/sci/ - Science & Math

>>3542425
This talk about the diving scooter is giving me some STRONG flashbacks from Startide Rising.

In it, the scooter had a large-enough canopy so that when you needed to go solo, you could put on a mask and a small rebreather. Of course, it also had several canopies for the use of sapient dolphins when doing work at depth, but that gets a bit off-topic.

>>3273971
The difference is that the brain is only powerful due to its massive parallelism: Neurons, after all, take whole milliseconds to fire. Computers are a lot faster but flat surfaces and single-processors, even this "multi-core" stuff is not true parallelism. The moment we develop parallel computation and 3D printing of microchips is both the moment computing power reaches unthought of levels and the moment heat radiation becomes and actual problem rather than "I'll troll /g/ by frying marshmallows on my NVIDIA".

>>3273913
Hoax. They have the computing power to run a few billion of these little "dots", but neither their properties or their structure resembles that of a cat in any way. Cat uploads are far off.

We have, on the other hand, simulated a nice portion of a mice's brain, their cerebellum, by duplicating the structure and abstracting away some of the behaviour of the neurons, and it showed cerebellum-like behaviour. We'll then upload a whole rat, and I expect it to show extremely rat-like behaviour, ie fucking awesome.
I think mind uploading is not going to be a problem with the brain stuff; scanning technology keep increasing it's resolution and eventually you might as well follow Anders Sandberg's advice: Freeze the brain, laminate it, scan with electron microscope. (Freezing a brain creates cracks and there is no activity to record, but you get the idea). I think the biggest problem of mind uploading will be defeating this brain/body dualism (Rather than mind/body dualism, which is so 20th century): That is, a thought can cause the heart to pump faster, feeding more Oxygen to the brain, pH and hormones can affect the way neurons work, et cetera. You get the idea. Isolating a brain and simulating it might be good enough, might not.

>>3184919

>BACK THEN, we were also, probably obviously, attempting to work out voice recognition, or as we called it utterance transcription. It seemed simple enough — sample the audio, keep up with binning it into smallish segments and converting them to spectral data via FFT (fast Fourier transforms), maybe do overlapping FFTs to autocorrelate out the binning and transient spectral noise, and feed the spectral stream to a higher level processor and set of algorithms that could then characterize the segments of the utterance, determine formants and zeros, and somehow turn all that back into phoneme streams, which at an even higher level could be transcripted through the newly developed combo of fuzzy logic and neural synthetic networks, back into the words that were uttered.

>IT WAS A FAILURE, completely. Not only was the computational oomph insufficient (without costing the Naval Research Laboratory's entire custom-chip design budget), but worse — even when utterances were recorded, then analyzed at 1÷1000TH realtime on fairly robust computers (the assumption being, “if we can sim it, eventually we can chip it”), waveforms didn't decode to nice phoneme streams, and nor did whatever was analyzed have enough information-to-noise to make it through huge neural networks to result in an output of words — and better, the original utterance, transcripted. It was depressing, actually.

I'm getting stronger and stronger David Brin-vibes here.
>humans and dolphins against hostile aliens
>human and dolphin sex
>human-dolphin equality

Gotta read the books again.