which are:

-either neurons move away from the stress caused by the probes

-or neurons are killed by the probes.

The lifetime of neural probes due to this problem is under a year, and let me remind you these direct probes need to be drilled into your skull. Neuralink is not solving anything: they are scoring cheap publicity points by doing something (signal analysis) nobody has done that much before, because it is too early to do it

These guinea pigs will be worse off after neuralink will have (ab)used them for this publicity stunt.

You say they're doing things that are cheap and trivial, but isn't the major innovation of Neuralink the fact that they're using very fine wire probes? No one else has done this and instead used relatively hard fixed arrays of needles. This causes damage to blood vessels among other things. Neuralink's advantage is they exactly void this type of brain damage.
It is not true that "no one else has done this". Do you think neural interface researchers didn't thought of something like that? There are plenty of other electrode designs apart from the "nails board", and none solve the problem I mentioned well enough yet.
Can you mention some? I've never heard of them.
10 & 8 years ago: https://youtu.be/Y4fmMVdDf3s?si=17rh9ydWHEwVcdO2 https://youtu.be/RbnRY3eH3AU?si=lskZuSL5PjsufjDw

6 years ago: https://youtu.be/nM973fDmD6E?si=YY02bo9fHNG_HNXy

11 months ago, but Lacour's lab was already on it 5 years ago: https://youtu.be/Qsvo_O4-amg?si=T1qucWm0vEN9jrNU

Lacour's lecture 8 years ago: https://youtu.be/Qsvo_O4-amg?si=T1qucWm0vEN9jrNU

(She's at the Wyss institute in Geneva)

2 years ago: https://youtu.be/aL7zEsfkibg?si=GkF6ll7rUDHSrcvl

What brain startup has the right approach?
Neuralink has previously discussed their implantations in monkeys, in which they claim implantations for > 1 year with increasing performance (e.g., in their deep dive on youtube [1]). That would seem to indicate that this wasn't seen in animal studies (maybe just a matter of degree). Figure 04 they have in the article suggests a steady state may have been reached with respect to some factors, but I guess they could continue to improve things on the model side while still experiencing degradation if the improvements outpace the effect of degradation. Since this doesn't seem to impact safety, it doesn't seem like a follow-up surgery would be needed as long as the device continues to be useful; if it essentially became a brick then the patient would probably want it removed (which requires another surgery). They've never talked about any surgeries for "maintenance" before, and given the size of the threads, I doubt the same previously implanted ones could be reinserted.

1. Neuralink Show and Tell, Fall 2022 (https://www.youtube.com/watch?v=YreDYmXTYi4)

> 2. Will it degrade further? Are follow-up surgeries needed?

They said in the weeks following and it doesn't seem to have continued to happen. However I would expect things to degrade eventually as this is still very early days.

Why are the electrodes retracting?
> During his first-ever research session, Noland set a new world record for human BCI cursor control of 4.6 BPS. He has subsequently achieved 8.0 BPS and is currently trying to beat scores of the Neuralink engineers using a mouse (~10 BPS).

This is what I was wondering since the original announcement. How does what they achieved compare to previous efforts and to a regular mouse. Pretty impressive I think! He won't be winning any Starcraft tournaments but it seems like a big improvement.

Every time Neuralink comes up, there's a whole lot of speculating in the comments. Is there no real, hard data from them?
There sure isn’t.