WEBVTT - IBM Chairman and CEO Arvind Krishna Talks AI Innovation 

0:00:02.520 --> 0:00:07.400
<v Speaker 1>Bloomberg Audio Studios, podcasts, radio news.

0:00:08.039 --> 0:00:11.600
<v Speaker 2>Oppenheim and she and coverage on IBM with an outperform

0:00:11.640 --> 0:00:15.600
<v Speaker 2>writing expecting double digit revenue growth in its software portfolio

0:00:15.680 --> 0:00:18.200
<v Speaker 2>that stocks up about a quarter of one percent. Let's

0:00:18.200 --> 0:00:20.480
<v Speaker 2>stick with IBM. Let's head to the World Government Summit

0:00:20.600 --> 0:00:23.119
<v Speaker 2>in Dubai, where bloombags Geamunta BASESSI. She is with a

0:00:23.239 --> 0:00:24.479
<v Speaker 2>special guest, Jamana, I have.

0:00:24.480 --> 0:00:25.000
<v Speaker 1>A CU.

0:00:27.960 --> 0:00:30.440
<v Speaker 3>Guys, thanks for that. Yes, so joining me right now

0:00:30.560 --> 0:00:33.280
<v Speaker 3>is the IBM CEO, Arvin Krishner. Great to have you

0:00:33.320 --> 0:00:35.600
<v Speaker 3>with us here the World Government Summits in the UAE

0:00:35.720 --> 0:00:39.240
<v Speaker 3>of all places. But look, I just want to start

0:00:39.280 --> 0:00:42.839
<v Speaker 3>off with a question that I think is dominating the

0:00:42.840 --> 0:00:46.560
<v Speaker 3>discussions these days, and that is visa VI Artificial intelligence.

0:00:46.680 --> 0:00:49.839
<v Speaker 3>You can't be a US tech company these days if

0:00:49.880 --> 0:00:52.920
<v Speaker 3>you're not seen to have a growing presence in AI.

0:00:53.880 --> 0:00:57.040
<v Speaker 3>How does innovation keep up with demand for a company

0:00:57.080 --> 0:00:57.520
<v Speaker 3>like yours?

0:00:57.880 --> 0:00:59.360
<v Speaker 1>Look, I mean innovation is our life.

0:00:59.400 --> 0:01:01.920
<v Speaker 4>No star there, And I do think that AI is

0:01:01.920 --> 0:01:05.360
<v Speaker 4>a transformative technology. So no company can be in tech

0:01:05.400 --> 0:01:07.720
<v Speaker 4>without having to have to play a role in AI.

0:01:08.319 --> 0:01:09.960
<v Speaker 4>If I look at the role that we want to play,

0:01:09.959 --> 0:01:12.000
<v Speaker 4>we are focused on the B to B market, So

0:01:12.040 --> 0:01:14.000
<v Speaker 4>we do not want to do B two C. So

0:01:14.040 --> 0:01:16.360
<v Speaker 4>B to B means you've got to take advantage of

0:01:16.400 --> 0:01:17.959
<v Speaker 4>an enterprise's private data.

0:01:18.680 --> 0:01:19.920
<v Speaker 1>Here's a shocking statistic.

0:01:20.240 --> 0:01:23.640
<v Speaker 4>Only one percent of enterprise data has found its way

0:01:23.640 --> 0:01:25.600
<v Speaker 4>into any form of AI model so far.

0:01:26.280 --> 0:01:28.679
<v Speaker 1>So that is the unlock that we want to do.

0:01:29.040 --> 0:01:31.960
<v Speaker 4>And as we bring that private data into models, whether

0:01:32.000 --> 0:01:35.280
<v Speaker 4>it's used to refine models or to construct unique use cases,

0:01:35.600 --> 0:01:38.040
<v Speaker 4>it unlocks a huge amount of value for the enterprise.

0:01:38.200 --> 0:01:41.800
<v Speaker 3>Yeah, how has the arrival of deep seek disrupted the

0:01:41.920 --> 0:01:44.920
<v Speaker 3>landscape for generative AI because it's a model that was

0:01:45.240 --> 0:01:49.160
<v Speaker 3>produced with a much lower cost but also equal utility

0:01:49.360 --> 0:01:51.760
<v Speaker 3>as some of the larger language models. Is this a

0:01:51.840 --> 0:01:52.720
<v Speaker 3>major game changer?

0:01:52.760 --> 0:01:53.160
<v Speaker 2>Do you think?

0:01:53.280 --> 0:01:53.680
<v Speaker 1>Actually?

0:01:53.880 --> 0:01:56.040
<v Speaker 4>You see we smile because I think it's a validation.

0:01:56.480 --> 0:01:58.600
<v Speaker 4>We've been on this point for a long time that

0:01:58.720 --> 0:02:00.760
<v Speaker 4>you do not have to spend so much money to

0:02:00.800 --> 0:02:03.280
<v Speaker 4>get these models if you're willing to make fit for

0:02:03.320 --> 0:02:05.920
<v Speaker 4>purpose models. We do believe that the cost should be

0:02:05.920 --> 0:02:08.120
<v Speaker 4>in the millions and not in the hundreds of millions.

0:02:08.639 --> 0:02:11.519
<v Speaker 4>And so how do you begin to distill models down

0:02:11.560 --> 0:02:14.200
<v Speaker 4>to smaller sizes, get them unique for a purpose and

0:02:14.280 --> 0:02:17.480
<v Speaker 4>then run them at two to three percent, so thirty

0:02:17.520 --> 0:02:21.240
<v Speaker 4>times cheaper than the big models, but as accurate and

0:02:21.280 --> 0:02:23.440
<v Speaker 4>as good for a domain specific task.

0:02:23.720 --> 0:02:25.480
<v Speaker 3>Do you think that therefore there could be a day

0:02:25.520 --> 0:02:28.160
<v Speaker 3>of reckoning for some of these big tech companies spending

0:02:28.600 --> 0:02:31.560
<v Speaker 3>billions and tens of billions, even one hundred billion dollars

0:02:31.600 --> 0:02:33.200
<v Speaker 3>on kafak spending funds these year.

0:02:33.280 --> 0:02:35.680
<v Speaker 1>That's probably beyond my ability to computer. Is it a

0:02:35.760 --> 0:02:36.320
<v Speaker 1>day erecting?

0:02:36.360 --> 0:02:38.440
<v Speaker 4>But I can tell you that we're going to find

0:02:38.480 --> 0:02:40.720
<v Speaker 4>that the usage is going to explode as the cost

0:02:40.760 --> 0:02:43.840
<v Speaker 4>comes down, So maybe there's enough quantity increase that all

0:02:43.840 --> 0:02:45.120
<v Speaker 4>of it maps out in the right way.

0:02:45.200 --> 0:02:48.040
<v Speaker 3>Yeah, You've always been a big supporter of open source models,

0:02:48.080 --> 0:02:49.800
<v Speaker 3>and I just wonder in the case of deep Seek,

0:02:49.880 --> 0:02:55.200
<v Speaker 3>whether that actually served as a reason for the disruption

0:02:55.360 --> 0:02:57.919
<v Speaker 3>to take place. Would we not have seen a deep

0:02:57.960 --> 0:03:00.359
<v Speaker 3>seek come to the market had there not been open

0:03:00.360 --> 0:03:01.160
<v Speaker 3>source models?

0:03:01.320 --> 0:03:04.360
<v Speaker 4>No, I don't believe so, because there are enough large

0:03:04.360 --> 0:03:06.720
<v Speaker 4>models that people are building, not in the hundreds, but

0:03:06.760 --> 0:03:07.800
<v Speaker 4>definitely in the tens.

0:03:08.040 --> 0:03:08.680
<v Speaker 1>And it's hard to.

0:03:08.639 --> 0:03:11.160
<v Speaker 4>Imagine that there's an ecosystem which doesn't have a large

0:03:11.200 --> 0:03:14.760
<v Speaker 4>model to start from. Look ideas tend to spread. This

0:03:14.800 --> 0:03:16.600
<v Speaker 4>has been true for two thousand years. People have written

0:03:16.600 --> 0:03:19.119
<v Speaker 4>about this. Once there's an idea, it gets penned down.

0:03:19.120 --> 0:03:22.200
<v Speaker 4>The idea spreads. Once an idea spread, smart people, everywey

0:03:22.240 --> 0:03:22.840
<v Speaker 4>can copy it.

0:03:23.040 --> 0:03:24.440
<v Speaker 1>Yeah, let me ask.

0:03:24.320 --> 0:03:26.520
<v Speaker 3>You a question about regulation. I started off the interview

0:03:26.760 --> 0:03:29.600
<v Speaker 3>asking you whether innovation can keep up with demands. Do

0:03:29.600 --> 0:03:33.200
<v Speaker 3>you think regulation is actually keeping up with the innovation

0:03:33.240 --> 0:03:34.960
<v Speaker 3>that we're seeing in artificial intelligence.

0:03:35.480 --> 0:03:37.040
<v Speaker 1>I'll actually go the other way around.

0:03:37.280 --> 0:03:42.120
<v Speaker 4>Too much regulation early stifles innovation and then doesn't allow

0:03:42.240 --> 0:03:45.200
<v Speaker 4>those companies and those nations where the regulation.

0:03:44.920 --> 0:03:48.119
<v Speaker 1>Is heavy to succeed. We can always look at the example.

0:03:48.200 --> 0:03:50.200
<v Speaker 4>I think that the EU is very good at lots

0:03:50.200 --> 0:03:54.120
<v Speaker 4>of things, but sometimes the over regulation is an inhibitor

0:03:54.320 --> 0:03:57.880
<v Speaker 4>on innovation, and I think we should balance that. I

0:03:57.880 --> 0:04:00.840
<v Speaker 4>think that that balance is incredibly critical. And I talk

0:04:00.920 --> 0:04:03.320
<v Speaker 4>about precision regulation as opposed to.

0:04:03.280 --> 0:04:04.640
<v Speaker 1>Sort of blunt force regulation.

0:04:05.120 --> 0:04:08.400
<v Speaker 4>So be precise and take a risk based approach where

0:04:08.520 --> 0:04:11.640
<v Speaker 4>only the most risky activities are regulated. So ANYI that

0:04:11.680 --> 0:04:15.040
<v Speaker 4>becomes the use cases. So yeah, be more careful around

0:04:15.200 --> 0:04:19.080
<v Speaker 4>life or death activities, but allow innovation in customer service,

0:04:19.640 --> 0:04:23.440
<v Speaker 4>in productivity for your programmers in terms of what we're

0:04:23.480 --> 0:04:26.960
<v Speaker 4>going to do around improving the customer experience, but maybe

0:04:27.040 --> 0:04:29.000
<v Speaker 4>be more careful in life or death activities.

0:04:29.160 --> 0:04:32.320
<v Speaker 3>Yeah, let me just ask you another question which is

0:04:32.400 --> 0:04:37.520
<v Speaker 3>also very important in the context of tech innovation, and

0:04:37.560 --> 0:04:40.200
<v Speaker 3>that is the rise of countum computing. I thought it

0:04:40.240 --> 0:04:42.560
<v Speaker 3>was really interesting that in your investor dage just a

0:04:42.600 --> 0:04:46.240
<v Speaker 3>week ago, you said, well, you updated the quantum computing roadmap,

0:04:46.279 --> 0:04:49.000
<v Speaker 3>and you said that you're aim at demonstrating the first

0:04:49.040 --> 0:04:52.000
<v Speaker 3>contum computer by twenty twenty eight. That's only three years away.

0:04:52.040 --> 0:04:55.240
<v Speaker 3>People are a bit skeptical that that can be achieved in.

0:04:55.160 --> 0:04:58.240
<v Speaker 4>That timeline, tolerant with era correction part of computer. At

0:04:58.240 --> 0:05:01.040
<v Speaker 4>twenty eight, we actually have third quantum computers on the

0:05:01.080 --> 0:05:04.480
<v Speaker 4>cloud today, actual quantum computers over one hundred cubits, so

0:05:04.600 --> 0:05:08.240
<v Speaker 4>not toys, really serious quantum computers. But we believe we're

0:05:08.279 --> 0:05:10.160
<v Speaker 4>going to be on the roadmap to do that by

0:05:10.240 --> 0:05:12.560
<v Speaker 4>twenty twenty eight, and that is a commitment that we

0:05:12.600 --> 0:05:15.039
<v Speaker 4>did make. But I also believe that quantum computers are

0:05:15.080 --> 0:05:17.040
<v Speaker 4>going to unlock a lot of value. We think about

0:05:17.040 --> 0:05:19.680
<v Speaker 4>half a trillion dollars worth of value for our customers

0:05:20.000 --> 0:05:24.200
<v Speaker 4>by the end of this decade. That's exciting materials, climate change,

0:05:24.240 --> 0:05:29.560
<v Speaker 4>carbon sequestration, better fertilizers, better batteries, exciting areas.

0:05:29.680 --> 0:05:32.039
<v Speaker 3>Yeah, do you feel that because the world is so

0:05:32.120 --> 0:05:33.960
<v Speaker 3>focused on one thing at a time and this time,

0:05:34.080 --> 0:05:37.559
<v Speaker 3>you know it's generative AI, we're perhaps not focusing enough

0:05:37.760 --> 0:05:41.440
<v Speaker 3>on how quantum computing is actually going to disrupt our

0:05:41.480 --> 0:05:42.080
<v Speaker 3>day to day lives.

0:05:42.320 --> 0:05:44.520
<v Speaker 4>What do we actually look with that we can do

0:05:44.560 --> 0:05:48.440
<v Speaker 4>our work and when we get there, it'll be a moment,

0:05:48.640 --> 0:05:51.080
<v Speaker 4>and that that moment is going to be important.

0:05:51.520 --> 0:05:52.960
<v Speaker 1>Our customers are pretty focused on it.

0:05:53.000 --> 0:05:55.440
<v Speaker 4>We have two hundred and eighty people different companies and

0:05:55.520 --> 0:05:59.200
<v Speaker 4>organizations who work with us on algorithms. The state of Chica,

0:05:59.560 --> 0:06:02.839
<v Speaker 4>Illinois in the United States just announce the National Quantum

0:06:02.880 --> 0:06:08.159
<v Speaker 4>Algorithmic Center, where multiple universities, startups, large companies, national labs,

0:06:08.240 --> 0:06:09.039
<v Speaker 4>all player role.

0:06:09.560 --> 0:06:13.320
<v Speaker 1>They'll be running on our quantum computer. They probably get others.

0:06:13.000 --> 0:06:16.400
<v Speaker 4>As well if they exist and do that. I think

0:06:16.440 --> 0:06:18.640
<v Speaker 4>that's the excitement that is there. But the fact that

0:06:18.640 --> 0:06:22.280
<v Speaker 4>two hundred and eighty institutions all around the world are

0:06:22.360 --> 0:06:25.559
<v Speaker 4>busy learning how to use them tells me they'll be ready.

0:06:25.680 --> 0:06:28.560
<v Speaker 3>Yeah, Well, we're here in the UAE, and I know

0:06:28.600 --> 0:06:31.279
<v Speaker 3>that you were in Saudi Arabia recently as well. What

0:06:31.560 --> 0:06:34.520
<v Speaker 3>opportunities are you seeing in this part of the world

0:06:34.640 --> 0:06:36.839
<v Speaker 3>for AI growth and growth in your business.

0:06:37.120 --> 0:06:39.400
<v Speaker 4>We are very bullish about our business in both both

0:06:39.480 --> 0:06:42.520
<v Speaker 4>these nations, both in Saudi Arabia and in the UE.

0:06:42.720 --> 0:06:46.440
<v Speaker 4>I think the appetite for digital innovation, both for government

0:06:46.520 --> 0:06:50.160
<v Speaker 4>services and for the private sector is incredible. I think

0:06:50.400 --> 0:06:52.560
<v Speaker 4>both nations have woken up that tech can be a

0:06:52.560 --> 0:06:53.839
<v Speaker 4>big part of their economy.

0:06:54.040 --> 0:06:55.880
<v Speaker 1>If you look globally four to five percent.

0:06:56.240 --> 0:06:58.480
<v Speaker 4>I think both these nations wanted to be ten percent,

0:06:58.760 --> 0:07:01.680
<v Speaker 4>and they're investing appropriate And you can look at the

0:07:01.720 --> 0:07:05.080
<v Speaker 4>services that are there in both countries that are enabled

0:07:05.080 --> 0:07:08.760
<v Speaker 4>by technology. That appetite for AI is massive. You look

0:07:08.760 --> 0:07:11.760
<v Speaker 4>at the investment that is done in the in the OE,

0:07:11.920 --> 0:07:17.000
<v Speaker 4>around nbc UI, around some of the Falcon models. You

0:07:17.040 --> 0:07:19.960
<v Speaker 4>look good too in the in Saudi Arabia and you

0:07:20.000 --> 0:07:22.560
<v Speaker 4>have the universities and you have the Alarm model.

0:07:23.040 --> 0:07:24.119
<v Speaker 1>I think that speaks for itself.

0:07:24.200 --> 0:07:26.320
<v Speaker 3>Yeah, well, I'm sure we'll be seeing a lot more

0:07:26.320 --> 0:07:28.360
<v Speaker 3>of you in the region. Then, Arvin Krishna, really good

0:07:28.360 --> 0:07:29.000
<v Speaker 3>to chat to you.