WEBVTT

00:00:00.000 --> 00:00:01.851 align:middle line:90%


00:00:01.851 --> 00:00:02.679 align:middle line:90%
Machine Learning

00:00:02.679 --> 00:00:03.181 align:middle line:90%
AI

00:00:03.181 --> 00:00:03.690 align:middle line:90%
AI

00:00:03.690 --> 00:00:04.912 align:middle line:90%
Machine Learning

00:00:04.912 --> 00:00:06.190 align:middle line:90%
Artificial Intelligence

00:00:06.190 --> 00:00:06.920 align:middle line:90%
Machine Learning

00:00:06.920 --> 00:00:08.161 align:middle line:90%
Machine Learning or AI

00:00:08.161 --> 00:00:09.050 align:middle line:90%
Machine Learning

00:00:09.050 --> 00:00:10.090 align:middle line:90%
Machine Learning

00:00:10.090 --> 00:00:10.850 align:middle line:90%
AI

00:00:10.850 --> 00:00:12.240 align:middle line:90%
Confusing, right?

00:00:12.240 --> 00:00:13.970 align:middle line:84%
Artificial intelligence
and machine learning

00:00:13.970 --> 00:00:15.930 align:middle line:90%
are often used interchangeably.

00:00:15.930 --> 00:00:18.020 align:middle line:84%
So they must relate
to each other.

00:00:18.020 --> 00:00:19.226 align:middle line:90%
But how?

00:00:19.226 --> 00:00:21.670 align:middle line:84%
AI or artificial
intelligence is the science

00:00:21.670 --> 00:00:24.700 align:middle line:84%
of training machines
to perform human tasks.

00:00:24.700 --> 00:00:27.107 align:middle line:84%
The term was
invented in the 1950s

00:00:27.107 --> 00:00:29.940 align:middle line:84%
when scientists began
exploring how computers could

00:00:29.940 --> 00:00:31.550 align:middle line:90%
solve problems on their own.

00:00:31.550 --> 00:00:34.490 align:middle line:84%
Most of the time when
you hear AI, you probably

00:00:34.490 --> 00:00:36.112 align:middle line:90%
see these ridiculous images.

00:00:36.112 --> 00:00:36.614 align:middle line:90%


00:00:36.614 --> 00:00:38.470 align:middle line:84%
ANDY RAVENNA: I
grew up at the time

00:00:38.470 --> 00:00:42.020 align:middle line:84%
when we were watching
"The Jetsons" and "2001:

00:00:42.020 --> 00:00:45.800 align:middle line:84%
A Space Odyssey" where they
had HAL the talking computer.

00:00:45.800 --> 00:00:49.160 align:middle line:84%
So for me Artificial
Intelligence

00:00:49.160 --> 00:00:52.630 align:middle line:84%
is really like a
computer or a machines

00:00:52.630 --> 00:00:56.589 align:middle line:84%
that are given properties,
human-like properties.

00:00:56.589 --> 00:00:59.200 align:middle line:84%
We take for granted how
our brains effortlessly

00:00:59.200 --> 00:01:02.960 align:middle line:84%
calculate the world around
us, every second of every day.

00:01:02.960 --> 00:01:06.580 align:middle line:84%
AI is the concept that a
computer can do the same.

00:01:06.580 --> 00:01:10.880 align:middle line:84%
BOB ROGERS: All of our speech
interfaces for our devices

00:01:10.880 --> 00:01:11.400 align:middle line:90%
are AI.

00:01:11.400 --> 00:01:12.260 align:middle line:90%
And it's incredible.

00:01:12.260 --> 00:01:13.538 align:middle line:90%
You can have an accent.

00:01:13.538 --> 00:01:16.410 align:middle line:84%
You can be speaking
a particular dialect

00:01:16.410 --> 00:01:19.190 align:middle line:84%
that, as long as there's
data on the Internet

00:01:19.190 --> 00:01:22.928 align:middle line:84%
with that language, these
AI systems can quickly

00:01:22.928 --> 00:01:24.650 align:middle line:90%
developed a way of interacting.

00:01:24.650 --> 00:01:26.761 align:middle line:84%
TOM SABO: I can pick
up my phone and I

00:01:26.761 --> 00:01:32.376 align:middle line:84%
can ask it questions like I
can say, "Siri say my name."

00:01:32.376 --> 00:01:33.320 align:middle line:90%
SIRI: You're Tom.

00:01:33.320 --> 00:01:36.730 align:middle line:84%
But since we're friends I
get to call you M-R Smiley.

00:01:36.730 --> 00:01:40.470 align:middle line:84%
While AI is the broad science
of mimicking human abilities,

00:01:40.470 --> 00:01:43.720 align:middle line:84%
machine learning is a
specific subset of AI

00:01:43.720 --> 00:01:46.710 align:middle line:84%
that trains a
machine how to learn.

00:01:46.710 --> 00:01:49.890 align:middle line:84%
Machine learning models
look for patterns in data

00:01:49.890 --> 00:01:52.930 align:middle line:84%
and try to draw conclusions
like you or I would.

00:01:52.930 --> 00:01:56.230 align:middle line:84%
BOB ROGERS: They're not
being explicitly programmed

00:01:56.230 --> 00:01:56.810 align:middle line:90%
by people.

00:01:56.810 --> 00:01:58.620 align:middle line:84%
You can actually
give some examples

00:01:58.620 --> 00:02:00.940 align:middle line:84%
and they're going to learn
what to do from those examples.

00:02:00.940 --> 00:02:04.828 align:middle line:84%
That's a huge difference because
it's much easier for us humans

00:02:04.828 --> 00:02:07.550 align:middle line:84%
to give examples than it
is for us to write code.

00:02:07.550 --> 00:02:09.940 align:middle line:84%
Once the algorithm gets
really good at drawing

00:02:09.940 --> 00:02:12.400 align:middle line:84%
the right conclusions,
it applies that knowledge

00:02:12.400 --> 00:02:13.511 align:middle line:90%
to new sets of data.

00:02:13.511 --> 00:02:15.500 align:middle line:84%
BOB ROGERS: That's
the lifecycle.

00:02:15.500 --> 00:02:19.320 align:middle line:84%
It's ask the question, collect
the data, train the algorithm,

00:02:19.320 --> 00:02:21.293 align:middle line:84%
try it out, collect
the feedback,

00:02:21.293 --> 00:02:24.460 align:middle line:84%
use the feedback to make the
algorithm better so that you

00:02:24.460 --> 00:02:27.830 align:middle line:84%
have increasing accuracy
and performance.

00:02:27.830 --> 00:02:28.330 align:middle line:90%
Ta-Da!

00:02:28.330 --> 00:02:29.030 align:middle line:90%
Machine Learning.

00:02:29.030 --> 00:02:31.419 align:middle line:84%
LISA LOFTIS: If you
look at the Google car,

00:02:31.419 --> 00:02:34.490 align:middle line:84%
it has lasers on the
top which are telling it

00:02:34.490 --> 00:02:37.620 align:middle line:84%
where it is in terms of
the surrounding area.

00:02:37.620 --> 00:02:40.050 align:middle line:84%
It has radar in
the front, which is

00:02:40.050 --> 00:02:43.780 align:middle line:84%
informing the car of the
speed and motion of all

00:02:43.780 --> 00:02:44.980 align:middle line:90%
the cars around it.

00:02:44.980 --> 00:02:47.800 align:middle line:84%
And it uses all of
that data to figure out

00:02:47.800 --> 00:02:50.730 align:middle line:84%
not only how to drive the
car but also to figure

00:02:50.730 --> 00:02:55.620 align:middle line:84%
out and predict what potential
drivers around the car

00:02:55.620 --> 00:02:56.630 align:middle line:90%
are going to do.

00:02:56.630 --> 00:02:59.640 align:middle line:84%
And that's almost a
gigabyte a second of data

00:02:59.640 --> 00:03:01.202 align:middle line:90%
that that car is processing.

00:03:01.202 --> 00:03:01.704 align:middle line:90%


00:03:01.704 --> 00:03:04.560 align:middle line:84%
ANDY RAVENNA: One of the
things that they're working on

00:03:04.560 --> 00:03:09.150 align:middle line:84%
is, for example, the
scanning of tumors.

00:03:09.150 --> 00:03:12.150 align:middle line:84%
Can you just imagine one day
where they could actually

00:03:12.150 --> 00:03:16.180 align:middle line:84%
do a scan and determine whether
a tumor is benign or not

00:03:16.180 --> 00:03:20.176 align:middle line:84%
instead of having to go in
and take a sample every time

00:03:20.176 --> 00:03:22.120 align:middle line:84%
they're trying to
figure out what's

00:03:22.120 --> 00:03:23.564 align:middle line:90%
going on inside your body?

00:03:23.564 --> 00:03:25.152 align:middle line:90%
It's very Star Trek, right?

00:03:25.152 --> 00:03:25.654 align:middle line:90%


00:03:25.654 --> 00:03:27.890 align:middle line:84%
LISA LOFTIS: And if
you look at the number

00:03:27.890 --> 00:03:30.360 align:middle line:84%
of continuous
streaming information

00:03:30.360 --> 00:03:33.080 align:middle line:84%
from IoT, from the
beacons and sensors

00:03:33.080 --> 00:03:36.200 align:middle line:84%
it gives us the ability to
understand our environment

00:03:36.200 --> 00:03:38.890 align:middle line:84%
much more intimately than
we ever could before.

00:03:38.890 --> 00:03:41.190 align:middle line:90%
And that's what AI and ML need.

00:03:41.190 --> 00:03:42.980 align:middle line:90%
They need granular data.

00:03:42.980 --> 00:03:45.200 align:middle line:84%
They need very large
volumes of data.

00:03:45.200 --> 00:03:48.220 align:middle line:84%
And they need extremely
diverse data sources

00:03:48.220 --> 00:03:50.544 align:middle line:84%
to be able to find the
patterns and learn.

00:03:50.544 --> 00:03:51.046 align:middle line:90%


00:03:51.046 --> 00:03:54.550 align:middle line:84%
KIRK BORNE: So if we can have a
chatbot tell us in the morning,

00:03:54.550 --> 00:03:57.090 align:middle line:90%
tell me the latest news.

00:03:57.090 --> 00:03:57.650 align:middle line:90%
Okay.

00:03:57.650 --> 00:03:59.428 align:middle line:90%
So it can read the stock market.

00:03:59.428 --> 00:04:02.332 align:middle line:84%
It can read all the financial
numbers that we need

00:04:02.332 --> 00:04:05.110 align:middle line:84%
and actually sort of not
just tell you the numbers

00:04:05.110 --> 00:04:07.666 align:middle line:84%
but tell you the story
behind the numbers in a

00:04:07.666 --> 00:04:09.440 align:middle line:90%
in a human interpretable way.

00:04:09.440 --> 00:04:10.740 align:middle line:90%
I think that's really valuable.

00:04:10.740 --> 00:04:12.530 align:middle line:84%
When I think about
what's new and coming up

00:04:12.530 --> 00:04:15.363 align:middle line:84%
I don't know if so much
the mathematics is changing

00:04:15.363 --> 00:04:17.279 align:middle line:90%
but the types of things we do.

00:04:17.279 --> 00:04:18.470 align:middle line:90%
There you have it.

00:04:18.470 --> 00:04:21.769 align:middle line:84%
AI is the science of
computers emulating humans

00:04:21.769 --> 00:04:24.300 align:middle line:84%
and machine learning is
the method behind how

00:04:24.300 --> 00:04:26.170 align:middle line:90%
machines learn from data.

00:04:26.170 --> 00:04:29.850 align:middle line:84%
There are so many problems
to solve based on data.

00:04:29.850 --> 00:04:32.910 align:middle line:84%
Why not let an algorithm
take care of a few?

00:04:32.910 --> 00:04:43.456 align:middle line:90%