WEBVTT

1
00:00:00.640 --> 00:00:04.160
What's going on inside this black box ?

2
00:00:04.160 --> 00:00:06.480
Complex AI algorithms are trying to solve

3
00:00:06.480 --> 00:00:08.800
the world's most difficult problems :

4
00:00:08.800 --> 00:00:10.080
they're fed with data

5
00:00:10.080 --> 00:00:12.800
which they analyze to find patterns.

6
00:00:12.800 --> 00:00:13.920
This is how they provide us

7
00:00:13.920 --> 00:00:16.760
with predictions to answer tricky questions :

8
00:00:16.760 --> 00:00:17.640
from how changes

9
00:00:17.640 --> 00:00:20.640
in a process affect production volume or speed,

10
00:00:20.640 --> 00:00:23.280
to how policies might impact society.

11
00:00:23.280 --> 00:00:25.920
But how can we trust or improve these predictions

12
00:00:25.920 --> 00:00:28.920
if we don't know what happens inside the box ?

13
00:00:28.920 --> 00:00:30.440
BCG’s FACET

14
00:00:30.440 --> 00:00:32.120
is switching on the light.

15
00:00:32.120 --> 00:00:34.280
Our open-source library provides

16
00:00:34.280 --> 00:00:36.480
an unprecedented view of the interactions

17
00:00:36.480 --> 00:00:37.960
between the different variables

18
00:00:37.960 --> 00:00:39.880
in machine learning models.

19
00:00:39.880 --> 00:00:41.520
It applies a unique algorithm

20
00:00:41.520 --> 00:00:43.520
for global model interpretation

21
00:00:43.520 --> 00:00:45.840
enabling data scientists to understand

22
00:00:45.840 --> 00:00:48.320
how the different factors influence each other.

23
00:00:48.320 --> 00:00:51.240
With the release of our new version available now,

24
00:00:51.240 --> 00:00:54.160
FACET offers an enhanced machine learning workflow,

25
00:00:54.160 --> 00:00:56.800
better model inspection, new visualizations,

26
00:00:56.800 --> 00:00:58.400
and much more !

27
00:00:58.400 --> 00:01:00.240
By making the black box transparent,

28
00:01:00.720 --> 00:01:03.440
we take the leap from prediction to explanation.

29
00:01:04.680 --> 00:01:07.000
If we know how parameters relate to each other,

30
00:01:07.000 --> 00:01:08.040
and to outcomes,

31
00:01:08.040 --> 00:01:09.760
we can run “what if” scenarios

32
00:01:09.760 --> 00:01:11.200
and optimize these outcomes

33
00:01:11.200 --> 00:01:13.120
with surgical precision.

34
00:01:13.120 --> 00:01:15.960
We can eliminate bias by detecting its source.

35
00:01:15.960 --> 00:01:18.480
We can also enable all those working with algorithms

36
00:01:18.480 --> 00:01:20.600
to make informed decisions.

37
00:01:21.400 --> 00:01:22.800
Curious ?