Linguist Forum

Specializations => Psycholinguistics => Topic started by: Corybobory on January 06, 2014, 04:54:12 AM

Title: A device that lets dogs speak - "No More Woof"
Post by: Corybobory on January 06, 2014, 04:54:12 AM
I read about this on Language Log and looked into it a bit.  "No More Woof" is an small EEG device that reads dogs brain signals and transforms them into vocalizations through a little computer and speaker on the device.

The description on their website (www.nomorewoof.com) says:

"The first device to translate animal thoughts into human language. No More Woof is a small gadget that uses the latest technology in micro computing and EEG to analyze animal thought patterns and spell them out in Human Language using a loudspeaker."

Here's a video by the designers showing a bit about the product:

http://www.youtube.com/watch?v=CweAeshjObA&feature=youtu.be

What do you think?  It's not really 'dog language' in that it's unintentional thought process that human technology is doing the interpreting of - but that doesn't detract from how amazingly cool it is!!! How about one for babies?  Or one for cats, complete with swearing?

They are still in the first stage of their research, so they are encouraging people to buy one to be part of the research process so they can improve later models.

I want one though.  And a dog...
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Daniel on January 06, 2014, 05:01:36 AM
There's an inherent flaw: the translations are based on our observations. So the device, at best, creates accurate translations of what we'd believe the dog is intending to say. This is a major communication barrier.

But it's certainly fun. Maybe more? Certainly it could be a step on the way to more.


Edit: thinking about it a bit more, the problem is that it isn't two-way communication. I wonder if it will ever evolve to allow anything like a conversation. One very odd thing is that it's reading the dogs thoughts, not "speech".

Also, does it intend to stop barking? If so, how? I mean, will dogs magically stop barking because they understand that they're making human noises? Seems odd. I guess that idea is that it clarifies why they are barking so you can solve the problem faster. ...assuming it's accurate and that you want to solve the problem at that moment. Do many dog owners really have a problem understanding, for example, that the dog wants attention? I'd imagine not.

I wonder what Con Slobodchikoff (the prairie dog guy) would have to say about something like this.
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: zaba on January 23, 2014, 03:06:19 AM
There is nothing serious about this study. It certainly doesn't know what dogs think. It's nonsense!
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Corybobory on January 23, 2014, 04:02:14 AM
^Really?  Why wouldn't it work?

If the eeg device is picking up on repeated patterns, and it could for certain emotions or thought processes, it could surely alert us with a vocal recording when the pattern is made.

It's the same logic used in a prosthetic limb which is informed by brain signals for how to move.
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: jkpate on January 23, 2014, 05:43:20 AM
At least in human EEG and MEG studies, electrical activity from muscle movements (such as eye movements) tends to swamp brain activity recordings of interest, and signals are subjected to post-processing algorithms to remove these artifacts before analysis. The same issue should arise with ear movements and eye movements in dogs, but the pictures show electrodes right on top of the ears and right next to the eyes, and they are marketing devices with one or two electrodes. I don't think (although I'm not absolutely positive) that the current artifact rejection technologies can work with such a small number of sensors that are right on top of the muscles. I suspect that the general patterns they say they have found are just indicators of when the ears or eyes have moved.
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: zaba on January 23, 2014, 07:09:17 AM
I agree with JKPATE.

Quote
If the eeg device is picking up on repeated patterns, and it could for certain emotions or thought processes, it could surely alert us with a vocal recording when the pattern is made.

Look, the device picks up repetitions which are a subset of the activity going on in a tremendously complex action -- but let's even forget about the meaning of EEG devices (which are in no way are capable of telling about what goes on in someone's thoughts, much less the thoughts of a dog) -- from the perspective of pure logic only, there is no way for us to assign meaning to those patterns. Period.

This is marketing, plain and simple. It's a joke (and it's supposed to be!)
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: freknu on January 23, 2014, 07:37:11 AM
The research in "mind reading" can be quite freaky (Reconstructing Visual Experiences from Brain Activity Evoked by Natural Movies (http://www.cell.com/current-biology/retrieve/pii/S0960982211009377?script=true)) but I doubt only a few electrodes are capable of accurately performing such a task.
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Daniel on January 23, 2014, 07:42:43 AM
There's also a question of precision: it may only assume something like 10 contrastive "thoughts" for dogs, in which case this is less crazy though still likely has some problems.
Jkpate, I agree, but maybe ear movements are enough in this case for a few basic "thoughts".
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Corybobory on January 23, 2014, 11:58:56 AM
The one with only 2 electrodes is the $40 cheap one, that is marketed to raise money for this prototype - sure, more nodes = more accuracy in picking up patterns.

But this isn't the only machine working on a brain-computer interface, like I said there are prosthetic limbs that run on the same idea, and through eeg as well.  I don't seem to understand why it's unfeasable?
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: lx on January 23, 2014, 12:49:56 PM
Quote
But this isn't the only machine working on a brain-computer interface, like I said there are prosthetic limbs that run on the same idea, and through eeg as well.  I don't seem to understand why it's unfeasable?

Neither do I.

I mean, you can fly a toy helicopter based on brain waves with a set of pretty ordinary-looking headphones, land it and everything. I know the science behind that is based on attention and meditation levels, but if you can get that and fly a toy helicopter with it for around £100, I don't see the big leap to why this sort of invention is still considered something way off in the depths of scientific future. Okay, it wouldn't be able to have any deep meaning and it might be based on quite crude assumptions, but maybe the brain structure of animals without higher cognitive processes is easier to discern, with fewer natural states and less messy intereferences (nb: I said 'maybe'). Okay, hunger is a process deep within the brain that probably isn't able to be detected by anything on the commercial market, but while I'm not advocating that this works (only read about it for the first time when Cory posted the link), I'm not left with a sense of disbelief (nb: I haven't read the claims of what it says it can say) but the general idea of basic-level attentional states is already on the market to fly toys for humans. I think if you've selected reliably-specific surface brain activity in animals, then at least for those feelings, I am not left with a strong taste of disbelief in my mouth at the idea of this being available (not specifically talking about this product itself).
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Daniel on January 23, 2014, 01:23:41 PM
The issue, I think, is that flying a toy helicopter is NOT language. That's what is misleading about this whole thing. Language, which is what this claims/hints at being, is way beyond the current state.

If you read some research on ERPs for language, you'll see just how extremely sensitive they are-- if you blink then the entire set of data must be thrown out!
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: lx on January 23, 2014, 01:26:07 PM
Quote
The issue, I think, is that flying a toy helicopter is NOT language. That's what is misleading about this whole thing.

But when was this about language? Isn't this about processing the feelings of dogs and then alerting their owners via language as to what they're feeling? I mean, it could be: - one bleep for happy - two bleeps for sad. Language is not the central issue to this discussion, unless I've really misinterpreted what we're talking about here (and if so: woops, my bad  :P).
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Corybobory on January 23, 2014, 02:20:27 PM
This is how I am understand it too, Alex, that it's about processing signals and alerting owners via language (with a cute speaker where you can choose the pre-recorded voices).  It's not 'letting a dog's inner voice out' or anything, it's just finding patterns and associations and mapping them onto the real world, and announcing it through an audible medium (a hilarious audible medium I'm sure :))
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: jkpate on January 23, 2014, 03:48:23 PM
I don't know, even if the idea in general is feasible, this group does not strike me as pursuing it in a serious way.

Also, the one sensor version is being advertised for $65, and the two sensor version is being advertised for $300. They don't say how many sensors the $1200 version will have, but the pictures don't have more than six sensors. (and two of the sensors are down around the dog's jaw!?)
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Corybobory on January 23, 2014, 04:00:44 PM
^perhaps recording salivation?

I participated in an EEG study with less than 10 sensors placed on my scalp - to monitor if after repeat trials of imagining the same hand movement, the patterns could be recognized.  It seemed enough to the researchers!

From the site:
"Right now we are only scraping the surface of possibilities; the project is only in its cradle. And to be completely honest, the first version will be quite rudimentary. But hey, the first computer was pretty crappy too."

They're not lying about its start up-complexity - but by selling the prototypes, they are able to fund further research and refinement, and get a bit group of beta testers for the products.
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Daniel on January 23, 2014, 09:41:57 PM
Quote from: Cory
It's not 'letting a dog's inner voice out' or anything
Cory and lx, you may have a point. But that's also why this is intriguing, you see. If it's just a button that the dog pushes with its ears (ok, not quite) that then says a word, it's not very interesting. It might be useful, but it doesn't have all of those implications they hint at.

There was a study out of UC Berkeley about vision that is relevant here (actually Freknu mentioned some of that research earlier-- I think it's the same project anyway). They played hundreds of youtube videos to subjects in an fMRI, watching brain activity. Then through statistics on that activity, they found patterns and began to learn what responded to what input. Fine. Then they were able to "see" it just from the brain activity! They actually learned how our brains hold images. But... wait. It's not quite that impressive. They were simply modeling activity and mapping it to similar activity, with very similar stimuli. That is, they didn't ask the subjects to imagine one of their favorite videos. It wasn't reading their mind. It was simply intercepting information being used by the visual neurons. Finally, it was only doing that based on statistical reasoning based on the samples and then blending the original input videos for a best fit approximation of the current input.
In short, while it seems very cool and might go to a next step, these technologies are very far from mind reading in any conscious sense-- they can see activity certainly, though.
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: zaba on January 24, 2014, 02:30:21 AM
Quote
Finally, it was only doing that based on statistical reasoning based on the samples and then blending the original input videos for a best fit approximation of the current input.

Bingo. That's (one reason) why its gobbledygook.
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: lx on January 24, 2014, 02:34:57 AM
Quote
Finally, it was only doing that based on statistical reasoning based on the samples and then blending the original input videos for a best fit approximation of the current input.

Bingo. That's (one reason) why its gobbledygook.


Daniel is not talking about No More Woof there, he's talking about the UC Berkeley study on humans. You do realise that, don't you?
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Daniel on January 24, 2014, 09:47:18 AM
In a way, it's similar: is this really a first step toward deeper understanding of the (human/canine) mind if it's just surface-level to start? Certainly it could be, but it's not a rough approximation of deep thought currently. It's a rough approximation of surface-level brain activity.

As Cory said, measuring salivation could quite literally also be useful, but it doesn't seem to get very deep into the mind of a dog.
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: lx on January 24, 2014, 09:58:15 AM
Quote
Certainly it could be, but it's not a rough approximation of deep thought currently. It's a rough approximation of surface-level brain activity.
Are we even comparing the same idea?  :o
You're reading some deep, deep complicated claims into their proposal, that I just can't see anywhere, and then (by that standard) quite rightly throwing those claims out of the window because they don't stand up to modern evidence/reality right now. What I see, is them saying over and over again that it's not a finished product, they're testing, they need resources and their plan is to increase it once they have more research and funding under their belt. Even with a banner along the bottom, "THIS IS NOT A FINISHED PRODUCT," then they say by helping out with the first edition, the main point is not that you get a product that does all this magical stuff, but rather you're allowing them to further the research.

I really think that the whole idea of what they're aiming for in their proposal video has been completely blown out of proportion here. Nowhere do they even get anywhere close to saying anything about translating deep thought. The closest thing is the opening line, which anyone in the real world knows is always a bit of an exaggeration to keep the interest of the viewer. It's their end goal.
Quote
is this really a first step toward deeper understanding of the (human/canine) mind if it's just surface-level to start?
Absolutely! Why would it not be? How often is it that new research instantly is at advanced stages? Isn't it the most natural thing from a scientific viewpoint to start off at a basic level like this?

I have my own doubts about this being possible, and I'm not saying I believe they will ever get anywhere close to achieving it but based on the current proposal, from a clean academic perspective, it looks like how a lot of scientific research goes on. Basic test, demonstrate end desire, basic product, get more funding, develop and hone technology/research towards end goal. They might not ever make the breakthrough but I don't believe anyone should be slamming them down for not delivering something they never implied they could deliver right now.

Besides, I think it'd be really creepy to know what my pets were thinking.  ;)
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Daniel on January 24, 2014, 11:15:15 AM
Hmm... fair enough. The question, as I see it, is whether that's a direction to actually get farther or just improve the current model. The accuracy might be 85% right now; if that was improved to 100% it would still not do "language" or anything like that. The question then is whether it could in theory do language if it were improved more. I'm not seeing that it could.

And, to be honest then, that's why this would be interesting. If not, it's just amusing and not too connected to "language", which is fine of course.

Quote
"Right now we are only scraping the surface of possibilities; the project is only in its cradle. And to be completely honest, the first version will be quite rudimentary. But hey, the first computer was pretty crappy too."
That's what I'm really responding to. I don't mind that it's in its infancy. But they imply that their invention has as improved a future as the computer. I'm not convinced. I think it may end at the point of doing a good job with surface-level brain monitoring, never getting anywhere near the thoughts of animals. If so, it might still be useful, but it's not very impressive as "mind reading" or anything like that.

Anyway, that's just the scientist in me talking. It's fun! I would be amused by animals talking to me (however accurate/precise it was).

Quote
Besides, I think it'd be really creepy to know what my pets were thinking.
Haha. Perhaps!
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: lx on January 24, 2014, 11:44:27 AM
Quote
And, to be honest then, that's why this would be interesting. If not, it's just amusing and not too connected to "language", which is fine of course.
Yeah, I agree. I didn't really link the claim of, "We'll let you know what your dog is thinking," too much with the fact that a system of language might be there. Well, I don't believe dogs have brains that have language in the way that humans did so maybe that's why it never even registered there was ever an attempt to formally develop a system that processed thoughts as we might expect as humans. That initial first connection with the principal of human language being brought out in dogs would only be relevant to the discussion if there was a presumption that dogs processed thoughts/language in the way humans do, and this device would allow for the "missing component" or something like that. That's not something you believe, is it? That's just how I saw it. It was always about registering emotions driven by primal behaviour and detecting what feelings/emotions were connected to what brain activity. I agree it might not be interesting as a study of language, but again, that's not what I think they're selling. They're floating the idea of being able to communicate what the dog wants/feels to pet owners, which is something that would make an absolute fortune and there would be an incredibly lucrative market for that. Let's not forget we were the ones that brought this into the discussion in a linguistics forum. We were the ones that started labelling this project on these terms - not them. So, it might not be interesting for language, but that wasn't the reason why we were talking about it in the first place. They've got an idea that, if brought to fruition, would be in incredible demand (in my opinion).
Quote
Anyway, that's just the scientist in me talking. It's fun! I would be amused by animals talking to me (however accurate/precise it was).
Exactly, me too. I'd pay for something to put on a dog and go round insulting everyone in the house, when it detected the dog was making eye contact or something.  ;D
Title: Re: A device that lets dogs speak - "No More Woof"
Post by: Daniel on January 24, 2014, 12:08:04 PM
Fair enough :)

Quote
That initial first connection with the principal of human language being brought out in dogs would only be relevant to the discussion if there was a presumption that dogs processed thoughts/language in the way humans do, and this device would allow for the "missing component" or something like that. That's not something you believe, is it?
I believe something between the two extremes. I do believe that dogs think/"speak" in a way that we don't fully understand, and I also believe that they are not like (or as complex as) humans. I don't see a sharp line distinguishing the two, but rather than dogs don't have as complex systems as humans.

So while I don't imagine a device like this (or any device) would make a dog interesting to talk to about existential philosophy, I wouldn't find it inconceivable that we could reach some level of language/communication, perhaps along the lines of the two word stage in young children or the level of sign language achieved by apes.

As an example, it would be great if this device could be used on bloodhounds and other similar dogs and tell us what kinds of smells they were perceiving, what direction the target was, how far, how certain they were, etc.