Elon Musk has tweeted about a successful implantation of a Neuralink device. Will it prove to be consciously controllable enough to send such a tiny flow of information? For the reference, a touch typist on a qwerty keyboard transmits 10-100 bits per second.
Resolves YES by a peer-reviewed paper, or multiple reputable media sources, demonstrating the capability.
Resolves NO in a year.
Related questions
What specifically do you want to see to have this resolved? Here is an hour long video clearing showing him using the device to do a broad range of things (way above 4 bits).
I do not see how this isn’t enough to resolve. Please be specific:
https://x.com/moddedquad/status/1771298116719002100?s=46&t=w8K1ymlMU0cSwtGlbRCU6Q
@a2bb They are live streaming the patient using it right now. This should be enough yo resolve yes: https://twitter.com/i/broadcasts/1ypJdkXjaLNGW
@GabeGarboden Just looking at the cursor movement, there's definitely more than 4 bits per second there. At about 2m 15s he makes a move in the chess game and them moves the cursor to the bottom of the screen, and then up slightly, within a second or two. That's a click and three mouse movements. Any movement of the mouse must be at least two bits just to identify which of four directions in 2D it is moving in, and there's obviously a lot more control than just linear movement for one second at a time at a fixed speed. So this is many bits per second we're seeing.
@a2bb Neuralink posted multiple video demonstrations of the capability. Most recently an hour long talk with the patient demonstrating him playing Mario cart and breaking many BCI world records.
Your resolution criteria says “OR multiple reputable media sources.” Multiple video demonstrations ARE reputable media source.
Also, just google Neuralink. Almost every major news network (very much reputable media sources) have published articles discussing this.
@a2bb could you clarify which part of the resolution criteria hasn't been met yet? This to me looks like multiple reputable media sources reporting on it:
@chrisjbillington Your screenshot isn't enough. And i couldn't find the evidence that would satisfy the information rate requirement. A few cuts of a couple of seconds worth of moving a cursor in a chess game doesn't seem representative of the capability, https://www.youtube.com/watch?v=5SrpYZum4Nk
https://manifold.markets/a2bb/will-a-human-use-neuralink-to-trans#tsSRYytDAzOA6ICCPowj
The user admits it's not stellar either.
Nevertheless, it's likely a YES resolution will occur soon if the video isn't cherry-picking a lucky success, as we see more evidence in a more speed/accuracy demanding setting than a chess game.
@a2bb Here is the original one hour video with multiple demonstrations, not the video you posted:
https://x.com/moddedquad/status/1771298116719002100?s=46&t=w8K1ymlMU0cSwtGlbRCU6Q
See 15:40 into video for a demonstration of user breaking the past world record
You also did not explain how the screenshot did not show “multiple reputable media sources” because it clearly does..
@a2bb I'm not in too much of a hurry, but you're not really answering the question - what aspect of the resolution criteria haven't been satisfied? Is it that the media sources aren't reliable enough? Is it that I posted a screenshot instead of links and/or quotes? Is that they don't specifically comment on the bitrate? Are you not happy to infer from the video that the bitrate is high enough? In my earlier comment I argued why it had to be - do you need such inference to be done by a third party? Is it just about who makes the claim, or are you actually not happy that the required bitrate has been demonstrated (to me this video demo is impossible without a bitrate higher than 4bps). Please be specific - it's all well and good to say it's not enough, but what specifically about it isn't enough?
Edit: OK, sounds like you want a demo that shows use over a minute or so. I think we could argue if we wanted that the the ~two seconds of use I referred to earlier has at least 120 bits of data in it, which would average to 4bps over a minute by itself. But you want to see more extended use?
What error rate is acceptable for this question to resolve "yes"? If an implant can transmit a 4bps signal but the error rate is 50%, the signal is statistically equivalent to noise and it seems obvious that would not count as "consciously controllable". But what if the error rate is 25%? What if it's 49.999%? What level of statistically significant difference from noise are we looking for?
@pyrylium 4bps error-free average over a meaningful time period (a minute?) after all the applied error corrections would do.
@a2bb I'm not sure, but what comes to mind is being able to control an AI agent with very low latency, over the alternative of the relatively high latency methods of using language and graphics. The team of cyborgs whose brains are plugged into a corporate AI are going to have an advantage over the dinosaurs who are merely using screens and keyboards, and their tongues, to achieve something worse a hundred times more slowly.
