WEBVTT
Kind: captions
Language: en

00:00:00.000 --> 00:00:01.960 align:start position:0%
 
you

00:00:01.960 --> 00:00:01.970 align:start position:0%
you
 

00:00:01.970 --> 00:00:07.280 align:start position:0%
you
[Music]

00:00:07.280 --> 00:00:07.290 align:start position:0%
 
 

00:00:07.290 --> 00:00:19.570 align:start position:0%
 
you

00:00:19.570 --> 00:00:19.580 align:start position:0%
 
 

00:00:19.580 --> 00:00:26.690 align:start position:0%
 
um<00:00:20.600><c> in</c><00:00:21.600><c> general</c><00:00:22.230><c> icai</c><00:00:22.830><c> as</c><00:00:25.160><c> software</c><00:00:26.160><c> that</c><00:00:26.490><c> uses</c>

00:00:26.690 --> 00:00:26.700 align:start position:0%
um in general icai as software that uses
 

00:00:26.700 --> 00:00:29.300 align:start position:0%
um in general icai as software that uses
statistics<00:00:27.510><c> in</c><00:00:27.750><c> order</c><00:00:28.560><c> to</c><00:00:28.740><c> make</c><00:00:28.890><c> intelligent</c>

00:00:29.300 --> 00:00:29.310 align:start position:0%
statistics in order to make intelligent
 

00:00:29.310 --> 00:00:35.360 align:start position:0%
statistics in order to make intelligent
decisions<00:00:29.610><c> and</c><00:00:31.189><c> specifics</c><00:00:32.189><c> in</c><00:00:32.310><c> my</c><00:00:32.430><c> role</c><00:00:32.730><c> i</c><00:00:34.370><c> SAT</c>

00:00:35.360 --> 00:00:35.370 align:start position:0%
decisions and specifics in my role i SAT
 

00:00:35.370 --> 00:00:39.319 align:start position:0%
decisions and specifics in my role i SAT
gigster<00:00:35.670><c> we</c><00:00:36.300><c> use</c><00:00:36.480><c> AI</c><00:00:36.690><c> in</c><00:00:37.290><c> order</c><00:00:37.829><c> to</c><00:00:38.329><c> match</c>

00:00:39.319 --> 00:00:39.329 align:start position:0%
gigster we use AI in order to match
 

00:00:39.329 --> 00:00:42.560 align:start position:0%
gigster we use AI in order to match
workers<00:00:39.870><c> to</c><00:00:40.230><c> gigs</c><00:00:40.880><c> automatically</c><00:00:41.880><c> price</c><00:00:42.180><c> gigs</c>

00:00:42.560 --> 00:00:42.570 align:start position:0%
workers to gigs automatically price gigs
 

00:00:42.570 --> 00:00:45.080 align:start position:0%
workers to gigs automatically price gigs
and<00:00:43.130><c> predict</c><00:00:44.130><c> when</c><00:00:44.370><c> gigs</c><00:00:44.670><c> are</c><00:00:44.850><c> going</c>

00:00:45.080 --> 00:00:45.090 align:start position:0%
and predict when gigs are going
 

00:00:45.090 --> 00:00:47.090 align:start position:0%
and predict when gigs are going
escalated<00:00:45.660><c> are</c><00:00:45.840><c> going</c><00:00:46.230><c> to</c><00:00:46.320><c> get</c><00:00:46.440><c> escalated</c><00:00:46.739><c> by</c>

00:00:47.090 --> 00:00:47.100 align:start position:0%
escalated are going to get escalated by
 

00:00:47.100 --> 00:00:49.310 align:start position:0%
escalated are going to get escalated by
monitoring<00:00:47.550><c> data</c><00:00:47.940><c> about</c><00:00:48.390><c> the</c><00:00:48.810><c> development</c>

00:00:49.310 --> 00:00:49.320 align:start position:0%
monitoring data about the development
 

00:00:49.320 --> 00:00:51.560 align:start position:0%
monitoring data about the development
lifecycle<00:00:49.470><c> and</c><00:00:50.250><c> client</c><00:00:50.760><c> developer</c>

00:00:51.560 --> 00:00:51.570 align:start position:0%
lifecycle and client developer
 

00:00:51.570 --> 00:00:58.150 align:start position:0%
lifecycle and client developer
communications<00:00:52.280><c> that</c><00:00:53.280><c> sort</c><00:00:53.520><c> of</c><00:00:53.550><c> thing</c><00:00:53.610><c> I</c>

00:00:58.150 --> 00:00:58.160 align:start position:0%
 
 

00:00:58.160 --> 00:01:00.850 align:start position:0%
 
think<00:00:59.160><c> AI</c><00:00:59.340><c> has</c><00:00:59.550><c> been</c><00:00:59.730><c> very</c><00:00:59.970><c> successful</c><00:01:00.180><c> in</c>

00:01:00.850 --> 00:01:00.860 align:start position:0%
think AI has been very successful in
 

00:01:00.860 --> 00:01:05.660 align:start position:0%
think AI has been very successful in
vision<00:01:01.860><c> in</c><00:01:02.330><c> self-driving</c><00:01:03.330><c> cars</c><00:01:03.720><c> in</c><00:01:04.110><c> text</c><00:01:04.979><c> so</c>

00:01:05.660 --> 00:01:05.670 align:start position:0%
vision in self-driving cars in text so
 

00:01:05.670 --> 00:01:06.859 align:start position:0%
vision in self-driving cars in text so
these<00:01:05.939><c> are</c><00:01:06.000><c> mediums</c><00:01:06.479><c> which</c><00:01:06.630><c> have</c><00:01:06.750><c> been</c>

00:01:06.859 --> 00:01:06.869 align:start position:0%
these are mediums which have been
 

00:01:06.869 --> 00:01:09.050 align:start position:0%
these are mediums which have been
studied<00:01:07.200><c> for</c><00:01:07.229><c> you</c><00:01:07.650><c> know</c><00:01:07.770><c> tens</c><00:01:08.130><c> of</c><00:01:08.220><c> years</c><00:01:08.280><c> and</c><00:01:08.729><c> I</c>

00:01:09.050 --> 00:01:09.060 align:start position:0%
studied for you know tens of years and I
 

00:01:09.060 --> 00:01:10.910 align:start position:0%
studied for you know tens of years and I
think<00:01:09.090><c> that</c><00:01:09.509><c> AI</c><00:01:09.840><c> has</c><00:01:10.170><c> really</c><00:01:10.560><c> been</c><00:01:10.710><c> hitting</c>

00:01:10.910 --> 00:01:10.920 align:start position:0%
think that AI has really been hitting
 

00:01:10.920 --> 00:01:12.200 align:start position:0%
think that AI has really been hitting
starting<00:01:11.310><c> to</c><00:01:11.369><c> hit</c><00:01:11.490><c> the</c><00:01:11.549><c> performance</c><00:01:12.000><c> peak</c>

00:01:12.200 --> 00:01:12.210 align:start position:0%
starting to hit the performance peak
 

00:01:12.210 --> 00:01:14.420 align:start position:0%
starting to hit the performance peak
especially<00:01:12.689><c> with</c><00:01:12.780><c> deep</c><00:01:12.960><c> learning</c><00:01:13.400><c> where</c><00:01:14.400><c> I</c>

00:01:14.420 --> 00:01:14.430 align:start position:0%
especially with deep learning where I
 

00:01:14.430 --> 00:01:16.520 align:start position:0%
especially with deep learning where I
think<00:01:14.700><c> AI</c><00:01:14.850><c> can</c><00:01:15.060><c> be</c><00:01:15.150><c> improved</c><00:01:15.450><c> is</c><00:01:15.780><c> an</c><00:01:16.350><c> other</c>

00:01:16.520 --> 00:01:16.530 align:start position:0%
think AI can be improved is an other
 

00:01:16.530 --> 00:01:18.560 align:start position:0%
think AI can be improved is an other
problem<00:01:16.920><c> domains</c><00:01:17.220><c> so</c><00:01:17.549><c> for</c><00:01:17.850><c> example</c><00:01:17.880><c> what</c><00:01:18.540><c> I</c>

00:01:18.560 --> 00:01:18.570 align:start position:0%
problem domains so for example what I
 

00:01:18.570 --> 00:01:20.330 align:start position:0%
problem domains so for example what I
just<00:01:18.780><c> gave</c><00:01:18.900><c> a</c><00:01:18.960><c> talk</c><00:01:19.200><c> on</c><00:01:19.229><c> was</c><00:01:19.619><c> about</c><00:01:19.829><c> using</c><00:01:20.159><c> AI</c>

00:01:20.330 --> 00:01:20.340 align:start position:0%
just gave a talk on was about using AI
 

00:01:20.340 --> 00:01:23.480 align:start position:0%
just gave a talk on was about using AI
in<00:01:20.729><c> generative</c><00:01:21.149><c> modeling</c><00:01:21.630><c> of</c><00:01:21.749><c> music</c><00:01:22.490><c> we're</c>

00:01:23.480 --> 00:01:23.490 align:start position:0%
in generative modeling of music we're
 

00:01:23.490 --> 00:01:25.669 align:start position:0%
in generative modeling of music we're
seeing<00:01:23.700><c> AI</c><00:01:23.909><c> being</c><00:01:24.390><c> used</c><00:01:24.630><c> for</c><00:01:24.749><c> art</c><00:01:25.049><c> now</c><00:01:25.259><c> as</c><00:01:25.499><c> well</c>

00:01:25.669 --> 00:01:25.679 align:start position:0%
seeing AI being used for art now as well
 

00:01:25.679 --> 00:01:27.739 align:start position:0%
seeing AI being used for art now as well
so<00:01:25.920><c> generative</c><00:01:26.369><c> models</c><00:01:26.700><c> are</c><00:01:27.060><c> big</c><00:01:27.299><c> field</c><00:01:27.539><c> I</c>

00:01:27.739 --> 00:01:27.749 align:start position:0%
so generative models are big field I
 

00:01:27.749 --> 00:01:29.809 align:start position:0%
so generative models are big field I
also<00:01:28.590><c> think</c><00:01:28.770><c> AI</c><00:01:28.979><c> can</c><00:01:29.249><c> be</c><00:01:29.340><c> improved</c><00:01:29.579><c> to</c>

00:01:29.809 --> 00:01:29.819 align:start position:0%
also think AI can be improved to
 

00:01:29.819 --> 00:01:32.200 align:start position:0%
also think AI can be improved to
actually<00:01:30.179><c> handle</c><00:01:30.359><c> large</c><00:01:30.810><c> amounts</c><00:01:30.899><c> of</c><00:01:31.229><c> data</c><00:01:31.409><c> so</c>

00:01:32.200 --> 00:01:32.210 align:start position:0%
actually handle large amounts of data so
 

00:01:32.210 --> 00:01:34.819 align:start position:0%
actually handle large amounts of data so
data<00:01:33.210><c> sets</c><00:01:33.420><c> are</c><00:01:33.479><c> getting</c><00:01:33.719><c> really</c><00:01:33.960><c> big</c><00:01:34.139><c> and</c><00:01:34.380><c> you</c>

00:01:34.819 --> 00:01:34.829 align:start position:0%
data sets are getting really big and you
 

00:01:34.829 --> 00:01:36.379 align:start position:0%
data sets are getting really big and you
know<00:01:34.920><c> the</c><00:01:35.130><c> most</c><00:01:35.310><c> sophisticated</c><00:01:35.490><c> of</c><00:01:36.090><c> models</c>

00:01:36.379 --> 00:01:36.389 align:start position:0%
know the most sophisticated of models
 

00:01:36.389 --> 00:01:38.569 align:start position:0%
know the most sophisticated of models
themselves<00:01:36.569><c> take</c><00:01:37.170><c> hours</c><00:01:37.560><c> maybe</c><00:01:37.770><c> days</c><00:01:37.889><c> to</c><00:01:38.369><c> run</c>

00:01:38.569 --> 00:01:38.579 align:start position:0%
themselves take hours maybe days to run
 

00:01:38.579 --> 00:01:40.760 align:start position:0%
themselves take hours maybe days to run
I<00:01:38.789><c> think</c><00:01:39.509><c> finding</c><00:01:40.079><c> computationally</c>

00:01:40.760 --> 00:01:40.770 align:start position:0%
I think finding computationally
 

00:01:40.770 --> 00:01:42.980 align:start position:0%
I think finding computationally
tractable<00:01:40.799><c> methods</c><00:01:41.639><c> and</c><00:01:41.850><c> smarter</c><00:01:42.450><c> algorithms</c>

00:01:42.980 --> 00:01:42.990 align:start position:0%
tractable methods and smarter algorithms
 

00:01:42.990 --> 00:01:45.379 align:start position:0%
tractable methods and smarter algorithms
to<00:01:43.229><c> process</c><00:01:43.560><c> larger</c><00:01:43.889><c> data</c><00:01:44.100><c> at</c><00:01:44.310><c> faster</c><00:01:45.090><c> scales</c>

00:01:45.379 --> 00:01:45.389 align:start position:0%
to process larger data at faster scales
 

00:01:45.389 --> 00:01:47.899 align:start position:0%
to process larger data at faster scales
is<00:01:45.659><c> really</c><00:01:46.350><c> where</c><00:01:46.469><c> the</c><00:01:46.590><c> next</c><00:01:46.859><c> frontier</c><00:01:47.609><c> for</c><00:01:47.639><c> AI</c>

00:01:47.899 --> 00:01:47.909 align:start position:0%
is really where the next frontier for AI
 

00:01:47.909 --> 00:01:53.370 align:start position:0%
is really where the next frontier for AI
is

00:01:53.370 --> 00:01:53.380 align:start position:0%
 
 

00:01:53.380 --> 00:01:55.780 align:start position:0%
 
the<00:01:54.380><c> problem</c><00:01:54.710><c> I</c><00:01:54.770><c> want</c><00:01:54.950><c> to</c><00:01:55.010><c> I've</c><00:01:55.250><c> been</c><00:01:55.430><c> focused</c>

00:01:55.780 --> 00:01:55.790 align:start position:0%
the problem I want to I've been focused
 

00:01:55.790 --> 00:01:58.840 align:start position:0%
the problem I want to I've been focused
recently<00:01:56.330><c> on</c><00:01:56.480><c> is</c><00:01:57.190><c> deploying</c><00:01:58.190><c> a</c><00:01:58.460><c> I</c><00:01:58.490><c> in</c>

00:01:58.840 --> 00:01:58.850 align:start position:0%
recently on is deploying a I in
 

00:01:58.850 --> 00:02:01.510 align:start position:0%
recently on is deploying a I in
production<00:01:59.360><c> so</c><00:01:59.840><c> a</c><00:02:00.170><c> I</c><00:02:00.560><c> you</c><00:02:00.740><c> know</c><00:02:01.130><c> there's</c><00:02:01.310><c> a</c><00:02:01.370><c> lot</c>

00:02:01.510 --> 00:02:01.520 align:start position:0%
production so a I you know there's a lot
 

00:02:01.520 --> 00:02:04.000 align:start position:0%
production so a I you know there's a lot
of<00:02:01.550><c> papers</c><00:02:02.000><c> about</c><00:02:02.240><c> it</c><00:02:02.480><c> we</c><00:02:03.260><c> see</c><00:02:03.410><c> a</c><00:02:03.440><c> lot</c><00:02:03.590><c> of</c><00:02:03.650><c> great</c>

00:02:04.000 --> 00:02:04.010 align:start position:0%
of papers about it we see a lot of great
 

00:02:04.010 --> 00:02:06.640 align:start position:0%
of papers about it we see a lot of great
empirical<00:02:04.460><c> results</c><00:02:04.580><c> inside</c><00:02:05.360><c> of</c><00:02:05.450><c> academia</c><00:02:05.780><c> but</c>

00:02:06.640 --> 00:02:06.650 align:start position:0%
empirical results inside of academia but
 

00:02:06.650 --> 00:02:08.020 align:start position:0%
empirical results inside of academia but
when<00:02:06.860><c> it</c><00:02:06.950><c> comes</c><00:02:07.100><c> to</c><00:02:07.250><c> industry</c><00:02:07.700><c> there's</c><00:02:07.940><c> a</c><00:02:08.000><c> lot</c>

00:02:08.020 --> 00:02:08.030 align:start position:0%
when it comes to industry there's a lot
 

00:02:08.030 --> 00:02:10.000 align:start position:0%
when it comes to industry there's a lot
of<00:02:08.119><c> concerns</c><00:02:08.600><c> that</c><00:02:08.869><c> academic</c><00:02:09.350><c> publications</c>

00:02:10.000 --> 00:02:10.010 align:start position:0%
of concerns that academic publications
 

00:02:10.010 --> 00:02:11.920 align:start position:0%
of concerns that academic publications
don't<00:02:10.220><c> address</c><00:02:10.430><c> so</c><00:02:11.150><c> some</c><00:02:11.330><c> examples</c><00:02:11.360><c> of</c><00:02:11.780><c> this</c>

00:02:11.920 --> 00:02:11.930 align:start position:0%
don't address so some examples of this
 

00:02:11.930 --> 00:02:14.020 align:start position:0%
don't address so some examples of this
include<00:02:12.200><c> versioning</c><00:02:13.040><c> data</c><00:02:13.280><c> sets</c><00:02:13.550><c> that</c><00:02:13.640><c> change</c>

00:02:14.020 --> 00:02:14.030 align:start position:0%
include versioning data sets that change
 

00:02:14.030 --> 00:02:16.300 align:start position:0%
include versioning data sets that change
over<00:02:14.390><c> time</c><00:02:14.650><c> versioning</c><00:02:15.650><c> models</c><00:02:15.980><c> that</c><00:02:16.100><c> change</c>

00:02:16.300 --> 00:02:16.310 align:start position:0%
over time versioning models that change
 

00:02:16.310 --> 00:02:18.070 align:start position:0%
over time versioning models that change
over<00:02:16.550><c> time</c><00:02:16.730><c> how</c><00:02:17.000><c> do</c><00:02:17.060><c> i</c><00:02:17.180><c> store</c><00:02:17.570><c> these</c><00:02:17.810><c> very</c>

00:02:18.070 --> 00:02:18.080 align:start position:0%
over time how do i store these very
 

00:02:18.080 --> 00:02:20.140 align:start position:0%
over time how do i store these very
large<00:02:18.170><c> models</c><00:02:18.830><c> and</c><00:02:19.010><c> how</c><00:02:19.340><c> do</c><00:02:19.400><c> i</c><00:02:19.490><c> deploy</c><00:02:19.850><c> them</c>

00:02:20.140 --> 00:02:20.150 align:start position:0%
large models and how do i deploy them
 

00:02:20.150 --> 00:02:21.940 align:start position:0%
large models and how do i deploy them
how<00:02:20.630><c> do</c><00:02:20.690><c> I</c><00:02:20.780><c> know</c><00:02:20.870><c> my</c><00:02:20.990><c> model</c><00:02:21.410><c> is</c><00:02:21.440><c> actually</c><00:02:21.620><c> doing</c>

00:02:21.940 --> 00:02:21.950 align:start position:0%
how do I know my model is actually doing
 

00:02:21.950 --> 00:02:23.800 align:start position:0%
how do I know my model is actually doing
better<00:02:22.070><c> and</c><00:02:22.460><c> it's</c><00:02:22.700><c> not</c><00:02:22.850><c> doing</c><00:02:23.090><c> better</c><00:02:23.270><c> can</c><00:02:23.720><c> I</c>

00:02:23.800 --> 00:02:23.810 align:start position:0%
better and it's not doing better can I
 

00:02:23.810 --> 00:02:25.270 align:start position:0%
better and it's not doing better can I
roll<00:02:24.050><c> back</c><00:02:24.260><c> in</c><00:02:24.500><c> a</c><00:02:24.590><c> cheap</c><00:02:24.800><c> and</c><00:02:24.950><c> efficient</c>

00:02:25.270 --> 00:02:25.280 align:start position:0%
roll back in a cheap and efficient
 

00:02:25.280 --> 00:02:26.920 align:start position:0%
roll back in a cheap and efficient
manner<00:02:25.310><c> so</c><00:02:25.730><c> really</c><00:02:26.030><c> a</c><00:02:26.060><c> lot</c><00:02:26.180><c> of</c><00:02:26.300><c> ergonomics</c>

00:02:26.920 --> 00:02:26.930 align:start position:0%
manner so really a lot of ergonomics
 

00:02:26.930 --> 00:02:33.970 align:start position:0%
manner so really a lot of ergonomics
about<00:02:27.170><c> practicing</c><00:02:27.770><c> AI</c><00:02:32.440><c> my</c><00:02:33.440><c> favorite</c><00:02:33.800><c> thing</c>

00:02:33.970 --> 00:02:33.980 align:start position:0%
about practicing AI my favorite thing
 

00:02:33.980 --> 00:02:36.880 align:start position:0%
about practicing AI my favorite thing
about<00:02:34.010><c> AI</c><00:02:34.460><c> by</c><00:02:34.730><c> the</c><00:02:34.790><c> bay</c><00:02:35.060><c> are</c><00:02:35.330><c> the</c><00:02:35.870><c> speakers</c><00:02:36.110><c> the</c>

00:02:36.880 --> 00:02:36.890 align:start position:0%
about AI by the bay are the speakers the
 

00:02:36.890 --> 00:02:39.610 align:start position:0%
about AI by the bay are the speakers the
lineup<00:02:37.250><c> this</c><00:02:37.370><c> year</c><00:02:37.430><c> is</c><00:02:37.850><c> amazing</c><00:02:38.570><c> I'm</c><00:02:39.230><c> super</c>

00:02:39.610 --> 00:02:39.620 align:start position:0%
lineup this year is amazing I'm super
 

00:02:39.620 --> 00:02:41.650 align:start position:0%
lineup this year is amazing I'm super
excited<00:02:39.740><c> to</c><00:02:40.130><c> see</c><00:02:40.280><c> like</c><00:02:40.850><c> Stuart</c><00:02:41.270><c> Russell</c>

00:02:41.650 --> 00:02:41.660 align:start position:0%
excited to see like Stuart Russell
 

00:02:41.660 --> 00:02:44.140 align:start position:0%
excited to see like Stuart Russell
George<00:02:42.380><c> Hut's</c><00:02:42.740><c> like</c><00:02:43.610><c> there's</c><00:02:43.850><c> some</c><00:02:44.000><c> really</c>

00:02:44.140 --> 00:02:44.150 align:start position:0%
George Hut's like there's some really
 

00:02:44.150 --> 00:02:45.970 align:start position:0%
George Hut's like there's some really
great<00:02:44.390><c> people</c><00:02:44.570><c> here</c><00:02:44.870><c> I'm</c><00:02:45.020><c> impressed</c><00:02:45.410><c> that</c><00:02:45.770><c> you</c>

00:02:45.970 --> 00:02:45.980 align:start position:0%
great people here I'm impressed that you
 

00:02:45.980 --> 00:02:47.140 align:start position:0%
great people here I'm impressed that you
were<00:02:46.010><c> able</c><00:02:46.100><c> to</c><00:02:46.340><c> get</c><00:02:46.459><c> all</c><00:02:46.580><c> of</c><00:02:46.670><c> them</c><00:02:46.760><c> here</c><00:02:46.970><c> and</c><00:02:47.060><c> I</c>

00:02:47.140 --> 00:02:47.150 align:start position:0%
were able to get all of them here and I
 

00:02:47.150 --> 00:02:56.020 align:start position:0%
were able to get all of them here and I
can't<00:02:47.390><c> wait</c><00:02:47.570><c> to</c><00:02:47.690><c> go</c><00:02:47.810><c> see</c><00:02:47.959><c> them</c><00:02:48.080><c> talk</c>

00:02:56.020 --> 00:02:56.030 align:start position:0%
 
 

00:02:56.030 --> 00:03:04.260 align:start position:0%
 
you

00:03:04.260 --> 00:03:04.270 align:start position:0%
 
 

00:03:04.270 --> 00:03:06.330 align:start position:0%
 
you