Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Gabe Newell on brain-computer interface technology

tvnz.co.nz

Post reply on HN
Print view View original

71–80 of 88 posts

Re: Gabe Newell on brain-computer interface technology

#71

Post by PH01 »

mattkrause wrote: post #66
PH01 wrote: post #58

Earlier quoted context omitted.

So, to be clear, the signal of interest does not go above 1kHz. Obviously if you want to sample a signal you have to use a higher frqeuency.

Eh, it depends on what you're calling "the signal of interest." A spike train, the output of a neuron, can reasonably be represented as a binary signal (1=spike, 0=no spike) at 1 kHz. In fact, I'd say this is nearly standard. The neuron's membrane potential, a combination of its current inputs, intrinsic properties, and recent history, changes on faster timescales. Whatever processing the brain does probably does not…

Quite a lot simpler to seperate the mixed signals using a modern multi-electrode array.

Re: Gabe Newell on brain-computer interface technology

#72

Post by johnnybaptist »

andygmb wrote: post #30
johnnybaptist wrote: post #16

I work at OpenBCI. It's been great getting to work with Gabe and the Valve team. Can't overstate how unique they are as partners on a project like this. Also cool to see OpenBCI (sortof) in the top 10 today :) Happy to answer any questions people have about OpenBCI https://openbci.com/

In the area of BCI, what is the most "Science-Fictiony" thing you have seen made into a reality?

I referenced this in my other reply, but this is one of the ones that blew my mind:

https://www.washington.edu/news/2019/07/01/play-a-video-game...

Re: Gabe Newell on brain-computer interface technology

#73

Post by throwsaways212a »

5440 wrote: post #43
5440 wrote: post #40

I do a significant amount of breakthrough device fda/eu regulatory work in the BCI area. For those interested, its worth reading the FDA guidance: https://www.fda.gov/regulatory-information/search-fda-guidan... .

Regarding security, I think people would be really freaked out by the scope of attacks. One infusion pump I worked on this year was succesfully attacked over 200 times.

That sounds interesting, is this a case of 200 vulnerabilities being found or 200 documented cases of attacks against (production?) devices?

Re: Gabe Newell on brain-computer interface technology

#74

Post by m463 »

Geee wrote: post #67

Imagine getting a cookie popup every time you look at something.

https://borderlands.fandom.com/wiki/Adbug

"While technically hostile, Adbugs do not directly attack Vault Hunters. Instead, they follow their targeted Vault Hunter around and beam a semi-transparent advertisement at them, which is placed in a random position on the field of vision, obscuring part of the field of view."

Re: Gabe Newell on brain-computer interface technology

#76

Post by px43 »

offtop5 wrote: post #9

This scares me a bit. Let's say you have certain nero indicators which indicate person X may do Y . Do you we then minority report lock up people. At the same time , it's going to be amazing for people in a coma or unable to communicate. We could have a fantastic amazing world, where we effectively live forever via some type of matrix like interface.

Eh, we'll have software for detecting and correcting criminal intent before the government gets its act together enough to start arresting people preemptively.

Re: Gabe Newell on brain-computer interface technology

#77

Post by Balgair »

PH01 wrote: post #71
mattkrause wrote: post #66

Earlier quoted context omitted.

Eh, it depends on what you're calling "the signal of interest." A spike train, the output of a neuron, can reasonably be represented as a binary signal (1=spike, 0=no spike) at 1 kHz. In fact, I'd say this is nearly standard. The neuron's membrane potential, a combination of its current inputs, intrinsic properties, and recent history, changes on faster timescales. Whatever processing the brain does probably does not…

Quite a lot simpler to seperate the mixed signals using a modern multi-electrode array.

I've written multi-tetrode cluster analysis SW. It is very much not simple. Deconvolving all the inputs is terribly difficult and computationally expensive. Generally we had undergrads as classifiers for the clustering ML, which never really worked all that well, TBH. The higher the sampling frequency, the better the spatiotemporal resolution, but the more data to work through. When you do get a cluster, the overlaps are very close together and it really is hard to say that they are 2 neurons. It takes a lot of data/time to be certain.

One thing to remember is that noise is a huge problem with all this. Not just the 'normal' electronic sources (how do you ground a brain well? It's intentionally evolved to be nothing but loops!), but the neuronal too. Neurons will often fire just because (or not), they will jiggle with animal movement, they will move with heartbeat, and they will die off or move on their own, the electrode dances about, the brain attacks the electrodes, some portion of the electrode snaps off, etc. It is a very hard thing to manage well.

Sorry, bad 'memories' of those projects.

Re: Gabe Newell on brain-computer interface technology

#78

Post by birdbrain »

PH01 wrote: post #71
mattkrause wrote: post #66

Earlier quoted context omitted.

Eh, it depends on what you're calling "the signal of interest." A spike train, the output of a neuron, can reasonably be represented as a binary signal (1=spike, 0=no spike) at 1 kHz. In fact, I'd say this is nearly standard. The neuron's membrane potential, a combination of its current inputs, intrinsic properties, and recent history, changes on faster timescales. Whatever processing the brain does probably does not…

Quite a lot simpler to seperate the mixed signals using a modern multi-electrode array.

If only this were true.

(By "modern" I assume you mean densely-spaced [~10um] multi-pad linear silicon probes, the NeuroPixel being the most celebrated example among the non-specialist public.)

What many pads close together can give you is the potential for the waveform from a spike nearby to register on more than one pad.

In theory this should make spike-sorting easier, because you can distinguish two neurons whose spikes might have the same waveform on pad #1 but different waveforms on pad #2.

In practice, spike sorting improves, but not by as much as you'd think.

Part of this is down to physical factors (which are improving): probe geometry, pad shape, pad material, pad impedance, and so on.

Another part is down to software (which is also improving). Single-channel spike-sorting is by now probably close to as good as it's going to get given the information content of its input, and the algorithms and software to perform it are well-understood and stable.

Algorithmic approaches to multi-channel spike sorting, however, are the subject of active research with multiple promising avenues of progress, and software to perform it is ... well, charitably, let's call it "rough-and-ready." (It's nearly all lab-grown software, which means it's written by enthusiastic amateur programmers [among whom I'd count myself, no shade intended here] who soon move on to new projects because of the structure of academic science. This means the software is buggy, poorly documented, inconsistently supported, and constantly evolving.)

Now, using dense silicon arrays does markedly increase the rate at which I can record well-isolated neurons, but a significant part of this increase is just having more pads in the target brain region - many of the neurons I get from these multi-channel spike-sorting programs only show significant power on one or two pads.

And all of this doesn't even touch on some of the significant challenges that come with using multi-electrode arrays, including higher initial inflammation, later gliosis, and data-collection and storage (a 64-channel array [NB: the NeuroPixel 1 can record from 384 pads] producing 16-bit ints at 20kHz [just about the minimum sampling rate for decent spike-sorting] generates a bit less than 10GB/hour - a volume that real big data people might laugh at, but it sure isn't small!).

And finally, all of this is done off-line. On-line spike-sorting is harder.

So I'm sorry to say that just having a modern multi-electrode array absolutely does not make things "quite a lot simpler."

Re: Gabe Newell on brain-computer interface technology

#79

Post by Scipio_Afri »

tashoecraft wrote: post #61
farias0 wrote: post #36

Earlier quoted context omitted.

What is the timeframe you imagine for us to see the first consumer BCI-based products? Gabe seems to talk a lot about "inserting" data (like feelings) into the brain, instead of reading it. Is the technology really there already? Can we reliably read data from the brain (i.e. using it as input for a digital system)? And regarding inserting, what is the coolest thing you've done that you can share with us?

Considering I own Neurosity's Notion: https://neurosity.co/ , which is a consumer BCI based product, I'd say a couple years ago.

Do you like it? Where is the science backing up it's ability to keep you focused?
Post reply on HN