1
00:00:04,454 --> 00:00:04,888
Hi

2
00:00:04,888 --> 00:00:05,872
My name is Thomas Baker

3
00:00:05,872 --> 00:00:08,641
and I'm a Canada Research Chair in the area of 

4
00:00:08,641 --> 00:00:11,811
quantum computing for the modeling of molecules and materials

5
00:00:11,811 --> 00:00:14,080
which essentially means that we come up with better strategies 

6
00:00:14,080 --> 00:00:17,017
to solve quantum chemistry on a quantum computer.

7
00:00:17,017 --> 00:00:18,835
Computing is all about information.

8
00:00:18,835 --> 00:00:20,186
On a classical computer,

9
00:00:20,186 --> 00:00:22,872
you're shuffling around bits, charges, electrons

10
00:00:22,872 --> 00:00:24,557
into different places on the computer

11
00:00:24,557 --> 00:00:27,677
so that you can manipulate information in a particular way.

12
00:00:27,677 --> 00:00:29,679
The weird fact about quantum computing

13
00:00:29,679 --> 00:00:32,615
is that particles can be in more than one state at a time, 

14
00:00:32,615 --> 00:00:34,634
and can be entangled with each other.

15
00:00:34,634 --> 00:00:37,787
And what this means is that we can have the opportunity

16
00:00:37,787 --> 00:00:40,590
to make a fundamentally new type of computation

17
00:00:40,590 --> 00:00:43,443
based on this new entangled information.

18
00:00:43,443 --> 00:00:44,694
So one of the things that we do is

19
00:00:44,694 --> 00:00:46,296
we try and come up with better algorithms 

20
00:00:46,296 --> 00:00:48,481
to solve quantum chemistry problems

21
00:00:48,481 --> 00:00:50,750
which essentially means that we want to make it

22
00:00:50,750 --> 00:00:54,087
so that we can simulate what would happen in a chemical reaction

23
00:00:54,087 --> 00:00:55,355
or in molecules

24
00:00:55,355 --> 00:00:57,574
and then come up with better strategies to solve them

25
00:00:57,574 --> 00:01:00,043
so that we can make newer and better technologies.

26
00:01:00,043 --> 00:01:01,795
It's not easy to simulate

27
00:01:01,795 --> 00:01:04,280
all of these quantum states on the quantum computer.

28
00:01:04,280 --> 00:01:06,349
And in order to fully understand

29
00:01:06,349 --> 00:01:10,036
how the quantum computer can best constructed and best used,

30
00:01:10,036 --> 00:01:13,573
we use high performance computing in order to try and simulate

31
00:01:13,573 --> 00:01:16,960
what are called low-depth circuits on the classical computer

32
00:01:16,960 --> 00:01:19,679
and basic components of the quantum computer

33
00:01:19,679 --> 00:01:21,881
so that we can predict what's going to happen

34
00:01:21,881 --> 00:01:23,633
when we finally get around to making them.

35
00:04:03,793 --> 00:04:03,977
What?

36
00:04:03,977 --> 00:04:07,313
I'm working on with high performance computing in the cluster

37
00:04:07,313 --> 00:04:10,166
is testing a certain algorithm

38
00:04:10,166 --> 00:04:10,733
to see

39
00:04:10,733 --> 00:04:14,187
if it can make matrix multiplication in graphics,

40
00:04:14,187 --> 00:04:18,808
machine learning, quantum computation. If we can make it faster.

41
00:04:18,808 --> 00:04:20,493
But in order to test these things,

42
00:04:20,493 --> 00:04:23,279
we couldn't just put it on a PC or a laptop.

43
00:04:23,279 --> 00:04:26,282
That wouldn't really work because it would take forever

44
00:04:26,282 --> 00:04:27,533
to actually run it.

45
00:04:27,533 --> 00:04:29,469
if we put it on the cluster,

46
00:04:29,469 --> 00:04:31,838
we can run 100 of these at once,

47
00:04:31,838 --> 00:04:33,923
and then we can throw it all onto the cluster.

48
00:04:33,923 --> 00:04:35,858
And that's a huge advantage to us.

49
00:04:35,858 --> 00:04:38,411
One really interesting result we found from this research

50
00:04:38,411 --> 00:04:39,762
is that in the data,

51
00:04:39,762 --> 00:04:42,231
there was some discrepancies depending on the input.

52
00:04:42,231 --> 00:04:45,285
We started asking questions as to why these were occurring.

53
00:04:45,285 --> 00:04:46,719
And diving into that.

54
00:04:46,719 --> 00:04:50,423
We couldn't have done that without all the data that we had accumulated

55
00:04:50,423 --> 00:04:53,226
from running these things on the cluster.

56
00:04:53,226 --> 00:04:56,679
From there, we were able to come up with a new algorithm

57
00:04:56,679 --> 00:04:59,949
that works around some of these issues

58
00:04:59,949 --> 00:05:04,537
and has mainly benefits over the original algorithm.

59
00:05:04,537 --> 00:05:05,772
They're marginal,

60
00:05:05,772 --> 00:05:09,425
but every bit you can get with something that takes so long

61
00:05:09,425 --> 00:05:12,545
is a giant improvement. Even if it's just 5%.