It's worth noting that this isn't new technology. This paper is specifically about how their new technique provides a small but statistically significant improvement on existing techniques.
The fact that they provide code and dataset is really praiseworthy.
I still think of this video often and wonder if it is building on any of that technology, almost 10 years old now. Just looking at this whitepaper it seems like they both use some kind of infrared transcranial light, but never imagined the machine in the original iteration was so big
[Regina Dugan's Keynote at Facebook F8 2017 | Inverse]
https://www.youtube.com/watch?v=kCDWKdmwhUI
So attended an interesting talk a couple years ago:
- fMRI and/or brain implants are the best to figure out brain waves
- but they are expensive or invasive
- EEG is a lot cheaper and easier but not as precise
- BUT what if you used LLMs to analyze EEG data taken at the same time as brain implants etc
The answer seemed to be that "yes, you can get better than traditional EEG data using EEG + LLMs". Curious to see where this ends up and hopefully not that like that Black Mirror episode with the brain scanning leading to murders.
If these systems ever become good enough to be useful, the privacy and consent questions need to be treated as core engineering requirements, not as a policy afterthought
Please see my bio for the full rant. The key take-away is:
> While we missed the boat on Internet tracking, there is still time to avoid sailing through the final frontier of neural tracking.
> Thanks to the BCI, we will soon be offered the trade of our privacy for the convenience of password-free login and faster typing. Next, there will be a quick TSA neural scan prior to boarding...
Yes, Meta focusing on this type of research does set off alarm bells for me. I suspect this isn’t about some altruistic intention to help people with disabilities but something a little more surveillance oriented.
Imagine the ad profile data value that this would unlock. CPC - > Cost per thought?
While allowing the user to type with their mind, we detected signs of depression. Full court press on all ad partners aiming for that group.
While allowing the user to type with their mind, we detected that this user pictures driving fast, our insurance partners have the user profile updated.
While allowing the user to type with their mind, we detected signs of <insert any fleeting thought that you would never act upon> - as required by local regulations, they are now labeled by the state as <that>.
It's good point! However, one of the main benefits of a technology like this, would be not really for everyday people, but for people with handicap or a speech impediment.
I personally have a stammer. While mine is less severe, and I doesn't need directly it, I know several people that would quite be glad of the benefits that it could bring to them. (Example: pass online interviews).
I agree however of the privacy concerns. We could limit it in a first time to medical devices for example, or have some privacy laws in place.
I totally respect that, and it would be really cool technology in general. However, it is truly last frontier of privacy. We can't just treat this as yet another little thing.
I ain't young. I have seen this all play out before. It's just extrapolating based on what we did with cookies, browsing history, cell tower data, etc... unless we pass very strict neural privacy laws, why wouldn't it go the same way?
The decent news is that if you search "neural privacy laws," you will find some states are already on top of it, a bit. We need national laws, in every country ASAP. This needs to happen before there are billions in economic inertia behind BCIs.
I’m trying to not be snarky here, but it is difficult to have an optimistic take about an advertising company working on mind reading technology. At some point we have to call a spade a spade, I think.
at this point i feel like its just painfully obvious that if this for whatever became feasible/widespread/scalable, it would absolutely be implemented by the powers that be and willed into fruition by gaslighting, fear mongering and misinformation. i dont know how much more proof this world needs that ruling classes can and do want to just subjugate the hell out of everyone & enforce their world view on the world around them
While I share your privacy concerns, one should also ask how realistic such a scenario is with non-invasive BCIs due to their limited nature. It does not mean that neural tracking isn't possible with this technology, however, I would question if this technology is feasible for consumer products and wide adaptation.
I'm not sure whether to be worried or not and I am not talking about whatever you wrote but for your own sake , it seems to be extremely paranoid style of writing
Mark always been framed/stated for stealing user's data or invasing user's privacy but apart from these, this guy has always been one step ahead in making new research, technology possible by experimenting new things. First with the AR/VR which didn't work I guess so he pivoted from that to Rayban glasses, and now this.
The "Metaverse" bet was a bold but lost bet, in hindsight. Mark is fortunate that the Facebook empire is still printing a lot of money to burn, so he could switch his focus to AI now that it's clear AI is the real thing for this decade. This brain-wave thing is a very small wrinkle in the big picture, not really a big strategy shift.
As far as I understand the underlying model is not multimodal. Maybe a quite naive question but can't we improve the performance by a joint embedding of EEG and MEG data? If it scales with log-linear data, maybe it would also improve with other data as well.
Interesting -- really excited for the future of human-brain interfaces and just in general more interface exploration enabled by large transformers. I'm already very excited by voice, although wish I could get something akin to the subvoc common in scifi novels. Seems like it would be an easier path than human-brain and would allow me to use voice models in public.
As an aside, disappointed by the very low quality of comments on this article here.
I imagine you’re wanting instead something the size of a hat, but smaller fMRI are certainly in development, for example [0] with discussion about the size at [1].
Not my field, but since they use SQUIDs I suspect most of the volume is taken up by cryo equipment. So, maybe if someone eventually discovers a room temperature superconductor.
Reminds me of the scene in series Incorporated where a megacorp uses this sort of tech to interrogate an employee from a competitor mega-corp to get at a trade secret.
It's a little spooky how real that could now be. Oh and that series was a dystopian series because ofc it was
The ultrasound technique there is more like MRI, static imaging, not measuring dynamic electrical signals. Also, regardless of static vs moving, all of this hinges on massive relatively expensive devices (ultrasound is never going to be dirt cheap or miniaturized compared to a smartwatch or even a VR headset) and for the subject to remain perfectly still and probably go through regular calibration sequences. All for <75% accuracy on simple classification tasks. The information mixes in the propagation out of the skull, erasing information. It's analogous to trying to do a row hammer attack on a CPU from outside the computer case.
Also, in the Meta result here, "while actively typing" is actually quite different than passive mind reading because the motor cortex sits nicely near the top surface of the skull, and the muscle memory from past typing makes for a nice well formed signal to measure and classify. It's the same trick over and over since the Brown BrainGate days where you can have people perform or imagine movements and get a decent but not good classification result, and it never gets much better after that trick is exploited. Project dies, VCs and grant writers forget or never appreciated the effect, time goes by, a grad student or corporate research lackey rediscovers it, media puts out an article claiming mind reading is here, and the cycle repeats...
This wouldn’t work. Apart from it being very difficult to get people to sleep deeply enough for them to dream in machines like this, the brain state is very different whilst asleep compared with awake. Also the data generated from typing would be very different than thought since it’s likely picking up on broad electrical activity in the primary motor cortex.
Source: spouse works in a sleep lab studying dreams with MRI
If one were to go about translating brain waves from dogs to meaning, we'd run into a big problem immediately: vocabulary resolution.
What I mean by that is we'll have a very limited number of words to which a dog's brainwaves can be translated to since we aren't able to understand them beyond their basic instincts of food, survival, fear, affection towards their owner etc.
There is just no way to go past what we have already observed by their behavior since dogs can't talk or write.
I do wonder how animals think. Perhaps this resolution would also be the theoretical maximum?
>There is just no way to go past what we have already observed by their behavior since dogs can't talk or write.
There are many dogs that have been trained to press buttons corresponding to words, in the extreme case tens/hundreds of buttons/words, and they can even construct rudimentary sentences. It doesn't seem insane to me that we could perhaps do a very rudimentary version of this for dogs, given a large enough training set.
The "vocabulary resolution being low" basically just means within our own limited context, it's low. But that doesn't mean it's a good measure. Heck, I'd say it isn't.
2035: every phone comes with one so you can can do things without clicking any yucky buttons
2036: China mandates phone manufacturers to capture and submit brain data to the state to protect the children and promote social harmony. EU and US condemn. the media condemns.
2037: the EU mandates phone manufacturers to capture and submit brain data to the state to protect the children and fight malinformation. the media applauds.
2038: the US, ruled by the blue party, mandates phone manufacturers to capture and submit brain data to the state to protect the children and combat white nationalism. the red party condemns.
2039: the US, now ruled by the red party, abolishes the previous law and introduces a new one, mandating phone manufacturers to capture and submit brain data to the state to protect the children and combat illegal immigration. the blue party condemns.
To take some specific piece of research going on in one corner of $BigCorp, and tar it with the brush of general sentiment about $BigCorp, is a classic example of generic tangent: where the topic goes from something specific-and-more-interesting, to something generic-and-more-indignant. We've learned over the years that this is exactly the wrong direction for HN threads to head in. Note this, from https://news.ycombinator.com/newsguidelines.html:
It's not a 'general sentiment' or tangent though, it's specifically THE technology that has the potential to make privacy nonexistent.
Funded by the company that in the recent past, got exposed trying to track employees computer interactions and grabbing screens. Not long before that, it was leaked that employees and contractors were watching videos recorded from meta/rayban smartglasses.
I'm a little bit baffled by the defense here. The point of having guidelines is to use your judgement when applying them.
I'm sorry but I have to disagree - I think that's quite a stretch, and an example of how small-specific topics tend to get sucked into the gravity field of big-generic ones, leading to less interesting discussion (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...).
They are already paying scarce labor like medical doctors and lawyers hundreds of dollars an hour to create training data. The RoI for training data is high because it can be used to train many models.
I hate this because we have crazy billionaires who want to abuse this technology. But apart from tht it's pretty cool, though I hate that it's being developed in this dark day and age. Miserable times.
The fact that they provide code and dataset is really praiseworthy.
- fMRI and/or brain implants are the best to figure out brain waves
- but they are expensive or invasive
- EEG is a lot cheaper and easier but not as precise
- BUT what if you used LLMs to analyze EEG data taken at the same time as brain implants etc
The answer seemed to be that "yes, you can get better than traditional EEG data using EEG + LLMs". Curious to see where this ends up and hopefully not that like that Black Mirror episode with the brain scanning leading to murders.
Not yet. There are plenty of transformer based models out there for EEG but they so far do not outperform the SOTA “traditional” (ML/DL) models.
> While we missed the boat on Internet tracking, there is still time to avoid sailing through the final frontier of neural tracking.
> Thanks to the BCI, we will soon be offered the trade of our privacy for the convenience of password-free login and faster typing. Next, there will be a quick TSA neural scan prior to boarding...
While allowing the user to type with their mind, we detected signs of depression. Full court press on all ad partners aiming for that group.
While allowing the user to type with their mind, we detected that this user pictures driving fast, our insurance partners have the user profile updated.
While allowing the user to type with their mind, we detected signs of <insert any fleeting thought that you would never act upon> - as required by local regulations, they are now labeled by the state as <that>.
I personally have a stammer. While mine is less severe, and I doesn't need directly it, I know several people that would quite be glad of the benefits that it could bring to them. (Example: pass online interviews).
I agree however of the privacy concerns. We could limit it in a first time to medical devices for example, or have some privacy laws in place.
The decent news is that if you search "neural privacy laws," you will find some states are already on top of it, a bit. We need national laws, in every country ASAP. This needs to happen before there are billions in economic inertia behind BCIs.
Something can have "good takes" but still run an unacceptable risk of ending badly
I think anyone in 1995, who was accurately predicting our current state of privacy would have sounded like a paranoid lunatic as well.
As an aside, disappointed by the very low quality of comments on this article here.
[0] https://med.umn.edu/news/university-minnesota-developing-com...
[1] https://designawards.core77.com/Strategy-Research/95329/A-ne...
How realistic would it be to make a smaller device?
It's a little spooky how real that could now be. Oh and that series was a dystopian series because ofc it was
https://news.ycombinator.com/item?id=48685558
Also, in the Meta result here, "while actively typing" is actually quite different than passive mind reading because the motor cortex sits nicely near the top surface of the skull, and the muscle memory from past typing makes for a nice well formed signal to measure and classify. It's the same trick over and over since the Brown BrainGate days where you can have people perform or imagine movements and get a decent but not good classification result, and it never gets much better after that trick is exploited. Project dies, VCs and grant writers forget or never appreciated the effect, time goes by, a grad student or corporate research lackey rediscovers it, media puts out an article claiming mind reading is here, and the cycle repeats...
Source: spouse works in a sleep lab studying dreams with MRI
- Let's go see what's on the other side of this door, friend, maybe there's food !!
Ok, friend, here you go. - opens door.
- Wow super cool, now let's go see what's on the other side of this door, friend, maybe there's food !!
What I mean by that is we'll have a very limited number of words to which a dog's brainwaves can be translated to since we aren't able to understand them beyond their basic instincts of food, survival, fear, affection towards their owner etc.
There is just no way to go past what we have already observed by their behavior since dogs can't talk or write.
I do wonder how animals think. Perhaps this resolution would also be the theoretical maximum?
There are many dogs that have been trained to press buttons corresponding to words, in the extreme case tens/hundreds of buttons/words, and they can even construct rudimentary sentences. It doesn't seem insane to me that we could perhaps do a very rudimentary version of this for dogs, given a large enough training set.
Might be of use.
there's a movie about that default human hubris, it spoils it though
The "vocabulary resolution being low" basically just means within our own limited context, it's low. But that doesn't mean it's a good measure. Heck, I'd say it isn't.
p.s. come on you guys - this is not what HN is for. You may not owe $megacorp better but you owe this community better if you're participating here.
When the bad guys try they just get the lyrics to Yoko Ono music.
2035: every phone comes with one so you can can do things without clicking any yucky buttons
2036: China mandates phone manufacturers to capture and submit brain data to the state to protect the children and promote social harmony. EU and US condemn. the media condemns.
2037: the EU mandates phone manufacturers to capture and submit brain data to the state to protect the children and fight malinformation. the media applauds.
2038: the US, ruled by the blue party, mandates phone manufacturers to capture and submit brain data to the state to protect the children and combat white nationalism. the red party condemns.
2039: the US, now ruled by the red party, abolishes the previous law and introduces a new one, mandating phone manufacturers to capture and submit brain data to the state to protect the children and combat illegal immigration. the blue party condemns.
Meta has shown remarkable disregard to users' and employees' privacy.
Why should that not come up when discussing an entirely new dystopian technology that allows them to invade privacy at scale?
To take some specific piece of research going on in one corner of $BigCorp, and tar it with the brush of general sentiment about $BigCorp, is a classic example of generic tangent: where the topic goes from something specific-and-more-interesting, to something generic-and-more-indignant. We've learned over the years that this is exactly the wrong direction for HN threads to head in. Note this, from https://news.ycombinator.com/newsguidelines.html:
"Eschew flamebait. Avoid generic tangents."
Funded by the company that in the recent past, got exposed trying to track employees computer interactions and grabbing screens. Not long before that, it was leaked that employees and contractors were watching videos recorded from meta/rayban smartglasses.
I'm a little bit baffled by the defense here. The point of having guidelines is to use your judgement when applying them.
Not only that but the less-interesting discussion usually turns more-indignant, which makes sense once the discussion abandons the actual content of an article (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&sor...).
https://news.ycombinator.com/newsguidelines.html
Better than text-stripping the internet - this thing will soon be pulling the logits as well.
"Don't be snarky."
https://news.ycombinator.com/newsguidelines.html
1. The post was obviously bullish / optimistic on the technical capabilities. Not in the least dismissive.
2. The economics extrapolation is obvious. See current precedent for paid access for purchased screen-casts of dev work: https://pdoom.org/open_calls/04_crowd_cast.html
they'll just put it buried on page 450 of the meta glasses 3 or something
There are no layoffs in Ba Sing Se.