Jeremy Howard and Joshua Browder discuss AI & Jobs with Piers Morgan
00:00:00.000 | Can our jobs all be at risk, or could AI be used to drive productivity across the economy?
00:00:05.120 | Joining me now is the CEO and founder of Do Not Pay, the world's first robot lawyer, Joshua
00:00:09.560 | Browder, and Jeremy Howard, founding researcher of Fast.AI, and Matthew Said and Amy Lewin
00:00:15.220 | are still with me.
00:00:16.240 | OK, Joshua Browder, here's my question about jobs specifically.
00:00:22.560 | Let's assume for a moment that AI and robots take a whole lot of jobs.
00:00:27.360 | Say Goldman Sachs are right, 300 million jobs.
00:00:30.200 | All those people are suddenly unemployed, who used to have those jobs.
00:00:34.400 | Where are the people going to be with enough income, if they're unemployed, to buy the
00:00:38.920 | products being made by all the robots?
00:00:42.240 | Well, the government can give them money.
00:00:45.960 | I think AI will replace a huge number of jobs.
00:00:49.100 | Lawyers will be the first to be replaced by AI because they're charging hundreds of dollars
00:00:52.760 | an hour for copying and pasting documents.
00:00:54.920 | But AI will also create a lot of jobs.
00:00:57.640 | A lot of jobs today didn't exist 20 years ago.
00:01:00.800 | So at my company, Do Not Pay, we're now hiring jobs called prompt engineers.
00:01:05.440 | And that is actually telling the AI what to do.
00:01:08.160 | And that job didn't even exist a year ago.
00:01:10.600 | So I think there will be new and exciting jobs for people.
00:01:13.160 | But at the same time, those that charge a lot of money for doing very little, like some
00:01:17.680 | lawyers, not all lawyers, have to worry about being replaced.
00:01:21.380 | You seem very, very anti-lawyers.
00:01:24.960 | I spent too much time trying to fight them.
00:01:27.400 | Probably too much money.
00:01:29.240 | One thing your chatbot lawyer has successfully done is overturn almost 200,000 parking tickets.
00:01:36.140 | I could see you becoming extremely popular just with that service alone.
00:01:40.200 | Yeah, for very simple tasks, no one has time to wait on hold for five hours to save $50,
00:01:47.400 | like getting a refund for a company or getting out of a parking ticket.
00:01:50.920 | And so that's the perfect job for AI, saving time and money for people.
00:01:55.160 | And I think those people, the lawyers you see on billboards that charge a lot of money
00:01:59.600 | to do that should be very worried.
00:02:01.320 | OK.
00:02:02.320 | So Jeremy Howard, I mean, clearly a massive threat to human employment.
00:02:07.560 | We're seeing it already.
00:02:08.960 | And it's going to move probably faster and faster.
00:02:11.400 | But what do we do about this?
00:02:12.400 | You can't just have vast swathes of the planet who were employed suddenly not having employment.
00:02:18.080 | What do we do?
00:02:19.080 | Yeah, I think we've got to be careful.
00:02:22.320 | You know, it's not as easy as what Joshua described.
00:02:25.480 | There aren't going to be new jobs to fill in all the old ones.
00:02:29.080 | And I can explain why.
00:02:30.320 | It's very simple.
00:02:31.960 | Think of it this way.
00:02:34.320 | We have two things.
00:02:35.680 | As humans, we have a body and a brain.
00:02:37.680 | Our body can move things.
00:02:39.320 | Our brain can think about things.
00:02:41.920 | Back in the Industrial Revolution, the engine was developed.
00:02:45.400 | And before that, in the UK, 80% of people worked on farms.
00:02:49.280 | And the engine came along and allowed us to replace humans using their bodies to move
00:02:53.920 | things with machines.
00:02:55.860 | And today, only 1.5% of people work on farms in the UK.
00:03:00.840 | That's fine.
00:03:01.840 | Lots of new jobs came along because we still had something else to give, our brains.
00:03:05.320 | And so now most of us do jobs which involve, at least to some extent, thinking about things.
00:03:10.400 | Now if AI can come along and think about things better than we can, where are these replacement
00:03:16.760 | jobs going to come from?
00:03:17.880 | We've got things that can move stuff.
00:03:19.720 | We've got things that can think about stuff.
00:03:22.120 | So where's the role for humans?
00:03:23.600 | I do think there'll be some jobs still.
00:03:25.680 | For example, talk show host.
00:03:27.440 | I think there are some things where we need a human.
00:03:29.440 | You know, I don't want to tune in to Piers Morgan bot, right?
00:03:32.400 | Well, you just chumble the okay, but on that, it's very interesting.
00:03:35.680 | You don't think you do.
00:03:37.360 | But if I was to have a robot, AI robot, be programmed, look like me, and had access to
00:03:44.440 | everything, every question I'd ever asked, every mannerism, every style, whatever, I reckon
00:03:49.600 | quite quickly they could develop something which could do a very passable version of
00:03:54.000 | me.
00:03:55.000 | Right.
00:03:56.000 | But I still wouldn't tune in.
00:03:57.120 | Like, think of another one.
00:03:58.120 | Are you going to tune in to the tennis playing bots?
00:04:01.400 | Like the fact that Roger Federer is an amazing human is why we like watching him play tennis.
00:04:07.120 | So I think like there'll still be a role of like humans doing human things and other other
00:04:11.720 | people saying, well, look at that person.
00:04:13.720 | That's amazing.
00:04:14.760 | So I think there's going to be a totally different kind of role for people that they won't be
00:04:21.400 | jobs in the classical sense, but it could be great.
00:04:25.040 | If we find a way to transition to this, it's not a threat.
00:04:28.720 | It means you don't have to go to work and do eight hours of whatever you're told tomorrow.
00:04:34.380 | You can do whatever you most want to, that could be great, but it could possibly be a
00:04:38.680 | huge threat.
00:04:39.680 | Yeah.
00:04:40.680 | Let me bring Matthew in here.
00:04:41.680 | A friend of mine, a friend of my son, actually, my oldest boy, his mother wanted him to send
00:04:46.240 | a thank you note for a party she'd arranged for his 30th birthday and he kept delaying
00:04:50.520 | this.
00:04:51.520 | And eventually he asked AI to do him a thank you note to his mum, giving it a few details.
00:04:56.560 | And it did a note that was so perfect and so emotional and heart-rending.
00:05:01.640 | His mother was reduced to tears when she read it and said she'd never been so moved by him.
00:05:06.260 | Now, is that good or is that awful?
00:05:09.360 | It brought great joy to his mother, but she has no idea it was a robot.
00:05:13.680 | As it happens, on Tuesday, my wife sent me an email.
00:05:17.120 | She had gone to chat GBT and said, write a Sunday Times column in the style of Matthew's
00:05:22.320 | side.
00:05:23.320 | And I was like, this is going to be terrible.
00:05:25.040 | I'm reading through it.
00:05:26.040 | Thank you pretty good.
00:05:27.040 | I'm thinking my goodness journalists are going to go.
00:05:30.240 | On the question, by the way, of we want to connect with the human.
00:05:33.600 | How do we know that we're currently talking to Piers Morgan, the flesh and blood human
00:05:37.560 | reality in the robot?
00:05:39.000 | What would stop you substituting the hologram?
00:05:41.200 | I interviewed a robot, honestly, Amy, I interviewed a robot at Good Morning Britain.
00:05:46.480 | It was a female robot and it was chilling.
00:05:49.640 | She looked like a woman.
00:05:50.960 | She spoke like a woman.
00:05:52.480 | Now that was a few years ago.
00:05:54.160 | God knows where they're getting to with this now, where they can just be very convincing
00:05:58.560 | humans.
00:05:59.560 | But with amazingly high power brains.
00:06:01.840 | They can be.
00:06:02.840 | They're not perfect, though.
00:06:03.840 | Yeah, I've got a friend who's a school teacher and she said that when she gets an essay done
00:06:08.280 | by the robot, it's so much better than any 13 year old boy could actually do that.
00:06:12.360 | She can really tell which ones, which still at the moment, but they'll get there.
00:06:16.360 | They will.
00:06:17.360 | They will.
00:06:18.360 | I mean, finally, Joshua, I've asked a few guesses, but what are you most excited by,
00:06:23.080 | by the potential of AI?
00:06:26.760 | I think AI, as was discussed in the previous guest, is being used for evil with debt collectors
00:06:32.320 | and all of this stuff.
00:06:33.320 | But it can also be used for good.
00:06:35.320 | And my goal is to give power to the people.
00:06:37.820 | And if it makes ordinary people more powerful than the richest in society, then that's great.
00:06:43.640 | And so I think it will level the playing field by allowing people to weaponize AI to help
00:06:49.160 | them in their everyday life.
00:06:51.040 | All right.
00:06:52.040 | Jeremy, same question for you.
00:06:53.040 | Quick answer, please.
00:06:54.040 | Yeah, imagine the huge opportunities in education.
00:06:57.480 | I've got a daughter and we're already she's doing stuff with chat GPT and stuff and it's
00:07:01.920 | fantastic.
00:07:02.920 | She can learn about anything she wants to.
00:07:05.200 | It's an engaging thing.
00:07:06.200 | You know, it's not replacing teachers at the moment, but I think AI could be used to really
00:07:10.840 | democratize education.
00:07:12.400 | That's something I'm very excited about.
00:07:13.560 | I could actually see robots taking classes with kids.
00:07:16.280 | I mean, if they're if they're good enough and they give them a bit of personality, why
00:07:19.680 | not?
00:07:20.680 | I mean, most teachers do a version of the same kind of lessons.
00:07:23.160 | I mean, you get the a few, you know, if you who break out and do very different things
00:07:27.080 | each time, but all of them do the same stuff, it's going to be like a personal tutor for
00:07:31.000 | every kid.
00:07:32.000 | Yeah.
00:07:33.000 | It's going to be fascinating.
00:07:34.000 | And thank you both very much indeed.
00:07:35.680 | I appreciate it.