In this episode, I'm meeting with Christian family Ramsey, who's been doing some crazy stuff with neuroscience on a whole new term called Effective Computing that you're going to learn all about Welcome, Christian.

Thanks.

I'm glad to be here.

So we did have another episode with your partner.

How hand where she was talking a lot about dialects macking on.

I know you're working on that too.

Could you tell us about it?

Sure.

So dialects marking our is again The diet part is about to as one.

So it's seeing to his one and then the machine eyes about really us working alongside a machine.

So truly a fancy way to say to people working with a machine nice on Dhe.

So what we're trying to build out is this effective lier, and to us it it means that well, currently, as it stands, if you look at the neuroscience literature, emotion is just as important as cognition.

But in today's society, we seem to value cognition over emotion, and we're kind of like throw the emotions away, get them out of here.

Let's make a reasonable decision.

But it turns out that to make a reasonable decision, you actually have to have a marker, right?

An emotional market that that's the right answer and that one's not the right answer interest.

And so we're kind of trying to build this effectively to bring emotion back to the forefront of what?

What the everyday life is already like.

So I can give you an example of that.

Okay, let's imagine we're working on the team and you've got all of a sudden this great idea that you want to tell everyone about it.

Um, and what's going to happen is most likely you're gonna have sympathetic projections down from your autonomic nervous system.

Your blood pressure's going to rise.

The skin temperature is going to rise.

A little bit of sweat is going to secrete from your sweat glands.

Sounds like I should have good ideas, and it's actually your body's way of getting you ready to say what you need to say.

OK, and in that moment, maybe a couple of seconds later, you might have some thoughts, like what if I said this and everyone thought it was stupid or what if I said this and everyone thought this isn't a great idea.

And all of a sudden I've got, like, brain regions like the Perry act, Ical gray or even the beloved of Magdala are going off at that time.

And those emotions are so strong that well, I'm not gonna say anything right, Okay.

And then a couple of days later, someone says it has a positive reaction in your and now I'm thinking about it, and now I'm ruminating.

And now I'm feeling really complex emotions like like resentment or regret or even envy, right?

And so it's like, How do we actually bring that emotional dreamy and make it live for that person to understand the during that they're taking with the things in their lives or even the relations with other people?

And so that's really what the effective layer is.

A ballot is about awareness, making people more aware of that.

Okay, so one of the things you mentioned in your sights that I really caught my eye was like, you're working at the intersection of effective computing and deep learning.

And so what is that intersection and have it?

Yeah, I think that's ah, really good question.

Ah, A long time ago we started studying the emotion on really effective neuroscience, which has been largely, uh, at first, emotions were like, Oh, it's a psychological construct and like, yes, sort of.

And then you had people like, Yeah, keeping sap and Antonio Damasio neuroscientists who showed like no, actually, these three brain regions in some sort of temporal fashion unveil when you get the motion of something like seeking or rage.

And so they were able to do the necessary research to show that emotions were, in fact, real and that they were real science and you could actually study them.

And most of that was done in, like rodents and and then eventually to humans by looking at patients who had lesions in certain parts of their brain, and they couldn't have certain types of emotions.

Okay, and it pointed directly to that.

And so the effect of neuroscience for us, the part is looking at the deep learning part is how do you understand?

Physiological signal, like heart rate and heart rate is largely modulated by the central nervous it by your brain, right?

And so what you want to know is what effective?

Say this happening in your brain that eventually transmits down to you having a higher or lower heart rate.

And so, in trying to model that, that's what you're modeling.

When you're trying to model emotion, you're trying to model something that's happening in the brain.

So we thought we needed to understand that so that we could better model these emotions.

So that's where the crossroad, I think, tensorflow, uh, points out of really good.

Almost the complexity of classifying those emotions is you gotta have something like a deep learning there all night.

And you gotta have even pushing state of the art like attention networks and going further and tensorflow provides that for us.

Interesting.

So you've been using Tensorflow quite a bit.

Yes, thank you.

So so so like just trying to get my hands around.

This would be that.

So when you can understand the emotions of a person may be using a product that will help you build a better product and design something because you're reflecting, their emotions are all right, right?

Right.

It's like any time you're trying to model something, there's this.

There's this unknown right part that your model calm predict.

It's why you can't get 100% on every model that you put out there, and a lot of that could be definitely subject to like the user's emotion at that time.

Why they choose to play the game at that time rather than another time could also be like their stress levels were high and they wanted to play video games or maybe the opposite.

But knowing that gives you that indicator like you're saying of what their state is and what they're motivated to do at that time.

So it's going to be no, the tool in the developers toolbox to help them build better products.

Exactly.

Exactly.

We hope that everyone has access to this effectively.

Er, you know, now I know you're also looking for people to help with this, right?

So could you tell us a little about that?

Sure, the effective layer is massive.

And if you just look at old, the literature, ineffective nervous signs, effective computing, it's so much work has been done.

But yet there's so much left to do and we're not gonna tackle it alone way Do wish we could say that, but we're going to need partners and work with people, so we think of it as there's people that the research end of the spectrum and the applications in the spectrum in the research.

There's a lab out of Stanford that focuses just on fear, right?

And so they're just looking at the physiological output of fear.

When you see a stimulus that scares you, right?

Okay.

And so we could actually partner with, AH, lab like that or any such lab and say OK, how do we bring this into deep learning?

And how do we make some sort of continuous algorithm or attention network that's deployed in the cloud toe, where people who have effective disorders with anxiety or fear anything like that, we could be predicting when they're going to be fearful off what and then help him get better from that interesting and then on the application site.

I think there's a lot of developers to work with who want to create this kind of effective computing stuff, but may not have the requisite or like background or studying in the effective or motion sciences.

And there we can help out by saying, Hey, how would you label something like disgust like, That's difficult.

Ryan set up the right experiment to be able or the labeling process and help with that.

And then the last part, I think, is like thinking of organizations if you wantto get the effective like style of your team or something like that, and so we could work with the company to see and is going to be able to further the research on the effective layer, but also give insights to whoever whoever we're working with, right so effective computing.

It's a phrase most of us probably haven't heard of right now, but maybe all of us will be using it in the not too distant future.

That would be great.

Thanks, Cristian.

And thanks everybody for watching this episode of Tensorflow meets.

If you've any questions for me, if you have any questions for Christian, just please leave him in the comments below.

And don't forget to hit that subscribe button.

Thank you.

Thank you.