00:00Are you ready?
00:01No.
00:02No.
00:04Please no.
00:05Please no.
00:07Yes.
00:08Yes.
00:10Yes.
00:12Is that a camera?
00:20Oh.
00:36What?
00:37No.
00:43No.
00:57Yes.
00:58No, no.
00:59Come on.
01:00No.
01:00I'm going to go to the next one.
01:54Yeah, when the testing is happening, I check everything is fine and any stream is my benefit.
02:06So, you will see that once she goes and eats, there is a different kind of signal.
02:15That's why we said, if you want to give all the answers, I'll give you an email.
02:21Today's session.
02:23Even though there is one who wants that, you are not a scientist, you are a teacher person.
02:42You are a teacher person.
02:47You are a teacher person.
02:49You are a teacher person.
03:05You are a teacher person.
03:07There's a teacher person.
03:08You are a teacher person.
03:13You are a teacher person.
03:30It's a very good place to see the world's best.
03:35The world's best.
03:35The world's best.
03:37The world's best.
04:08So, let's get started.
04:31My foundation, the literature exists in multiple journals, there have been about 50 peer-reviewed
04:38papers on how dogs can sniff out cancers, COVID-19, other kinds of disorders like Parkinson's,
04:45engineering problem. And what we did is we basically built sensors that sit on the dogs.
04:50It's a helmet and a harness that is fully non-invasive. It's like a doggy Iron Man suit,
04:55which measures their brain activity, their respiration patterns, their body language.
05:00And then we have machine learning models, which can take that data and predict whether the dog
05:04is smelling a certain odor, like cancer, and then present just like humans. They can have bad days,
05:11they can, you know, perform differently on one day versus another day. And then how do you scale this,
05:16right? How do you have a small team of dogs sniff out hundreds of thousands of samples?
05:21Both of these problems aren't actually about the dog's nose. So a dog sniffs at a frequency of five
05:26times a second. In an hour, they can do thousands of sniffs. Each sniff can actually evaluate a sample.
05:32So it's never been about performance. So the dog learns the machines. Basically it learns,
05:37okay, every time I sniff a sample that has this specific funky smell, I get a reward for it.
05:43So basically on day one, the dog steps on a platform and it just presents, you know,
05:50a lineup of samples. And out of curiosity, the dogs engages and smells the samples. On some samples,
05:57nothing happens. On the other samples, uh, you know, like hear this like sound and food comes out and
06:03basically a positive reinforcement, a very simple classic, uh, conditioning.
06:08Do the same behavior. The dog, when it gets reinforced on, in like the training phase,
06:15it gets reinforced on the, on like the cancel smell, it kind of like adopts its own behavior.
06:21So you'll see some of the dogs run there because of the feeder expecting reward. Some dogs freeze.
06:26Some dogs be scratched, uh, uh, at the sample with excitement. And we don't really care. As long as
06:32the machine that learns how the dogs react to the sample is very consistent and you know,
06:38it's very accurate and predicting the dogs like react.
