\n
![](\"https:\/\/static.scientificamerican.com\/sciam\/cache\/file\/1CD46EE9-93BC-4C05-89A881B0B9F55955_source.gif\")
<\/p>\n [ad_1]
\n<\/p>\n
Sophie Bushwick: <\/strong>Can you tell what a cat is thinking, just from looking at it?\u00a0<\/p>\n Tulika Bose:<\/strong> Probably not! Cats evolved to be solitary hunters stalking their prey, not social animals like us humans.<\/p>\n Bushwick: <\/strong>And that poker face might be handy while you\u2019re out stalking prey, but it\u2019s a real problem if humans are trying to figure out whether their pets are\u00a0 in pain. So that\u2019s why researchers are trying to read more into feline feelings<\/p>\n Bose: <\/strong>\u2014 [sings] using artificial intelligence!<\/p>\n Bushwick:<\/strong> \u2014[sings] using artificial intelligence! As they love to do. I\u2019m Sophie Bushwick, tech editor at Scientific American.\u00a0<\/p>\n Bose: <\/strong>I\u2019m Tulika Bose, multimedia editor.<\/p>\n Bushwick:<\/strong> And you\u2019re listening to Tech Quickly, the version of Scientific American\u2019s Science Quickly podcast that just wants to boop AI on the nose.<\/p>\n Bose:<\/strong> Cute.\u00a0<\/p>\n [INTRO MUSIC]<\/p>\n Bushwick: <\/strong>Right now, if you want to see whether a cat\u2019s in pain, you have to take a quiz called the Glasgow Feline Composite Measure Pain Scale.\u00a0<\/p>\n Bose: <\/strong>Interesting.\u00a0<\/p>\n Bushwick:<\/strong> It\u2019s also just called Glasgow scale.\u00a0<\/p>\n Bose:<\/strong> Ok, hmmm.\u00a0<\/p>\n Bushwick:<\/strong> But The question is, what\u2019s on the scale?<\/p>\n Bose:<\/strong> Yeah, what is on the scale?<\/p>\n Bushwick:<\/strong> Well, I just happen to have printed out a couple copies for us to peruse!\u00a0<\/p>\n Bose:<\/strong> You did not.\u00a0<\/p>\n Bushwick:<\/strong> I did indeed.\u00a0<\/p>\n Bose: <\/strong>Wow<\/p>\n Bushwick: <\/strong>Take a look, take a look at this.<\/p>\n Bose:<\/strong> This looks like a Buzzfeed quiz for cat pain, but uh \u2014\u00a0<\/p>\n Bushwick: <\/strong>Yeah, the least favorite kind of Buzzfeed quiz.\u00a0<\/p>\n Bose: <\/strong>Yeah, the least favorite. Ok, wow. So I see some little cartoon cats here.<\/p>\n Bushwick:<\/strong> Uh huh.<\/p>\n Bose:<\/strong> And I see their little ears squashed, or upright \u2014\u00a0<\/p>\n Bushwick: <\/strong>\u2014 And then you have to say which one is your cat.\u00a0<\/p>\n Bose:<\/strong> Yeah! And then, I see uh, a second little set of cartoon cats here, and their little mouths look to be squished in different ways too.\u00a0<\/p>\n Bushwick: <\/strong>Yeah, so you choose the picture that best matches the cat. And you also answer questions about its behavior \u2014\u00a0<\/p>\n Bose: <\/strong>\u2014 The cat \u2014<\/p>\n Bushwick:<\/strong> How it responds to being pet \u2014<\/p>\n Bose: <\/strong>Ok, ok.<\/p>\n Bushwick:<\/strong> And then each one of those questions gets a little number associated with it. Higher or lower, depending on how much pain it indicates. And then you add it all up at the end to get your pain score.\u00a0<\/p>\n Bose:<\/strong> [Flips page.] This looks kind of hard actually.\u00a0<\/p>\n Bushwick:<\/strong> It\u2019s tricky! I mean can you tell the difference between a cat that is tense \/ crouched \u2014\u00a0<\/p>\n Bose: <\/strong>No!<\/p>\n Bushwick: <\/strong>And one that is rigid \/ hunched?\u00a0<\/p>\n Bose:<\/strong> No that would be like the same thing to me!\u00a0<\/p>\n Bushwick: <\/strong>It sounds exactly the same.\u00a0<\/p>\n Bose:<\/strong> Yeah, exactly!\u00a0<\/p>\n Bushwick:<\/strong> So it\u2019s pretty subjective, you have to make some judgment calls\u2026\u00a0<\/p>\n Bose:<\/strong> mmhmmm\u00a0<\/p>\n Bushwick:<\/strong> And if you\u2019re not a veterinarian, who sees cats and uses this scale all the time, it could be an issue.\u00a0<\/p>\n Bose: <\/strong>Absolutely. More subjective right?\u00a0<\/p>\n Bushwick: <\/strong>Mmhmm exactly.\u00a0\u00a0<\/p>\n Bose: <\/strong>So the idea is that AI might be able to do better?<\/p>\n Bushwick: <\/strong>Or at the very least, it could do it faster and way more easily than an untrained human could.\u00a0\u00a0<\/p>\n Bose: <\/strong>So I know freelance science writer Leila Okahata wrote about a study that put AI to the pain-detection test.\u00a0<\/p>\n Bushwick: <\/strong>That\u2019s right. Researchers started with photos of 84 cats that were being admitted to a veterinary\/animal hospital in Germany \u2014\u00a0<\/p>\n Bose: <\/strong>Interesting \u2014\u00a0<\/p>\n Bushwick:<\/strong> \u2014 and then they gave these cats a pain score based on the Glasgow scale that we talked about \u2014\u00a0<\/p>\n Bose: <\/strong>Right \u2014\u00a0<\/p>\n Bushwick: <\/strong>And also on the amount of pain you\u2019d expect a cat to be in for whatever ailment or injury brought them in, you know like a bone fracture. That\u2019s gonna be pretty painful.\u00a0<\/p>\n Bose:<\/strong> Yowch.<\/p>\n Bushwick:<\/strong> You mean \u201cmeow-ch\u201d?<\/p>\n Bose: <\/strong>Uh, no, no I don\u2019t.<\/p>\n Bushwick: <\/strong>[Laughs]\u00a0<\/p>\n Bose<\/strong>: Ok. diabolical laugh. Diabolical laugh.\u00a0<\/p>\n Bushwick:<\/strong> I\u2019m not laughing at the cat\u2019s pain. I\u2019m laughing at the cat\u2019s pun.\u00a0<\/p>\n Bose: <\/strong>Uh, wow. Did you just make another pun?\u00a0<\/p>\n Bushwick: <\/strong>I can\u2019t help myself.<\/p>\n Bose: <\/strong>Oh my gosh, ok, ok.\u00a0<\/p>\n Bushwick: <\/strong>[Continues laughing.]<\/p>\n Bose: <\/strong>Ok, ok.\u00a0<\/p>\n Bushwick: <\/strong>I\u2019m just littering the script with them! [Continues laughing.]<\/p>\n Bose: <\/strong>Yes you are!\u00a0<\/p>\n Bushwick<\/strong>: [Continues laughing.] Littering? Litterbox?\u00a0<\/p>\n Bose:<\/strong> Oh my God. You did it again.\u00a0<\/p>\n Bushwick:<\/strong> Yeah. Anyway, the photos are labeled with the pain scores, and the researchers fed them into two different machine learning algorithms.\u00a0<\/p>\n Bose: <\/strong>Interesting.\u00a0<\/p>\n Bushwick:<\/strong> One broke down each photo in great detail, by examining four dozen specific sites on the cats\u2019 faces. These are called called landmarks.\u00a0<\/p>\n Bose: <\/strong>Got it.\u00a0<\/p>\n Bushwick: <\/strong>The second AI program was a deep learning model that just looked at each photo as a whole.<\/p>\n Bose:<\/strong> Got it. Moment of truth though \u2014 could the AI read the cats\u2019 faces?<\/p>\n Bushwick:<\/strong> It did pretty well!\u00a0<\/p>\n Bose:<\/strong> Huh!\u00a0<\/p>\n Bushwick:<\/strong> The landmark-based model was 77 percent accurate \u2014<\/p>\n Bose:<\/strong> Interesting \u2014<\/p>\n Bushwick:<\/strong> \u2014 and the deep learning one was a little less good. That was 65 percent accurate. But they think the deep learning one could improve if it has more photos to train on\u2014this particular kind of model is really data hungry.<\/p>\n Bose: <\/strong>It needs more pictures of cats.\u00a0<\/p>\n Bushwick:<\/strong> Exactly. Everyone needs more pictures of cats.\u00a0<\/p>\n Bose: <\/strong>We all do.\u00a0<\/p>\n Bushwick:<\/strong> Including AI.\u00a0<\/p>\n Bose: <\/strong>Right! So can we learn anything from the AI about reading our cats\u2019 faces?<\/p>\n Bushwick: <\/strong>Yeah, that\u2019s one of the cool things about this study! So it\u00a0 turns out that the eyes are not the windows to your cat\u2019s soul. When it comes to accurate pain recognition, you need to look at your cat\u2019s mouth.<\/p>\n Bose: <\/strong>It\u2019s meowth.\u00a0<\/p>\n Bushwick: <\/strong>Aaaaah! You did it!\u00a0<\/p>\n Bose:<\/strong> I did it.\u00a0 Can we go beyond pain here? Like, is there an AI that can tell me if a cat is happy, or sad, or about to knock a glass of water on my laptop?\u00a0<\/p>\n Bushwick:<\/strong> That\u2019s the eventual goal! A researcher who\u2019s not involved in this AI study recently published a paper showing that cats can produce almost three hundred facial expressions.\u00a0<\/p>\n Bose: <\/strong>Yeah like I knew this cat that had this expression when it was going to take a bite of MY sandwich\u2014\u00a0<\/p>\n Bushwick: <\/strong>[laughs] That\u2019s very specific.\u00a0<\/p>\n Bose: <\/strong>Yeah. But I feel like that was its sandwich-eating, sandwich-stealing \u2014\u00a0<\/p>\n Bushwick:<\/strong> Sandwich-eating sandwich steal-ing face. The classic expression.\u00a0<\/p>\n Bose:<\/strong> Yeah! It\u2019s like I\u2019m going to swipe your food, stupid human.\u00a0<\/p>\n Bushwick: <\/strong>[laughs] I mean I feel like all cats have that expression at all times.\u00a0<\/p>\n Bose: <\/strong>Yeah! Just like, hello, dumb human. Feed me. Give me your stuff.\u00a0<\/p>\n Bushwick:<\/strong> [laughs]<\/p>\n Bose: <\/strong>I\u2019m taking your bed.\u00a0<\/p>\n Bushwick: <\/strong>[laughs] \u2014 and your laptop.\u00a0<\/p>\n Bose:<\/strong> \u2014 everything that you own.\u00a0<\/p>\n Bushwick:<\/strong> \u2014 Everything that you own is mine now.\u00a0<\/p>\n Bose: <\/strong>It\u2019s mine.\u00a0<\/p>\n [CLIP] Cat meow<\/p>\n Bushwick: <\/strong>So yeah, once you know that you can recognize these different expressions,\u00a0<\/p>\n Bose:<\/strong> Right.\u00a0<\/p>\n Bushwick: <\/strong>This researcher wants to join forces with the AI researchers to create a model that can detect more specific cat emotions based on whatever face it\u2019s making. The goal is to be able to look at an image of the cat and have the AI detect whatever emotion that cat is feeling.\u00a0<\/p>\n Bose: <\/strong>Got it, got it. Ok \u2014 But like\u00a0 I\u2019ve heard that emotion recognition AI is really flawed though.\u00a0<\/p>\n Bushwick:<\/strong> Oh, for sure. For instance, what do you think I\u2019m feeling now? [makes face]<\/p>\n Bose: <\/strong>Hmmm. [pause] Sad. You look sad.\u00a0<\/p>\n Bushwick:<\/strong> Yeah, Wrong! I\u2019m feeling smug!<\/p>\n Bose:<\/strong> Oh!\u00a0<\/p>\n Bushwick: <\/strong>\u2014 and sneaky, because I just tricked you into thinking I\u2019m feeling an emotion that I\u2019m not.\u00a0<\/p>\n Bose:<\/strong> [laughs.] Wow, a politician.\u00a0<\/p>\n Bushwick: <\/strong>I mean think about it! If I just smile I could be like I\u2019m being a politician, I could be saying \u2018oh I feel happy\u2019 or I could be saying \u2018Im just being polite\u2019 \u2014\u00a0<\/p>\n Bose: <\/strong>Ohhh.\u00a0<\/p>\n Bushwick: <\/strong>Or it could mean, I\u2019m holding in a fart right now!\u00a0<\/p>\n Bose: <\/strong>[Laughs.]<\/p>\n Bushwick:<\/strong> And the AI \u2014 it can\u2019t tell that we\u2019re a bunch of lying liars who lie like that.\u00a0<\/p>\n Bose:<\/strong> Interesting! So the idea is that cats don\u2019t lie about their emotions the way we humans do?<\/p>\n Bushwick: <\/strong>Well, it\u2019s more that with cats and with other pets we have no way to ask them how they feel<\/p>\n Bose:<\/strong> Right.<\/p>\n Bushwick: <\/strong>So even an imperfect tool, like an emotion recognition AI \u2014 it\u2019s valuable because it can help us cross that communication barrier.\u00a0<\/p>\n Bose:<\/strong> Got it.\u00a0<\/p>\n Bushwick:<\/strong> The researchers also want to expand their work to study emotions in other animals.\u00a0<\/p>\n Bose:<\/strong> Right.\u00a0<\/p>\n Bushwick: <\/strong>We\u2019ve also talked on this podcast about using AI to decipher wild animal communication.\u00a0<\/p>\n Bose: <\/strong>Interesting!\u00a0<\/p>\n Bushwick:<\/strong> So we might just be a few datasets away from letting our pets practically \u201ctalk\u201d to us.\u00a0<\/p>\n Bose: <\/strong>Ok that would be so cool!\u00a0<\/p>\n