Microsoft CEO unravels ChatGPT, ethical AI, and going bust

Insider

Microsoft CEO Satya Nadella sat down with Axel Springer's CEO Mathias Döpfner to discuss AI and the partnership of Microsoft and OpenAI.

Transcript

00:00 Jeff Bezos once famously said, "One day, Amazon is going to be bankrupt."

00:04 Were you worried that Microsoft could also go bankrupt?

00:07 You know, look, for sure, there is no guarantees.

00:11 I think at the end of the day, a business is only good and should exist in the world

00:17 if it's doing useful things for the world.

00:19 Yes, it's answerable to its shareholders.

00:22 It has to be profitable.

00:23 It has to create innovation.

00:25 But at the end of the day, it is performing a societal need.

00:29 And so to me, that's kind of what will keep us, whether going or irrelevant,

00:34 is all dependent on whether we wake up tomorrow and produce things that the world needs more of.

00:39 How long are you going to remain CEO of the company?

00:43 And what are your goals beyond business, beyond Microsoft?

00:47 That's like a disclosure event, isn't it?

00:49 It depends on your answer.

00:50 I never dreamt that one day, you know, people ask me,

00:54 "Hey, did you someday think that you'll be CEO of Microsoft?"

00:57 I said, "God, I'm glad I got a job at Microsoft and I held it this long."

01:02 Is China going to win the AI arms race?

01:07 Look, I mean, the interesting thing is, you know,

01:12 if you had asked two years ago, three years ago, where the world was,

01:17 I think everyone was talking about how China may be ahead or what have you.

01:20 And now I think everyone thinks the US is ahead.

01:23 I feel inherently the approach that we are taking,

01:28 whether it's in the United States or in Europe, the AI Act in Europe,

01:31 these are the ways to go about it.

01:33 You do it in the open, you do it with real responsibility to the broader society.

01:39 This is not about one company or one country.

01:41 These are foundational technologies that are going to have real impact.

01:44 And so therefore, I think I'll give us a much better chance of being able to

01:51 build technology that can be deployed with trust more broadly around the world.

01:56 In that context, one of your biggest recent coups may play a role.

01:59 That was the early investment into open AI.

02:02 And with JetGPT, BingChat, you really have challenged your biggest competitor,

02:10 Google, if I may say so.

02:11 So could you share with us a little bit the story?

02:15 When did you meet Sam Altman first time and how did this whole deal develop?

02:20 Yeah, I mean, I've known Sam a long time.

02:22 In fact, I met him first when he was just dropping out from school to start his first startup back

02:28 in 2007 or 2008, I think is the first time I met him.

02:32 And then 2018 is when Sam sort of talked about their need for compute

02:39 to build at that time, GPT-25 and 3.

02:43 And so we were, you know, said, let's give it a go.

02:45 And that's where the investment came.

02:48 And then my confidence really went up when we went from GPT-25 to 3.

02:55 But more importantly, when we built the first product, GitHub Copilot.

02:58 And it has the potential to change user habits big time.

03:01 It may even have the potential to disrupt search completely.

03:05 You once said, I want to see Google dance.

03:07 Is Google dancing now?

03:08 Look, I mean, I call it the over-exuberance of somebody who has 3% share versus somebody

03:15 who has 97% share.

03:18 But that's the humbleness of a 2.5 billion market cap company.

03:22 But that said, we are very excited about the potential here.

03:27 If you say the 70 years of computing was always about searching for what's the most natural

03:33 user interface.

03:34 Finally, we have something that you, the computers that can understand you.

03:38 It can basically, it's multi-modal.

03:40 That is, it can be text or speech or image or video in and out.

03:45 It can be multi-turn.

03:46 It can be as many turns of the conversation.

03:48 And it can be multi-domain.

03:50 I can speak about German history or I can speak about solar energy and both of them,

03:55 you know, it'll understand.

03:56 So the ability to have that type of a user interface with a new type of reasoning engine,

04:01 right?

04:01 That's the other thing, which is you have essentially a neural reasoning engine to put

04:06 on top of all the data you've collected.

04:08 And so these two things are going to change every software category.

04:11 Elon Musk said AI is more dangerous than nukes.

04:16 Do you share that?

04:17 Look, if we lose, I think the thing that he's referencing, which is I think a good thing

04:25 for people to worry about is if there is a very powerful new technology that we lose

04:30 control of, then that's a problem.

04:32 I think the thought experiment is what if you had something which is an AI that is self-improving

04:38 with no human intervention, what will it decide to do is the existential question.

04:43 And I look at it and say, look, there's lots of steps along the way.

04:46 I mean, you could have said cars could be just driving around and or planes could be

04:51 flying around without FAA or what have you.

04:53 And we've figured out as human beings how to use very powerful technology with lots

04:58 of rules, lots of regulations and loft a lot of safety standards.

05:01 Be very mindful as people who are early in producing technology, not to sort of wait

05:06 for the unintended consequences to be sprung on the world, but to think about it at design

05:12 time.

05:13 But that said, let us also not abdicate our ability to control this.

05:16 Yeah, one of the smartest brains of AI developments, Mustafa Suleiman, is convinced that AI is

05:23 particularly strong with emotion.

05:26 We tend to think we always want to set limits.

05:29 We want to say AI cannot be creative.

05:31 AI cannot be cannot develop a sense of humor.

05:37 AI has no emotion.

05:38 These things, AI is particularly strong in emotion.

05:42 How do you see that?

05:42 And I mean, look, I mean, it trains and it learns and it's consistent, which is you can

05:48 say that about AI.

05:50 You can't say that about us humans.

05:51 So you would say AI is better in faking emotions than human beings.

05:55 You make you're making a good point.

05:56 I mean, that is the issue.

05:57 The point that is the ethical issue is when we're dealing with it, it's about, yes, it's

06:04 being empathetic.

06:05 But is it manipulating or is it being empathetic to be helpful?

06:09 AI is very efficient, very smart, knows everything, has emotions, has a sense of humor.

06:14 What is left for human beings or which form of intelligence should we focus?

06:20 Look, I think I think some of the core things that we value today, like critical thinking,

06:26 will be as relevant, if not more like everybody.

06:29 Every time we've thought, you know, had a new tool that had some superpower, we've always

06:34 come to question what do we do?

06:36 In fact, you know, the famous quote, which actually is from, you know, Steve Jobs had

06:43 said that when he talked about PCs and the computers first as bicycles of the mind, I

06:49 think we now have a steam engine for the mind that we have got an upgrade.

06:53 And the question is, what can we do with the steam engine of the mind?

06:56 And so one of the things I recently was reading is a professor sort of gave the output to

07:02 students and said, come back and submit your prompts.

07:05 And I thought that that's a beautiful inversion, right?

07:07 I mean, it just shows that human creativity and human learning is still going to be very

07:13 much in the case because this is still a tool.

07:15 OK, two more prompts.

07:16 Will human beings serve the machines or will the machines serve human beings?

07:22 We will have to ensure that the machines serve the human beings and human interest.

07:27 On the other hand, we thought we cannot let you go without anything.

07:31 So it needs a little trophy.

07:32 Thank you so much.

07:39 There you go.

07:41 Is there really a degree of humor, emotion, creativity?

07:46 Let's check it out.

07:48 We asked Bing Chat a prompt.

07:51 [BING CHAT PROMPT]

07:57 Ladies and gentlemen, humans and their favorite devices.

08:00 Greetings.

08:01 I am your friendly neighborhood AI here to pay tribute to the wizard behind the digital

08:06 curtain.

08:07 None other than Satya Nadella, the coding conjurer at the helm of Microsoft.

08:13 Now, before you start wondering...

Category

Transcript

Recommended