Can sound levitate objects and create holograms you can touch? with Sriram Subramanian

0:06 [applause] [applause] [applause]

0:18 So, when we think about magic or all stories of magic,

0:22 people always talk about floating objects.

0:26 [snorts] I mean it's nothing better than talking about Professor Flitwick.

0:31 I'm sure the first thing he would want is to see some objects float.

0:36 What do you say, Richie?

0:50 The thing about giving live demos is you never know.

0:55 But I hope

1:04 [snorts]

1:13 We're learning our wizard skills.

1:18 [laughter]

1:27 Well, finally.

1:32 [applause]

1:33 Now, what you're seeing there is a small object that's held in midair.

1:38 And the the most important question there is how is it being held?

1:42 It's not quite magic.

1:44 It's mostly with sound.

1:46 And it's useful to talk a little bit about what we mean by sound waves.

1:52 So I'm not going to dwell on all the basic things here.

1:56 The most important thing that's useful to know is the phase.

2:00 And the thing that we're going to manipulate

2:02 in the sound wave is the phase of the wave,

2:04 which is the the difference or it's it's how far along

2:10 the wave has come from this beginning point to its peak.

2:14 Right?

2:14 So, that's the phase.

2:15 And what we do is we manipulate this to do fun things with it.

2:23 [clears throat] So, I've shown you some levitation.

2:27 It's actually very simple to do levitation.

2:31 All you need is two transducers, two speakers, any frequency,

2:35 and you point them at each other, they create a standing wave pattern.

2:39 And if you if you have a light object, something that's light enough,

2:43 and you place it somewhere between those anti-nodes,

2:47 you will most likely levitate it.

2:49 It's going to work.

2:50 Okay?

2:52 The The question is then, can we do more?

2:56 Can we do something fun with it?

2:57 Right?

2:58 So, [clears throat] let me see if this works, right?

3:01 We thought, can we do some levitation?

3:03 So, I got someone to try some coffee beans, some milk, some meat, lettuce.

3:09 I guess you'd call this a burger.

3:11 You'd like your coffee.

3:14 [snorts] So, that's quite simple thing, right?

3:16 So, it's nice to do.

3:17 It's nice to see what we can do with those things.

3:19 Um But, one thing we did was we um [snorts]

3:26 we made an array, which is with lots of transducers, lots of speakers.

3:30 So, something that looks like this.

3:33 So, it's an array of speakers with ultrasound.

3:35 So, we're talking about 40 kHz sound, so it's just outside your audible range.

3:39 And the thing we can do with this is we can control every speaker individually.

3:43 Right?

3:44 And then I can change the timing between

3:46 the speakers to introduce different phase delays between them.

3:52 [snorts] Um So, can I do levitation with this array, right?

3:55 If I have something like this instead of two speakers pointing at each other,

4:00 how does one do levitation?

4:02 So, when I think about levitation what I'm trying

4:04 to do is I'm trying to hold an object with sound.

4:08 Imagine if it's with my hand.

4:09 So, I have my hand with me here, my fingers.

4:12 I pinch them.

4:13 I grasp them.

4:14 And an object is going to sit over here, right?

4:16 So, it's going to be held by my fingers.

4:18 This is truly all we're trying to do, right?

4:21 So, what we're trying to do is to see if I can shape the sound in such

4:25 a way that all the forces from the sound

4:28 field come together and converge at this point.

4:32 That's good enough.

4:34 Now, the the thing is if I want to hold an object,

4:40 I can have a bowl and the bowl is upside down.

4:43 I place something on top of the bowl.

4:44 Most likely it'll stay there unless you touch it a little bit, right?

4:48 If I touch it, it's going to fall off.

4:49 It's like but if the bowl is with a cup

4:53 at the bottom and if I put something on top,

4:55 it's going to come down to the bottom and stay there.

4:57 It'll be stable.

4:59 [snorts]

4:59 It's this stability that we're looking for.

5:02 So, if you talk about levitation it's a particle that's trapped by sound.

5:06 Uh, so we're talking about acoustic levitation here.

5:09 And a trap is a point where all the forces converge.

5:13 That alone is sufficient for you to get levitation to work.

5:17 But it's not going to be stable.

5:20 What we want for stability is the pressure amplitude to be minimum as well.

5:25 The [snorts] it's it's When you talk about sound,

5:31 we're talking about a small object.

5:33 So, if I make an assumption that this object is very small,

5:36 that means it's smaller than the wavelength of the sound.

5:39 That means this object is going to be held between those two uh, anti-nodes.

5:44 Uh, there's this thing called the Gor'kov potential

5:47 that tells you exactly how to compute the forces.

5:50 The nice thing is that you don't need to do all

5:52 the uh you you can just add up the forces, right?

5:54 Because it's a small object.

5:55 It's a point object.

5:56 You can add the forces.

5:59 [gasps] So, the objective function there is the more interesting thing, right?

6:02 Um what it says is that it's the sum of the pressure amplitudes and the force.

6:07 There are few terms there.

6:08 There's a there's the amp pressure amplitude term,

6:11 then there's the uh weights for the force.

6:16 [gasps] So, we did the optimization.

6:19 It's a non-linear optimization here.

6:21 Um so, we did that, and we found some interesting things.

6:25 I mean, we did the op we just got some solutions, right?

6:27 So, let me show you what this looks like.

6:29 Um so, you're seeing how the force looks in two different axes.

6:36 Um you can see it's really high intensity.

6:40 And then there's uh so, it's looks like my finger, right?

6:42 So, if it's like twin trap, it's like my two fingers pinching together,

6:45 so it's really dark and bright over there, and the object is held in the middle.

6:49 Or you have a vortex where it's like I'm holding it slightly differently, right?

6:55 And the bottle is like I'm holding it in the bottle.

6:58 That's good enough to do levitation, but it was a non-linear optimization.

7:03 It was quite complicated to get the solution.

7:06 Uh so, we thought, "Let's take a look at what the speaker looks like." So,

7:10 what is it What are the face patterns that are coming out of the speaker?

7:15 Uh we want to know what should I emit for a twin trap or a vortex or bottle,

7:22 and how does that affect levitation?

7:23 So, it looks like something like this over there.

7:26 I'm sure you see it it there seems to be some pattern over there,

7:31 but you cannot quite say what that pattern is.

7:34 Not sure what what's exactly going on there, right?

7:39 But, let's come back to this analogy of the hand, right?

7:42 So, when I use my fingers to grab an object,

7:45 the one thing I know is that wherever the object is,

7:49 most of my fingers are going to be there.

7:51 Right?

7:51 So, it's that's where the focus of the object is.

7:54 So, we thought what if we figure out what that location is going to be.

7:59 So, what is the focus?

8:00 How does my finger all come together?

8:02 So, I got that pattern.

8:04 That's easy enough to compute.

8:08 And then I take that away from this field that we computed, right?

8:15 If I do that, amazingly, we get something like this.

8:20 Right?

8:20 This is a static pattern, right?

8:22 So, what it says is actually levitation

8:25 with sound is not a optimization problem.

8:28 You can optimize, of course, and you can do a lot of optimization there.

8:32 But, it's actually a much simpler problem.

8:34 Two steps that you can add together.

8:37 You calculate the focal point, and you add that trap on the top, right?

8:41 It's a signature.

8:43 Now, the focal point calculation, if you do if you've done A-level physics,

8:49 you would have learned how to do this in A-level physics.

8:51 It's just interference superposition calculation.

8:56 So, what I can do instead of computing

8:58 those twin traps is I can compute this focal point.

9:02 I can add the signature.

9:04 And this focal point calculation is not just conceptually simple,

9:10 it's computationally super efficient, right?

9:12 I I can calculate it really, really fast.

9:16 Um So, Rui, should we show what we can do?

9:34 So, uh should we do the levitation first?

9:44 Should we get someone, maybe?

9:46 I don't know how any magicians, wizards here.

9:49 All right.

9:50 Come on.

9:56 [applause] No, no, not so far.

10:07 Yeah.

10:12 You need to say "Wingardium Leviosa." So, this particles move Oh,

10:22 you're Yeah, so it's it's moving with her finger, right?

10:28 I say I can go I can do fast.

10:30 I can do the calculations quite fast.

10:32 So, move your finger a little faster.

10:37 There you go.

10:38 So, I don't know if you're able to see on the camera there.

10:45 So, what you're seeing is the particles going so fast that you stop seeing

10:50 this polystyrene bead that we are using

10:52 and you're seeing the shape the whole shape.

10:57 If you move with your finger, it will move with your finger.

11:00 Uh she stopped it.

11:01 Yes.

11:05 This is amazing.

11:05 Yes.

11:06 So, we're not done.

11:08 We have another demo.

11:09 Yes.

11:12 [applause]

11:14 So, we we can calculate these these focal points at 40,000 frames per second.

11:20 Even that's nothing.

11:21 We can go even faster than that.

11:23 And you go so fast that we can have we're able to update

11:26 the position of the particle much faster than your eyes can see the movements.

11:31 So, you see the whole shape.

11:32 Right?

11:33 In fact, I said I can go faster than 40,000.

11:35 So, what happens if I'm holding an object in air with my finger, right?

11:39 I can move my finger.

11:40 If I'm fast, I can let go and catch it again.

11:42 And if I'm because I'm so fast, the object is not going to fall.

11:46 Right?

11:46 I can do the same thing here.

11:48 We can let go, do something else, come back, and catch it again.

11:52 Right?

11:53 Um So, what do we do when we let go of this object?

11:57 We thought, what if we play that same field in front of your hand,

12:02 so you can feel what's happening, right?

12:04 So, you can feel something, some forces on your finger or your palm,

12:09 and then the levitation happens as well on the side.

12:11 Should we just do the haptics so people can see?

12:14 Maybe.

12:15 Do we have another volunteer?

12:18 Anyone else want to You want to come back?

12:19 Yes.

12:20 Yeah.

12:21 [laughter] [applause]

12:32 Have you ever felt sound before?

12:34 Have you felt the sound?

12:35 Yes.

12:38 Yeah, exactly, right?

12:39 So, yeah, precisely.

12:42 So, what we're doing is no different from that, right?

12:43 Except it's it's uh ultrasound at 40 kHz.

12:47 So, instead of beating in your chest, the wavelengths are small,

12:49 so it's in the palm of your hand.

12:51 So, you want to try Put your hand over there.

12:53 Put your whole palm.

12:53 Yeah.

13:00 It tickles.

13:00 It almost feels like wind.

13:02 Yes.

13:03 Yeah.

13:04 So, so we can do many We can move the pattern around.

13:09 We can give you some different shapes.

13:10 Uh we So, we And we can multiplex the two of them, right?

13:13 Because we're doing it so fast, we can do we can give the we can

13:17 have a physical object that's floating there as a hologram,

13:20 and we can also give haptic feedback.

13:23 There's another piece of this demo that I'm not able to do quite here,

13:25 so I'm going to play the video to show you what's the last piece of Yes, please.

13:28 Thank you so much.

13:29 Thanks a lot.

13:35 Thank you, Richie.

13:36 So, let me just play the video so you can see the whole thing, right?

13:48 [music] [music] [music] [music] [music] [music] [music] [music] [music]

14:39 So, what we're going to do now is

14:41 you're going to start hearing the object speak, right?

14:45 [music] It's coming up.

14:55 So, that audio doesn't sound impressive, right?

15:00 I mean, like I can say it's 3-2-1-0 myself,

15:03 but the the the thing there is that we I said it's all ultrasound speakers.

15:08 So, there's only ultrasound, and still you heard the sound.

15:11 That's because we had so much time that we could modulate the amplitude.

15:15 So, we we took the audio signal and put

15:17 it on top of everything else we were doing, and then you hear the beat sing.

15:21 So, actually the sound is coming from that object there.

15:24 So, it feels like it's singing, it's dancing, and you can feel it as well.

15:28 So, that's the multimodal that we can generate there.

15:32 [snorts] The So, yeah, so that's how we do it.

15:34 So, we we take the three stimulus that you want to create.

15:37 So, there's the shape, then there's the tactile sensation or the haptics

15:41 that you want to create with it, and then the audio source, right?

15:44 So, we take them all, mix them together, and play it.

15:46 And then yeah, you you get all three of them at the same time.

15:50 on this object.

15:51 Uh so, this works really well, and it's it's nice to see this, and you

15:58 can see that we can create holograms and so on.

16:01 But, everything only works if it's a single object.

16:04 Single polystyrene bead I can move around.

16:07 What if I need to have more?

16:09 What if I have a whole shape?

16:11 Right?

16:11 So, let's say I'm given this image that I want [snorts] to create with sound.

16:18 How do I do that?

16:19 This is an inverse problem, right?

16:20 So, the Let me start over here.

16:23 So, actually it's quite uh interesting that I'm

16:29 told 65 years ago today or to the month,

16:33 Dennis Gabor uh was here giving lecture in this very same theater on holography.

16:39 Uh and what we're talking about here is also about holography.

16:43 But, we're talking about acoustic holography instead of optical

16:45 holography that Dennis Gabor got the Nobel Prize for.

16:48 Uh So, what we're what we're trying to do

16:52 is we're given some image that we have,

16:56 some target, and we want to know what should

16:59 be the amplitudes and the phases of the speakers, right?

17:03 So, what what should be the activation of the speakers to generate that shape?

17:11 This is not so difficult.

17:13 Sorry, what happened?

17:14 The problem This is the problem, right?

17:16 So, we know what pattern we want that's in the target.

17:21 We don't know what signals to send to the speakers.

17:24 But, I want the speakers to pump up as much

17:27 energy as it possible as it possibly can, right?

17:30 So, we want maximum energy out of the speakers.

17:33 That [snorts] means I'm going to turn it all on to maximum amplitude.

17:36 So, the only thing I have control of is the phase of these um speakers.

17:42 So, the question is what set of phase

17:44 patterns will produce the target image that I want?

17:50 Actually, this problem can also be quite simple to solve if and only

17:53 if I tell you what the target pattern is in fully, right?

17:59 Typically, with all such scenarios,

18:02 people can give you what the image energy should be,

18:05 but they cannot tell you what the phase is going to be.

18:07 If I think of an image, an image over here,

18:10 if you take a picture of me standing here,

18:13 you it's very easy for the camera to capture

18:15 all the amplitudes of the signal that generates this image,

18:19 but the camera is not going to tell you

18:20 what the phases are for the wave to generate it.

18:23 If it can, then it's quite easy

18:25 to calculate what the sources should emit, right?

18:30 [snorts] In the absence of that, how do we solve this problem?

18:32 So, we don't know what the right phases are.

18:36 We just take a random guess.

18:38 Uh and this is a nice thing.

18:40 So, you just take a guess.

18:41 You assign those random phases

18:42 to those to the amplitudes of the transducer, right?

18:46 Uh and then you propagate that field to the the target location you want.

18:53 Unsurprisingly, this is not going to give you anything interesting or correct.

18:57 But, we have something now we can improve on.

19:01 What we can do is we can take whatever we find at the target.

19:04 So, we've got some new phases and new amplitudes at the target.

19:08 But, at the target, I know that I need to have a different pattern,

19:12 the pattern that's given.

19:14 So, I'll take away all the rubbish I got from my calculation

19:17 and replace it with the actual one that I want, the target.

19:22 And then I take it all and recalculate

19:24 all the calculations back to the transducers.

19:28 Once I come back to the transducers,

19:29 I'm going to have a new set of phases and amplitudes.

19:32 I'm going to replace all the amplitudes to one,

19:36 keep the phases, and go back again.

19:38 Right, so I keep doing this iterate this back and forth.

19:41 So, each correction breaks one constraint, but respects the other.

19:45 And alternating between them should satisfy more or less both of them.

19:50 Right, so the amazing thing is it converges.

19:53 I cannot theoretically show.

19:54 Maybe we can show theoretically that it converges,

19:56 but the um you do 25 times, 30 times, you're there.

20:01 Right, you have a solution that looks pretty much final.

20:06 This works, and it works in optics quite well.

20:10 It'll also work in acoustics.

20:12 But, uh we have an advantage in sound [snorts]

20:16 manipulation that is not easy to achieve in optics, right?

20:21 All our All my speakers here, I can control both the amplitude and the phase.

20:25 I can turn it on at whatever amplitude I want on and phase I want.

20:29 Usually, with light, you can either control one or the other, but not both.

20:35 So, this is the algorithm.

20:36 This is called Gerchberg-Saxton algorithm.

20:38 It came in around 1976, a few years after Gabor.

20:42 But, we can play a little trick on it, right?

20:45 So, uh and I'm saying this trick is [snorts]

20:47 useful to show you because that's what gives us the speed,

20:50 and we always care about speed.

20:52 Maybe it's my bias.

20:53 Um So, what we can do is because I can turn on at different amplitudes,

20:59 I don't need to change set the amplitude at the transducers to one, right?

21:03 I can just keep leave it away.

21:04 If I remove that bit,

21:05 what happens is that I can just combine the going forward and back,

21:09 they both became the same thing because [snorts]

21:12 that F times B, this matrix is a constant.

21:15 You can take it out.

21:16 You can pre-compute the whole thing.

21:18 Then now you can go back to this 10,000 updates per second that I can get you.

21:23 Right?

21:23 So, when I start doing that, you can see over here I can

21:27 give you the sound and I have two different objects that are levitating.

21:38 So, what I've tried to show you here is we can given any pattern calculate what

21:48 the transducers amplitudes and phases should be so

21:53 that we can do levitation with it, right?

21:55 I mean I'm sure you're all wondering just like Professor Flitwick

22:01 would have been wondering where is the feather in all of this?

22:05 I mean, there's no levitation without a feather.

22:08 Right?

22:09 So, we have to come back to it.

22:18 Anybody feels like Wingardium Leviosa?

22:23 Do you want to volunteer to do Yes, come along.

22:29 Come.

22:32 [applause] How's your wizardry skills?

22:42 Wow, it's quite strong as you can see.

22:44 Yeah, I mean, he just had to be here.

22:47 And it's We need to start again.

22:53 Can I go on?

22:54 Yeah.

22:55 Wingardium There you go.

23:02 [applause] So, I mean Thank you.

23:10 [applause]

23:13 Uh yeah, so what we can do is we can take these polystyrene beads,

23:17 we can attach them to an object and lift the whole object up, right?

23:21 So, that's what we did.

23:23 So, we um [snorts] I'll show you how we did it just quickly.

23:32 [music]

23:32 So, we have a little tool in which you can place the location of the objects.

23:35 You can [music] say which wings of the butterfly

23:38 can flap and where it shouldn't flap,

23:40 [music] and then we can run some optimization to adjust all the parameters.

23:47 Then we can Yeah, we can levitate the object the whole object.

23:50 Once we can levitate the whole object,

23:52 nothing is stopping us from integrating some simple game engine to this, right?

23:56 It's always fun.

23:58 So, when you start doing game engines

24:00 on you can make them move like physical objects.

24:03 You can anthropo anthropomorphize them even, right?

24:06 So, you can see over here.

24:22 So, Yeah, so we can make more complicated objects.

24:36 We can do smaller objects, many objects, and so on, right?

24:39 But one thing that you will notice is that there's nothing there.

24:44 It's empty, right?

24:45 The only thing we can do is the object.

24:47 So, the levitation only works if the volume that we're operating in is empty.

24:54 So, can we change that?

24:56 Can we do something about this?

24:57 right?

24:57 This was something that we were thinking about.

24:59 Um so, this is an example of what we tried.

25:07 We took the same object.

25:12 So, what the all the trick I've been

25:16 talking about about the the adding the signatures

25:20 doesn't quite work when you have an object

25:22 here because when you have some object,

25:24 it's going to reflect and scatter the sound

25:28 in different ways than when there's no object there.

25:32 So, what we have to do is we have to mesh it.

25:35 So, we break this object into tiny rectangles, triangles,

25:39 and for the each triangle, we try and compute the reflections from it.

25:45 Um so, this is like a boundary element

25:47 method if uh if you're interested in that.

25:50 The um so, then the total contribution of What is

25:54 the total pressure that you get at any focal control point?

25:57 Is the sum of the direct path and the reflected path.

26:02 Now, if I assume that this object,

26:05 the bunny in this case, is static, so it doesn't move, right?

26:09 Then, I can precompute all of those things, right?

26:12 So, all of that can be calculated ahead of time so that I

26:17 store it in memory and just use it instead of calculating it repeatedly.

26:22 And then, the only thing I need to do

26:23 is this reflection from the bunny and the direct path.

26:27 That's always going to be different.

26:29 So, we have to um calculate them.

26:32 The So, yeah, so we we can do this and then we can actually go quite Yeah,

26:38 we can do similar things then with the whole with the objects.

26:43 [music] So, we can go under different things.

26:52 [music] Yeah.

27:12 I've been talking you through all the different tricks we've

27:15 done to to make the levitation more exciting exciting and interesting.

27:22 But I'm sure you're wondering what can we do with this?

27:25 So what?

27:25 I mean like what's the application of this?

27:27 So what we have is a way in which we can

27:30 control the sound field in in many different scenarios and contexts.

27:35 So let me go back to the very beginning, right?

27:37 So uh the lady over there she when she came up she could feel something uh

27:42 on this transducer board when I So you can

27:45 imagine taking that whole space of haptics forward, right?

27:49 So this is something that we did so where you can you can generate the sound

27:54 field in on the palm of your hand and start feeling the sense of haptics.

28:00 This I don't know if this video is going to play.

28:04 Uh there is sound but something happened with the sound it's off.

28:06 So basically the person is describing exactly it's it's incredible

28:10 that he was saying exactly like what you were saying, right?

28:13 So it feels like feather something tickling in the hand and soft.

28:19 So you feel the sensation but it's not so

28:22 strong that it's going to stop you from doing anything.

28:24 But the sensations are enough to notice that something is there.

28:28 So if you imagine you're playing a game of I I I am

28:31 with any gaming device like an Xbox or a or a Nintendo Wii you

28:38 you can hit the ball or you can if you're playing volleyball or tennis

28:41 but you don't feel that you touch something with those 3D gaming systems.

28:45 So now you can project the sense of haptics onto the palm of the hand.

28:49 So the person can feel the object as well.

28:53 Um in fact, automotive companies were excited about this for uh [snorts]

28:58 the dashboards of the car.

29:09 [music] [music] [music] [music] [music] [music] [music]

30:05 So the production quality of these videos

30:07 go up as it becomes commercial products.

30:11 The Coming back to the research side of things, right?

30:16 So here's a I want to show this funny little video.

31:02 [laughter] Yeah.

31:03 Imagine watching the same video with this haptic feedback, right?

31:06 So, you you imagine you're sitting uh and you get the sensation.

31:10 I I I at least couple of you were scared at the moment the crocodile came out.

31:15 Imagine the same thing when you were when

31:18 you're watching with this tactile sensation coming to you.

31:21 It's much more powerful with that and evocative

31:24 when you have this kind of tactile sensations.

31:27 So, there's a lot we can do with haptics.

31:30 Uh it was a fun journey to take this uh into the real world and see how

31:36 customers react to it and what the opportunities are

31:39 for um embedding these systems in the real scenarios.

31:44 Um but I've been I've not uh all the things I've

31:50 been talking about for the last half an hour or so,

31:53 mixed haptics and levitation, right?

31:55 So, um we have to answer the question, what do we do with the levitation part?

32:00 Uh so, I want to show you one more video on these things.

32:08 [music] The nice thing about levitation is that it's material agnostic,

32:12 so we can it's not just solid objects.

32:16 We can also levitate liquids, fluids, powders.

32:23 [music] [music] [music] [music] [music]

33:02 So, the advantage of this kind of printing

33:05 is that you can print from any direction.

33:07 So, you don't have to always come from the top to print on the bottom.

33:10 You can print omnidirectionally.

33:13 The other thing is you can mix different types of materials, right?

33:15 So, you can mix uh you can mix biological samples

33:19 if you want with um uh other kinds of fluids.

33:23 So, there's a lot of flexibility in how you can 3D print with this thing.

33:28 Um so, my multimaterial 3D printing is something that's fascinating uh and uh

33:36 [clears throat] something that we can work with in levitation.

33:38 So, here's another one that I want to share,

33:41 which is about seed separation, right?

33:43 I mean the the the amazing thing for me is that when we started doing this work,

33:47 we we start getting all kinds of application

33:50 interest from different customers and end users.

33:52 So, we had 3D printing interest,

33:55 and then we had some uh seed companies who wanted to know if

33:59 we can use this for sorting seeds in a in a conveyor belt.

34:05 I went, "Oh, that doesn't sound like levitation,

34:07 but actually it's all about applying forces and moving objects."

34:10 And some of the nice things over here is um [snorts]

34:15 we Usually, when you sort seeds and they're talking about

34:18 tons and tons of seeds coming down a conveyor belt,

34:20 um you only put them in bin two bins, A or B.

34:24 But, what we can do is put them in many different bins.

34:27 And you can also align them like in a row so that you can

34:32 pick different rows for different uh imaging

34:34 purposes or pick them up and see them.

34:37 So, I don't know how many of you knew

34:39 that a tomato seed is about as expensive as gold, right?

34:44 So, and and what they want to be able to do is to be able

34:47 to see whether this tomato seed is going

34:50 to germinate all in the same time or not.

34:53 So, that's that's something that farmers care about.

34:55 What the last thing you want as a farmer is you have all these tomato

34:59 plants and they all start producing fruits different times of the of the week.

35:04 You want them all to come fruit at the same time.

35:07 So, what they do is they want to be able to individually inspect every

35:10 seed to check that it's um it's good before it this they package it.

35:16 So, that's all possible with uh this kind of platform.

35:22 And again, we were in amazed to hear uh companies that were coming there.

35:28 But, in in some sense,

35:32 where we where I'm going with this is that we've we've talked about

35:37 how we can shape the sound field to do many different things, right?

35:40 We can we can give you sensations in the palm of your hand.

35:43 We can give you sensations to levitate an object.

35:46 It can be a powder, it can be a liquid, it can be a solid.

35:49 Uh we can we can move seeds.

35:54 I feel like where we are really heading towards this is we

35:58 can control the placement of physical matter anywhere in 3D space, right?

36:03 So, we we can be um it can be biological samples,

36:10 it can be agricultural products,

36:13 it can be 3D printing materials, or anything, right?

36:16 So, it's it's all about uh contact-free manipulation of materials.

36:20 So, in some sense, this is [music] [music] [music] [music]

36:51 Yeah, so this I feel is where

36:53 we're going with levitation and acoustic holography.

36:58 And so with that I'm going to open up for any questions if you want,

37:02 so thank you very much.

37:06 [applause]