WEBVTT
Kind: captions
Language: en

00:00:10.480 --> 00:00:13.110 align:start position:0%
 
my<00:00:10.599><c> name</c><00:00:10.759><c> is</c><00:00:10.880><c> Nick</c><00:00:11.120><c> Hill</c><00:00:11.840><c> I</c><00:00:12.320><c> work</c><00:00:12.559><c> for</c><00:00:12.719><c> IBM</c>

00:00:13.110 --> 00:00:13.120 align:start position:0%
my name is Nick Hill I work for IBM
 

00:00:13.120 --> 00:00:15.990 align:start position:0%
my name is Nick Hill I work for IBM
research<00:00:14.120><c> I</c><00:00:14.440><c> work</c><00:00:14.679><c> on</c><00:00:15.040><c> infrastructure</c><00:00:15.679><c> for</c>

00:00:15.990 --> 00:00:16.000 align:start position:0%
research I work on infrastructure for
 

00:00:16.000 --> 00:00:22.790 align:start position:0%
research I work on infrastructure for
serving<00:00:16.720><c> uh</c><00:00:16.840><c> large</c><00:00:17.160><c> language</c>

00:00:22.790 --> 00:00:22.800 align:start position:0%
 
 

00:00:22.800 --> 00:00:26.189 align:start position:0%
 
models<00:00:23.880><c> so</c><00:00:24.880><c> I</c><00:00:25.160><c> don't</c><00:00:25.439><c> work</c><00:00:25.640><c> on</c><00:00:25.920><c> the</c>

00:00:26.189 --> 00:00:26.199 align:start position:0%
models so I don't work on the
 

00:00:26.199 --> 00:00:28.310 align:start position:0%
models so I don't work on the
applications<00:00:26.800><c> as</c><00:00:26.960><c> such</c><00:00:27.199><c> I</c><00:00:27.320><c> work</c><00:00:27.640><c> more</c><00:00:27.800><c> on</c><00:00:28.000><c> the</c>

00:00:28.310 --> 00:00:28.320 align:start position:0%
applications as such I work more on the
 

00:00:28.320 --> 00:00:31.189 align:start position:0%
applications as such I work more on the
the<00:00:28.640><c> the</c><00:00:28.800><c> actual</c><00:00:29.119><c> platform</c><00:00:29.840><c> um</c><00:00:30.080><c> so</c><00:00:30.400><c> there</c><00:00:30.679><c> AI</c>

00:00:31.189 --> 00:00:31.199 align:start position:0%
the the actual platform um so there AI
 

00:00:31.199 --> 00:00:32.830 align:start position:0%
the the actual platform um so there AI
I'm<00:00:31.359><c> supporting</c><00:00:31.840><c> all</c><00:00:31.960><c> of</c><00:00:32.119><c> the</c><00:00:32.279><c> other</c><00:00:32.520><c> sort</c><00:00:32.680><c> of</c>

00:00:32.830 --> 00:00:32.840 align:start position:0%
I'm supporting all of the other sort of
 

00:00:32.840 --> 00:00:36.590 align:start position:0%
I'm supporting all of the other sort of
AI<00:00:33.200><c> applications</c><00:00:33.680><c> that</c><00:00:33.800><c> are</c><00:00:33.920><c> built</c><00:00:34.120><c> on</c><00:00:34.399><c> top</c><00:00:35.600><c> um</c>

00:00:36.590 --> 00:00:36.600 align:start position:0%
AI applications that are built on top um
 

00:00:36.600 --> 00:00:39.950 align:start position:0%
AI applications that are built on top um
really<00:00:36.879><c> looking</c><00:00:37.200><c> to</c><00:00:37.520><c> maximize</c><00:00:38.320><c> the</c><00:00:39.280><c> uh</c><00:00:39.440><c> use</c><00:00:39.680><c> of</c>

00:00:39.950 --> 00:00:39.960 align:start position:0%
really looking to maximize the uh use of
 

00:00:39.960 --> 00:00:43.709 align:start position:0%
really looking to maximize the uh use of
the<00:00:40.200><c> the</c><00:00:40.320><c> gpus</c><00:00:40.920><c> and</c><00:00:41.200><c> get</c><00:00:41.399><c> as</c><00:00:41.559><c> many</c><00:00:42.520><c> um</c><00:00:43.280><c> requests</c>

00:00:43.709 --> 00:00:43.719 align:start position:0%
the the gpus and get as many um requests
 

00:00:43.719 --> 00:00:51.670 align:start position:0%
the the gpus and get as many um requests
through<00:00:44.039><c> as</c><00:00:44.160><c> we</c><00:00:44.320><c> can</c><00:00:44.879><c> for</c><00:00:45.879><c> a</c><00:00:46.079><c> small</c>

00:00:51.670 --> 00:00:51.680 align:start position:0%
 
 

00:00:51.680 --> 00:00:54.869 align:start position:0%
 
footprint<00:00:52.680><c> so</c><00:00:53.440><c> I</c><00:00:53.640><c> talked</c><00:00:53.960><c> a</c><00:00:54.079><c> little</c><00:00:54.239><c> bit</c><00:00:54.600><c> about</c>

00:00:54.869 --> 00:00:54.879 align:start position:0%
footprint so I talked a little bit about
 

00:00:54.879 --> 00:00:57.389 align:start position:0%
footprint so I talked a little bit about
this<00:00:55.039><c> in</c><00:00:55.280><c> in</c><00:00:55.399><c> the</c><00:00:55.520><c> talk</c><00:00:55.719><c> I</c><00:00:55.879><c> just</c><00:00:56.079><c> gave</c><00:00:56.440><c> but</c><00:00:56.840><c> uh</c>

00:00:57.389 --> 00:00:57.399 align:start position:0%
this in in the talk I just gave but uh
 

00:00:57.399 --> 00:01:00.709 align:start position:0%
this in in the talk I just gave but uh
there's<00:00:58.399><c> a</c><00:00:58.559><c> lot</c><00:00:58.719><c> of</c><00:00:58.960><c> change</c><00:00:59.320><c> in</c><00:00:59.519><c> this</c><00:00:59.960><c> space</c><00:01:00.199><c> of</c>

00:01:00.709 --> 00:01:00.719 align:start position:0%
there's a lot of change in this space of
 

00:01:00.719 --> 00:01:03.630 align:start position:0%
there's a lot of change in this space of
uh<00:01:00.800><c> llm</c><00:01:01.199><c> inferencing</c><00:01:01.800><c> right</c><00:01:01.960><c> now</c><00:01:02.359><c> like</c><00:01:03.359><c> every</c>

00:01:03.630 --> 00:01:03.640 align:start position:0%
uh llm inferencing right now like every
 

00:01:03.640 --> 00:01:04.910 align:start position:0%
uh llm inferencing right now like every
other<00:01:03.879><c> week</c><00:01:04.080><c> it</c><00:01:04.199><c> seems</c><00:01:04.439><c> there's</c><00:01:04.600><c> a</c><00:01:04.720><c> new</c>

00:01:04.910 --> 00:01:04.920 align:start position:0%
other week it seems there's a new
 

00:01:04.920 --> 00:01:06.510 align:start position:0%
other week it seems there's a new
project<00:01:05.239><c> coming</c><00:01:05.479><c> out</c><00:01:05.840><c> and</c><00:01:06.000><c> they're</c>

00:01:06.510 --> 00:01:06.520 align:start position:0%
project coming out and they're
 

00:01:06.520 --> 00:01:08.469 align:start position:0%
project coming out and they're
incorporating<00:01:07.280><c> the</c><00:01:07.400><c> techniques</c><00:01:07.799><c> of</c><00:01:08.200><c> other</c>

00:01:08.469 --> 00:01:08.479 align:start position:0%
incorporating the techniques of other
 

00:01:08.479 --> 00:01:10.749 align:start position:0%
incorporating the techniques of other
projects<00:01:08.920><c> and</c><00:01:09.159><c> adding</c><00:01:09.479><c> their</c><00:01:09.640><c> own</c><00:01:10.119><c> and</c><00:01:10.439><c> saying</c>

00:01:10.749 --> 00:01:10.759 align:start position:0%
projects and adding their own and saying
 

00:01:10.759 --> 00:01:13.630 align:start position:0%
projects and adding their own and saying
they're<00:01:11.000><c> much</c><00:01:11.200><c> faster</c><00:01:11.759><c> so</c><00:01:12.600><c> um</c><00:01:12.960><c> our</c><00:01:13.159><c> sort</c><00:01:13.360><c> of</c>

00:01:13.630 --> 00:01:13.640 align:start position:0%
they're much faster so um our sort of
 

00:01:13.640 --> 00:01:15.429 align:start position:0%
they're much faster so um our sort of
main<00:01:14.000><c> I</c><00:01:14.080><c> guess</c><00:01:14.240><c> ongoing</c><00:01:14.680><c> project</c><00:01:15.000><c> is</c><00:01:15.159><c> to</c><00:01:15.280><c> try</c>

00:01:15.429 --> 00:01:15.439 align:start position:0%
main I guess ongoing project is to try
 

00:01:15.439 --> 00:01:16.789 align:start position:0%
main I guess ongoing project is to try
and<00:01:15.560><c> keep</c><00:01:15.720><c> tabs</c><00:01:16.000><c> and</c><00:01:16.119><c> all</c><00:01:16.280><c> these</c><00:01:16.439><c> things</c><00:01:16.640><c> and</c>

00:01:16.789 --> 00:01:16.799 align:start position:0%
and keep tabs and all these things and
 

00:01:16.799 --> 00:01:19.270 align:start position:0%
and keep tabs and all these things and
figure<00:01:17.040><c> out</c><00:01:17.320><c> okay</c><00:01:17.520><c> which</c><00:01:17.680><c> ones</c><00:01:18.400><c> can</c><00:01:18.600><c> we</c><00:01:19.119><c> uh</c>

00:01:19.270 --> 00:01:19.280 align:start position:0%
figure out okay which ones can we uh
 

00:01:19.280 --> 00:01:21.149 align:start position:0%
figure out okay which ones can we uh
leverage<00:01:19.840><c> and</c><00:01:20.079><c> where</c><00:01:20.240><c> do</c><00:01:20.400><c> we</c><00:01:20.560><c> sort</c><00:01:20.720><c> of</c><00:01:20.880><c> add</c><00:01:21.040><c> our</c>

00:01:21.149 --> 00:01:21.159 align:start position:0%
leverage and where do we sort of add our
 

00:01:21.159 --> 00:01:24.670 align:start position:0%
leverage and where do we sort of add our
own<00:01:21.439><c> things</c><00:01:21.680><c> in</c><00:01:22.600><c> um</c><00:01:22.880><c> so</c><00:01:23.600><c> yeah</c><00:01:23.759><c> really</c><00:01:24.400><c> you</c><00:01:24.520><c> know</c>

00:01:24.670 --> 00:01:24.680 align:start position:0%
own things in um so yeah really you know
 

00:01:24.680 --> 00:01:28.670 align:start position:0%
own things in um so yeah really you know
looking<00:01:25.000><c> at</c><00:01:25.680><c> uh</c><00:01:26.119><c> how</c><00:01:26.240><c> do</c><00:01:26.360><c> we</c><00:01:26.560><c> keep</c><00:01:26.759><c> up</c><00:01:27.280><c> and</c><00:01:27.680><c> um</c>

00:01:28.670 --> 00:01:28.680 align:start position:0%
looking at uh how do we keep up and um
 

00:01:28.680 --> 00:01:30.469 align:start position:0%
looking at uh how do we keep up and um
and<00:01:29.040><c> which</c><00:01:29.439><c> which</c><00:01:29.600><c> Tech</c><00:01:29.880><c> Technologies</c><00:01:30.240><c> do</c><00:01:30.360><c> we</c>

00:01:30.469 --> 00:01:30.479 align:start position:0%
and which which Tech Technologies do we
 

00:01:30.479 --> 00:01:37.429 align:start position:0%
and which which Tech Technologies do we
choose<00:01:30.720><c> out</c><00:01:30.840><c> of</c><00:01:31.040><c> this</c><00:01:31.200><c> set</c><00:01:31.520><c> that</c><00:01:31.720><c> that</c><00:01:31.840><c> keeps</c>

00:01:37.429 --> 00:01:37.439 align:start position:0%
 
 

00:01:37.439 --> 00:01:40.109 align:start position:0%
 
changing<00:01:38.439><c> I</c><00:01:38.600><c> really</c><00:01:38.799><c> like</c><00:01:39.159><c> the</c><00:01:39.520><c> the</c><00:01:39.680><c> keynote</c>

00:01:40.109 --> 00:01:40.119 align:start position:0%
changing I really like the the keynote
 

00:01:40.119 --> 00:01:43.749 align:start position:0%
changing I really like the the keynote
from<00:01:40.360><c> Chris</c><00:01:40.600><c> latner</c><00:01:41.439><c> um</c><00:01:41.759><c> I</c><00:01:41.880><c> think</c><00:01:42.240><c> the</c><00:01:42.680><c> uh</c><00:01:43.640><c> uh</c>

00:01:43.749 --> 00:01:43.759 align:start position:0%
from Chris latner um I think the uh uh
 

00:01:43.759 --> 00:01:45.990 align:start position:0%
from Chris latner um I think the uh uh
the<00:01:43.880><c> new</c><00:01:44.399><c> the</c><00:01:44.640><c> work</c><00:01:44.799><c> he's</c><00:01:44.960><c> doing</c><00:01:45.240><c> with</c><00:01:45.399><c> the</c><00:01:45.520><c> MOJ</c>

00:01:45.990 --> 00:01:46.000 align:start position:0%
the new the work he's doing with the MOJ
 

00:01:46.000 --> 00:01:49.030 align:start position:0%
the new the work he's doing with the MOJ
language<00:01:46.600><c> is</c><00:01:46.960><c> um</c><00:01:47.759><c> is</c><00:01:48.000><c> really</c><00:01:48.280><c> interesting</c><00:01:48.759><c> and</c>

00:01:49.030 --> 00:01:49.040 align:start position:0%
language is um is really interesting and
 

00:01:49.040 --> 00:01:51.310 align:start position:0%
language is um is really interesting and
yeah<00:01:49.159><c> looking</c><00:01:49.360><c> forward</c><00:01:49.600><c> to</c><00:01:49.680><c> seeing</c><00:01:49.920><c> how</c><00:01:50.320><c> that</c>

00:01:51.310 --> 00:01:51.320 align:start position:0%
yeah looking forward to seeing how that
 

00:01:51.320 --> 00:01:53.109 align:start position:0%
yeah looking forward to seeing how that
that<00:01:51.520><c> pans</c><00:01:51.840><c> out</c><00:01:52.360><c> might</c><00:01:52.520><c> be</c><00:01:52.680><c> something</c><00:01:52.880><c> that</c><00:01:53.000><c> we</c>

00:01:53.109 --> 00:01:53.119 align:start position:0%
that pans out might be something that we
 

00:01:53.119 --> 00:01:57.230 align:start position:0%
that pans out might be something that we
can<00:01:53.240><c> use</c><00:01:53.640><c> in</c><00:01:53.759><c> our</c><00:01:53.960><c> work</c><00:01:54.960><c> uh</c><00:01:55.240><c> at</c><00:01:55.360><c> some</c><00:01:55.560><c> point</c><00:01:56.240><c> um</c>

00:01:57.230 --> 00:01:57.240 align:start position:0%
can use in our work uh at some point um
 

00:01:57.240 --> 00:01:59.350 align:start position:0%
can use in our work uh at some point um
and<00:01:57.719><c> uh</c><00:01:57.920><c> I'm</c><00:01:58.039><c> also</c><00:01:58.280><c> interested</c><00:01:58.600><c> in</c><00:01:58.920><c> yeah</c><00:01:59.159><c> other</c>

00:01:59.350 --> 00:01:59.360 align:start position:0%
and uh I'm also interested in yeah other
 

00:01:59.360 --> 00:02:02.389 align:start position:0%
and uh I'm also interested in yeah other
talks<00:01:59.600><c> on</c><00:01:59.920><c> llms</c><00:02:00.399><c> there's</c><00:02:00.560><c> one</c><00:02:00.759><c> about</c><00:02:01.280><c> pytorch</c>

00:02:02.389 --> 00:02:02.399 align:start position:0%
talks on llms there's one about pytorch
 

00:02:02.399 --> 00:02:05.350 align:start position:0%
talks on llms there's one about pytorch
2.1<00:02:03.399><c> new</c><00:02:03.600><c> features</c><00:02:04.039><c> tomorrow</c><00:02:04.520><c> which</c><00:02:04.960><c> we</c><00:02:05.079><c> use</c><00:02:05.240><c> a</c>

00:02:05.350 --> 00:02:05.360 align:start position:0%
2.1 new features tomorrow which we use a
 

00:02:05.360 --> 00:02:09.389 align:start position:0%
2.1 new features tomorrow which we use a
lot<00:02:05.479><c> of</c><00:02:05.719><c> those</c><00:02:06.719><c> um</c><00:02:07.320><c> and</c><00:02:07.560><c> actually</c><00:02:08.119><c> I</c><00:02:09.039><c> uh</c><00:02:09.239><c> quite</c>

00:02:09.389 --> 00:02:09.399 align:start position:0%
lot of those um and actually I uh quite
 

00:02:09.399 --> 00:02:10.749 align:start position:0%
lot of those um and actually I uh quite
interested<00:02:09.720><c> in</c><00:02:09.840><c> some</c><00:02:09.959><c> of</c><00:02:10.080><c> the</c><00:02:10.160><c> rust</c><00:02:10.399><c> talks</c>

00:02:10.749 --> 00:02:10.759 align:start position:0%
interested in some of the rust talks
 

00:02:10.759 --> 00:02:12.589 align:start position:0%
interested in some of the rust talks
because<00:02:11.080><c> I've</c><00:02:11.200><c> been</c><00:02:11.319><c> using</c><00:02:11.520><c> rust</c><00:02:11.800><c> a</c><00:02:11.879><c> lot</c><00:02:12.040><c> more</c>

00:02:12.589 --> 00:02:12.599 align:start position:0%
because I've been using rust a lot more
 

00:02:12.599 --> 00:02:18.030 align:start position:0%
because I've been using rust a lot more
uh

00:02:18.030 --> 00:02:18.040 align:start position:0%
 
 

00:02:18.040 --> 00:02:20.869 align:start position:0%
 
lately<00:02:19.040><c> I'd</c><00:02:19.200><c> say</c><00:02:19.400><c> it</c><00:02:19.560><c> has</c><00:02:19.720><c> a</c><00:02:20.280><c> different</c><00:02:20.519><c> feel</c>

00:02:20.869 --> 00:02:20.879 align:start position:0%
lately I'd say it has a different feel
 

00:02:20.879 --> 00:02:22.550 align:start position:0%
lately I'd say it has a different feel
to<00:02:21.000><c> a</c><00:02:21.080><c> lot</c><00:02:21.200><c> of</c><00:02:21.319><c> other</c><00:02:21.480><c> conferences</c><00:02:22.120><c> it's</c><00:02:22.360><c> sort</c>

00:02:22.550 --> 00:02:22.560 align:start position:0%
to a lot of other conferences it's sort
 

00:02:22.560 --> 00:02:25.030 align:start position:0%
to a lot of other conferences it's sort
of<00:02:22.840><c> much</c><00:02:23.040><c> more</c><00:02:23.800><c> sort</c><00:02:23.959><c> of</c><00:02:24.080><c> friendly</c><00:02:24.519><c> developer</c>

00:02:25.030 --> 00:02:25.040 align:start position:0%
of much more sort of friendly developer
 

00:02:25.040 --> 00:02:28.509 align:start position:0%
of much more sort of friendly developer
focused<00:02:26.400><c> uh</c><00:02:27.400><c> I</c><00:02:27.560><c> really</c><00:02:27.760><c> like</c><00:02:27.959><c> the</c><00:02:28.080><c> way</c><00:02:28.200><c> it's</c><00:02:28.319><c> a</c>

00:02:28.509 --> 00:02:28.519 align:start position:0%
focused uh I really like the way it's a
 

00:02:28.519 --> 00:02:31.790 align:start position:0%
focused uh I really like the way it's a
mix<00:02:28.760><c> of</c><00:02:29.080><c> different</c><00:02:30.000><c> uh</c><00:02:30.200><c> subject</c><00:02:30.640><c> so</c><00:02:31.160><c> like</c><00:02:31.640><c> I</c>

00:02:31.790 --> 00:02:31.800 align:start position:0%
mix of different uh subject so like I
 

00:02:31.800 --> 00:02:34.350 align:start position:0%
mix of different uh subject so like I
mentioned<00:02:32.200><c> rust</c><00:02:32.720><c> there's</c><00:02:32.879><c> a</c><00:02:33.080><c> you</c><00:02:33.200><c> know</c><00:02:33.879><c> rust</c>

00:02:34.350 --> 00:02:34.360 align:start position:0%
mentioned rust there's a you know rust
 

00:02:34.360 --> 00:02:36.150 align:start position:0%
mentioned rust there's a you know rust
representation<00:02:35.120><c> also</c><00:02:35.360><c> distributed</c><00:02:35.840><c> systems</c>

00:02:36.150 --> 00:02:36.160 align:start position:0%
representation also distributed systems
 

00:02:36.160 --> 00:02:38.030 align:start position:0%
representation also distributed systems
is<00:02:36.319><c> something</c><00:02:36.560><c> I'm</c><00:02:36.720><c> involved</c><00:02:37.040><c> in</c><00:02:37.280><c> also</c><00:02:37.519><c> AI</c>

00:02:38.030 --> 00:02:38.040 align:start position:0%
is something I'm involved in also AI
 

00:02:38.040 --> 00:02:39.869 align:start position:0%
is something I'm involved in also AI
obviously<00:02:38.400><c> so</c><00:02:38.599><c> it's</c><00:02:38.800><c> like</c><00:02:39.239><c> the</c><00:02:39.400><c> combination</c>

00:02:39.869 --> 00:02:39.879 align:start position:0%
obviously so it's like the combination
 

00:02:39.879 --> 00:02:42.110 align:start position:0%
obviously so it's like the combination
of<00:02:40.040><c> these</c><00:02:40.239><c> things</c><00:02:40.519><c> I</c><00:02:40.599><c> think</c><00:02:40.800><c> is</c><00:02:41.200><c> uh</c><00:02:41.959><c> it's</c>

00:02:42.110 --> 00:02:42.120 align:start position:0%
of these things I think is uh it's
 

00:02:42.120 --> 00:02:44.190 align:start position:0%
of these things I think is uh it's
really<00:02:42.360><c> nice</c><00:02:42.560><c> to</c><00:02:42.800><c> you</c><00:02:42.879><c> know</c><00:02:43.000><c> be</c><00:02:43.120><c> able</c><00:02:43.239><c> to</c><00:02:43.360><c> see</c>

00:02:44.190 --> 00:02:44.200 align:start position:0%
really nice to you know be able to see
 

00:02:44.200 --> 00:02:45.630 align:start position:0%
really nice to you know be able to see
talks<00:02:44.480><c> on</c><00:02:44.680><c> each</c><00:02:44.840><c> and</c><00:02:45.040><c> somewhere</c><00:02:45.480><c> they</c>

00:02:45.630 --> 00:02:45.640 align:start position:0%
talks on each and somewhere they
 

00:02:45.640 --> 00:02:56.470 align:start position:0%
talks on each and somewhere they
combined

00:02:56.470 --> 00:02:56.480 align:start position:0%
 
 

00:02:56.480 --> 00:02:59.480 align:start position:0%
 
so