←back to thread

360 points danielmorozoff | 1 comments | | HN request time: 0.205s | source
Show context
arnaudsm ◴[] No.45033273[source]
[flagged]
replies(19): >>45033369 #>>45033390 #>>45033472 #>>45033493 #>>45033528 #>>45033691 #>>45033977 #>>45034012 #>>45034276 #>>45034628 #>>45034780 #>>45034920 #>>45035070 #>>45035284 #>>45035406 #>>45035412 #>>45035600 #>>45035609 #>>45035620 #
modeless ◴[] No.45033691[source]
Classic HN middlebrow dismissal, only upvoted because people dislike Musk. Word error rate of 25% is unusable. Also it needs extensive retraining every few days. They used four fixed electrode arrays, like pushing a miniature bed of nails into your brain, which is far more invasive and less advanced than Neuralink's one device with threads individually implanted by robot. Neuralink is not exclusively for speech, focusing more on general computer use. This is mostly about where the device is implanted, not the device's capabilities.

As for FSD, it leads by far for systems you can own, and while it is not as good as Waymo it is much cheaper and still rapidly improving. It is too early to say which approach will ultimately win.

replies(4): >>45033727 #>>45034073 #>>45034258 #>>45035246 #
arnaudsm ◴[] No.45033727[source]
The 25% error rate is for 100k+ vocabulary, it drops under 10% for smaller vocab. Meaningful bits per minute would be a better metric indeed.

I appreciate comments like yours that actually contribute to the debate. We need critical thinking and data. Not one-sided puff pieces out of context.

replies(1): >>45033745 #
1. modeless ◴[] No.45033745[source]
9.1% word error rate on a 50-word vocabulary is not that great either.