WEBVTT

00:00:00.000 --> 00:00:04.557
Copilot or anywhere else.
It's just a diffusion of the tools.

00:00:04.640 --> 00:00:07.997
I'll just give you you know one of the
best examples at Microsoft was I was

00:00:08.080 --> 00:00:12.197
talking to an engineer in our networking
our AI one, right? I mean I was talking

00:00:12.280 --> 00:00:16.917
about man the AI one is growing like
crazy. The number of fiber operators we

00:00:17.000 --> 00:00:21.677
work with is like some you know insane
number and by the way all the devops on

00:00:21.760 --> 00:00:24.797
it right all you know if there's a fiber
cut somewhere is all manual, right?

00:00:24.880 --> 00:00:27.517
Literally it is emails and calls and so

00:00:27.600 --> 00:00:32.357
on. So she just said okay I can't deal
with this. I have to really automate

00:00:32.440 --> 00:00:36.237
this because otherwise you know all of
Microsoft will be just working on devops

00:00:36.320 --> 00:00:39.117
of fiber. And so she
just built an agent and in

00:00:39.200 --> 00:00:41.517
fact a multi-agent orchestrator, right?

00:00:41.600 --> 00:00:45.197
And by the way she just did this using
some of the low code no code and the

00:00:45.280 --> 00:00:47.877
foundry tools, right? That ability to

00:00:47.960 --> 00:00:51.557
empower our people with tools and agency

00:00:51.640 --> 00:00:55.557
to go and look for the
workflow closest to them.

00:00:55.640 --> 00:00:57.877
That I think is what needs to happen.

00:00:57.960 --> 00:01:02.637
Just like how we were able to use you
know I think I I always say Excel is the

00:01:02.720 --> 00:01:06.077
greatest programming most ubiquitous
programming tool on the world.

00:01:06.160 --> 00:01:10.597
And all of us learned how to program
in Excel essentially and that's the thing

00:01:10.680 --> 00:01:12.277
that I think is going to happen again.

00:01:12.360 --> 00:01:15.917
And we don't have much time here but I
want to quickly touch on proactive

00:01:16.000 --> 00:01:20.157
agents. So behind the scenes we actually
got a demo of like on device proactive

00:01:20.240 --> 00:01:23.597
agent working on the
Copilot plus PCs where

00:01:23.680 --> 00:01:27.797
this agent essentially opened up

00:01:27.880 --> 00:01:31.917
Outlook and summarize email without
internet. But that's just one example.

00:01:32.000 --> 00:01:34.797
Are there any examples
that you have that you're

00:01:34.880 --> 00:01:38.637
really excited for? Yeah, so I think
that if I look at the continuum, right?

00:01:38.720 --> 00:01:43.357
So I I'm really looking for in some
sense I'm a big fan of this Mark Weiser

00:01:43.440 --> 00:01:46.477
quote about [music] ubiquitous
computing, right? Where

00:01:46.560 --> 00:01:50.157
technology is powerful enough
to disappear, right? [music] So therefore

00:01:50.240 --> 00:01:52.317
to some degree the march of natural user

00:01:52.400 --> 00:01:54.237
interface and so on

00:01:54.320 --> 00:01:57.797
is about doing all the things that I

00:01:57.880 --> 00:02:02.437
intend to be done with the least amount of

00:02:02.520 --> 00:02:04.557
I would say friction, right? So that's

00:02:04.640 --> 00:02:06.917
why I think when I want like interpret

00:02:07.000 --> 00:02:10.397
my intent and plan and get it done,

00:02:10.480 --> 00:02:14.477
right? So that's where I think some
of these proactive agent stuff comes in.

00:02:14.560 --> 00:02:17.757
And I I do I want us be in control though.
Like for example, that's why I

00:02:17.840 --> 00:02:20.277
think the the core agents for example,

00:02:20.360 --> 00:02:25.117
the computer use agents have to sort
of get to a point one that got be fast and

00:02:25.200 --> 00:02:27.277
better and more reliable. But the more

00:02:27.360 --> 00:02:32.597
importantly they have to be chained as
part of some intent of mine, right? It's

00:02:32.680 --> 00:02:36.757
not like I'm going to go give [music]
task task task task. I'm going to have a

00:02:36.840 --> 00:02:41.437
high-level intent someone's going off
and doing a bunch of work and I can

00:02:41.520 --> 00:02:45.517
inspect. So that's why I think like this
get up you know coding agent the session

00:02:45.600 --> 00:02:47.837
log, it's pretty cool. Like I go back

00:02:47.920 --> 00:02:50.197
after having a sign it's like I'm

00:02:50.280 --> 00:02:52.637
watching the session log with all the

00:02:52.720 --> 00:02:55.477
draft commits, right? So that I think is

00:02:55.560 --> 00:02:59.797
the balance, right? So in an interesting
way we talk about the proactive agent

00:02:59.880 --> 00:03:03.117
and a session log and that
way then it's sort of more

00:03:03.200 --> 00:03:06.077
transparency of it and then when you

00:03:06.160 --> 00:03:08.437
invoke multiple of them, you are able to

00:03:08.520 --> 00:03:13.437
still track effectively what's happening.
One last question for you and

00:03:13.520 --> 00:03:16.517
this is something that recently went viral

00:03:16.600 --> 00:03:21.677
about two months ago. You said
you mentioned that AGI is just some

00:03:21.760 --> 00:03:23.277
nonsensical benchmark hacking [music]

00:03:23.360 --> 00:03:26.517
and the true value of AI is global

00:03:26.600 --> 00:03:31.557
economic growth. Where do you see agents
having the most value in that regard

00:03:31.640 --> 00:03:36.197
first? It's a great one. I mean to me
take health care. What is it? 19-20% of

00:03:36.280 --> 00:03:39.997
our GDP in the United States is in health

00:03:40.080 --> 00:03:44.477
[music] care. A lot of that cost the
inefficiency. If you talk to any of our

00:03:44.560 --> 00:03:47.797
providers [music] is all
in the workflow, right? They're

00:03:47.880 --> 00:03:50.917
working so hard to provide unbelievable

00:03:51.000 --> 00:03:55.477
care except sort of you know and they're
using all these digital systems but

00:03:55.560 --> 00:03:57.837
they're really not able to

00:03:57.920 --> 00:04:01.077
tame the complexity and reduce the cost.

00:04:01.160 --> 00:04:04.997
if there is going to be one place
where I would love to love to see that

00:04:05.080 --> 00:04:08.677
productivity gain would
be like that you know that

00:04:08.760 --> 00:04:12.957
multi-agent orchestrator for health care
that we just launched used by Stanford

00:04:13.040 --> 00:04:17.557
become ubiquitous. And there are stories
that where people are saying oh wow,

00:04:17.640 --> 00:04:22.037
this has been the biggest thing that you
know every provider is able to provide

00:04:22.120 --> 00:04:23.877
better health care at lower cost, right?

00:04:23.960 --> 00:04:25.637
So something like that has to happen

00:04:25.720 --> 00:04:27.757
because my only comment there was not

00:04:27.840 --> 00:04:31.157
about any sort of shadow on any great AI

00:04:31.240 --> 00:04:33.397
research. It is more

00:04:33.480 --> 00:04:35.797
I think we as a society

00:04:35.880 --> 00:04:38.997
celebrate tech companies far too much

00:04:39.080 --> 00:04:41.637
versus the impact of technology. Quite

00:04:41.720 --> 00:04:43.757
honestly, if there was a more balanced

00:04:43.840 --> 00:04:46.877
way to talk about sort of not the tech

00:04:46.960 --> 00:04:49.517
industry but the use of technology,

00:04:49.600 --> 00:04:54.437
right? That's why even the you know
Copilot being an Think about the decades

00:04:54.520 --> 00:04:56.117
all of us have spent saying >> [music]

00:04:56.200 --> 00:04:58.757
>> can we have one tech intervention that

00:04:58.840 --> 00:05:01.837
makes a damn difference in education?

00:05:01.920 --> 00:05:05.957
Right? Finally, you have real
statistical evidence thanks to that

00:05:06.040 --> 00:05:10.317
World Bank study in Nigeria. That you
know you give people Copilot there or

00:05:10.400 --> 00:05:13.637
any other agent it doesn't
matter something changes.

00:05:13.720 --> 00:05:17.197
That's the story to me, right? That's
why I joined the tech industry. I mean

00:05:17.280 --> 00:05:20.477
it was a very different time
and suddenly it became like you know the

00:05:20.560 --> 00:05:22.677
place you were celebrating ourselves and

00:05:22.760 --> 00:05:27.800
I just hate it. [music] I
just want to get to a place

00:05:29.470 --> 00:05:31.490
>> [music]