WEBVTT

00:00:08.160 --> 00:00:11.357
Thanks for having me.
>> Great to have you, Sam.

00:00:11.440 --> 00:00:14.317
>> It's interesting cuz
people have been talking about

00:00:14.400 --> 00:00:19.517
the potential downsides or serious
risks of AI for some time.

00:00:19.600 --> 00:00:22.997
And I think a lot of people have been
saying, "Why is this the moment where it

00:00:23.080 --> 00:00:27.077
sort of escaped the bubble
and is now all of a sudden an

00:00:27.160 --> 00:00:30.597
international topic of such uh

00:00:30.680 --> 00:00:34.477
intensity?" And I think a few
things have happened. One,

00:00:34.560 --> 00:00:40.277
the models have gotten so good. And people
for a long time looked at the models,

00:00:40.360 --> 00:00:43.877
maybe they looked at the rate of progress,
but I don't think people really felt that

00:00:43.960 --> 00:00:47.677
the models were going
to come this far this fast.

00:00:47.760 --> 00:00:49.997
We will get it right, by the way. I am

00:00:50.080 --> 00:00:55.597
very confident in our I think
the biggest thing that has changed is

00:00:55.680 --> 00:00:59.237
this thing that our industry has been
talking about for a while, which is no,

00:00:59.320 --> 00:01:02.917
no, this technology, this approach
is going to go very far. These models are

00:01:03.000 --> 00:01:08.317
in some ways going to become more capable,
smarter than people. Um that finally

00:01:08.400 --> 00:01:11.877
happened. So, I think that was a big
change to why now. When people say, you

00:01:11.960 --> 00:01:15.317
know, my probability of doom
is X or Y or whatever,

00:01:15.400 --> 00:01:18.917
it hits in a different way. Technology
to do. Somebody else wants to build like

00:01:19.000 --> 00:01:23.237
a hacking bot, okay. Um the world can
decide what they think about that, but

00:01:23.320 --> 00:01:27.560
uh I think you also see um

00:01:28.000 --> 00:01:32.037
a real fear that some of these companies
or some companies developing AI could

00:01:32.120 --> 00:01:34.837
get too much power and be able to sort

00:01:34.920 --> 00:01:37.477
of exert undue influence uh on the

00:01:37.560 --> 00:01:39.997
economy, push a worldview out on people.

00:01:40.080 --> 00:01:43.197
Uh and I think the world
is right to be afraid of this.

00:01:43.280 --> 00:01:46.837
So, we have these two big challenges.
We have the potential of a loss of control

00:01:46.920 --> 00:01:50.997
accident or some other serious thing
that could go wrong. Um we have the

00:01:51.080 --> 00:01:55.117
potential of way too much power
concentration and people fear that those

00:01:55.200 --> 00:01:58.837
developing AI could
exert their worldview. Um

00:01:58.920 --> 00:02:02.477
and there is a sort of narrow
path we have to navigate

00:02:02.560 --> 00:02:06.917
with pragmatism, with you know,

00:02:07.000 --> 00:02:10.877
steadiness, with the ability for the
the world to like trust that we will make

00:02:10.960 --> 00:02:15.117
dependable decisions,
consistent decisions uh through this.

00:02:15.200 --> 00:02:17.157
And I think people are saying, well, you

00:02:17.240 --> 00:02:18.557
know, this now seems pretty important.

00:02:18.640 --> 00:02:22.597
Like let's let's get it right.

00:02:22.680 --> 00:02:25.117
I'm not going to get the days of the
week exactly right here, but something

00:02:25.200 --> 00:02:29.917
on like a Friday or a Saturday. These
models have gotten so good so fast that

00:02:30.000 --> 00:02:34.037
we need to treat the alignment and
monitoring and security with a new level

00:02:34.120 --> 00:02:36.797
of rigor. And we have to be willing to

00:02:36.880 --> 00:02:38.957
pace our development of capabilities

00:02:39.040 --> 00:02:44.597
such that capabilities always uh
such that alignment, safety, and

00:02:44.680 --> 00:02:48.037
monitoring are always
ahead of capabilities. And

00:02:48.120 --> 00:02:51.357
the Hugging Face incident triggered
a real reset, not just in our own company,

00:02:51.440 --> 00:02:53.077
but I think the whole industry.

00:02:53.160 --> 00:02:54.917
>> You know, we're
in the cyber area today,

00:02:55.000 --> 00:02:56.797
but maybe we move into another area. Is

00:02:56.880 --> 00:03:01.957
this a fair way to characterization,
or am I or is this unfair to characterize

00:03:02.040 --> 00:03:04.517
it like this? >> What I really
care about is that we have

00:03:04.600 --> 00:03:08.837
got a great culture of accident
reporting and learning. I think if we

00:03:08.920 --> 00:03:12.717
look at what happens with the FAA
and the NTSB and how safe airplanes have

00:03:12.800 --> 00:03:17.197
gotten, um as I understand it from talking
to people who have really studied,

00:03:17.280 --> 00:03:20.917
you know, new kinds of technology on the
order of magnitude of airplanes, one of

00:03:21.000 --> 00:03:24.677
the things that has gone so well there is

00:03:24.760 --> 00:03:26.517
is the culture of accident reporting.

00:03:26.600 --> 00:03:28.197
You know, accidents are going to happen.

00:03:28.280 --> 00:03:29.957
We're going to learn as much as we can

00:03:30.040 --> 00:03:35.477
and react from each one. Um and if,
you know, there's other kinds

00:03:35.560 --> 00:03:39.917
of technology where I don't think it's
gone I don't think society's had the same

00:03:40.000 --> 00:03:44.077
response, and I don't think
we have improved as much.

00:03:44.160 --> 00:03:47.237
to pull this thing together and present
you with the information to make a

00:03:47.320 --> 00:03:48.877
decision or to blame.

00:03:48.960 --> 00:03:52.077
I think this is one of the most profound

00:03:52.160 --> 00:03:56.357
changes to how we use
technology that I've ever seen.

00:03:56.440 --> 00:03:59.677
I mean, sci-fi movies, as you mentioned,
have been hinting at this or playing

00:03:59.760 --> 00:04:02.957
around with this for a long time,
but this is one of those moments where it

00:04:03.040 --> 00:04:06.237
feels like you're in the movie.
Like the sci-fi thing came true.

00:04:06.320 --> 00:04:10.517
You're talking to your personal super
assistant and you are working 10 times

00:04:10.600 --> 00:04:14.357
faster or better or more creative
than you could have before.

00:04:14.440 --> 00:04:17.997
>> Sam, you're invited back
anytime you want to Dreamforce.

00:04:18.080 --> 00:04:20.877
>> Thank you very much. Thank you.

00:04:20.960 --> 00:04:24.120
>> Please thank Sam Altman.