New AI may threaten 'safe' office jobs, say researchers | BBC News
BBC News- 840 views
- 30 Nov 2023
Humans already work alongside robots in industries such as manufacturing. Now, for knowledge workers, the threat of AI (artificial ...
Catching stories in the Context, it is time for our new weekly look at all things artificial intelligence. Welcome to AI Decoded, our weekly appointment with some of the most eye-catching stories in the world of artificial intelligence. We're going to take a look this week at how this technology is going to affect work. Here's a story from the FT who report that it's the high flying professionals in the city of London who will be most at risk from the advances in generative artificial intelligence. In line with that, the Times asked whether workers who are replaced by AI should receive compensation. The paper says the rise of tools like ChatGPT will inevitably lead to certain jobs being replaced. And in similar vein, this from Stat, medical news, a warning that even in the hiring process, these algorithms could and perhaps already are discriminating without any nuance against certain types of people, notably women, who've been pregnant, who will have missed work for understandable reasons. In other news, Reuters looks at how Google DeepMind is using artificial intelligence to develop the high tech materials of the future for energy conversion, data processing, all manner of things. The US researchers have employed it to look at two million new materials, which could lead us to a new generation of technologies.
The Sun says ancient text scribbled on 5,000 year old tablets are being decoded by AI technology that makes sense of these texts. It's a system similar to Google Translate, which could render Indiana Jones redundant. And an equally mysterious text has appeared in sports illustrated magazine who are accused of using artificial intelligence to create authors and writers who don't seem to exist. We will tell you all about that with me tonight in the studio, my own super sleuth, Stephanie Hair, author and commentator on all things artificial intelligence. Welcome to The Prone. Good to have you in person here with us. Let's start with this story in the financial times, because I think we've all been trying to quantify what difference AI would make to the UK job market. Now they've looked at this in-depth, that different sectors, different occupations, what does it tell us?
Well, it tells us that this is something that's going to be affecting professional workers, white-collar workers, far more than other people. That's going to be management consultants, financial managers, accountants, psychologists, interestingly.
Economists, and.
Then the very hard-done-dry lawyers.
The very hard-done-to-law. No one feels sorry for lawyers today, but maybe they should, because actually, what you're looking at here is disruption towards people with high-level qualifications.
That's right. Now, the question is, is it really going to be replacing them? Or, to take our example of the lawyers, do you need to look at 500 contracts to learn how to draw up a contract? Or can you use these tools to help you get up to speed faster when you're going through your training journey and then be doing much higher value tasks? That's the real debate within the legal profession, and I think we're going to see that, frankly, across all of these. It's really easy to say, Oh, all these jobs are going to be gone. I don't know if they are. I think they're going to change and transform. Hopefully, people can get rid of the stuff that they find tiresome and tedious and get into the stuff that's actually innovated. It's hard to innovate. 80% of your job is just grinding.
Right. We have got a problem in your mind. If you can talk up, we'll try and fix that as we go. I'm looking at the data on this, and I see a third of UK jobs subject to some automation over the next 20 years. That's an awful lot of jobs.
But again, is it the jobs or is it the tasks within the jobs? Now, some jobs are going to go, and we've always seen that throughout human history and then other jobs come up. What we are really trying to talk about here is for people who lose their jobs, will they be getting some training, re-education, etc, that they can go into the new jobs? Or are we going to see people getting enhanced in their jobs, or existing jobs, so they stay a lawyer, they stay an accountant.
They become a.
Super empowered lawyer or accountant.
This takes us to this story in times too today, and it's a big question about whether you compensate people for jobs if they lose jobs to artificial intelligence. They've asked two people in this field. One has an opinion that, yes, they should be compensated, one doesn't. Let's focus on whether there should be fair compensation. This system, these systems, are going to create extraordinary wealth. You could see a scenario where there could be a tax on AI technology that could pay for those of us who are not in work.
Yeah, sorry, from Soak the Rich to soak the AI. Yeah.
Do you think there could be a scenario, though, where companies that use this who are looking for cost savings could be forced to pay such attacks?
Potentially. I think we're going to have to really see how this plays out. Right now, everybody's making a lot of very big statements, but we don't actually know. Let's imagine a case where we see unemployment spike and we're going to need to retrain people. You might say, Okay, we tax all industries and people who have a lot of money, more than others, perhaps this is what we're going to have to do. Maybe the question is, what would that money be used for? If it's just going to pay off your national debt or go into your defense industry to start more wars, for instance, that might be quite controversial. If the money is going back and being used to re-educate and retrain your workforce, I don't know if people would have a problem with that necessarily.
When you talk to me about this, though, I find it all the more surprising that we don't have an industrial strategy because how can you possibly know where we're going? Unless every sector, every industry is sitting down saying, This is how we are going to change. There are transformations in every economy all the way back to the Industrial Revolution. This is going to change so fast. How could you do it without a plan?
Well, it's difficult to have an industrial strategy with AI if you haven't had an industrial strategy before AI. I wouldn't wish to name any countries, but some of them need to step up. Some countries do have an industrial strategy with AI. The United States is one, China is another. The ones who are going big on AI are thinking about that. The UK doesn't. I wouldn't wish to name the names being discreet.
There are other people who do not think there should be compensation because they would say AI is not about replacement; it's about enhancement, as you say. Get with the program is the blunt message from some people who are working with it.
I can see that point because we didn't compensate people when calculators came on or Microsoft Excel and PowerPoint, the vein of every office worker's existence. Nobody's done any compensation there. That said, I think the Nordic economies offer a really interesting model. They do a higher tax policy that then is used to retrain people who are out of work. This exists. We don't have to reinvent the wheel here. We may take inspiration, if it's necessary, from countries that are already doing this, but in other ways.
Maybe we've not got to the point where employers are replacing people, but we're certainly at a point where employers are turning to AI in their employment decisions. There's this story in stat news today about algorithms that are programmed explicitly to look for the traits people have. Now, there are some people, I'm thinking of pregnant women, who may be absent from work from time to time for understandable reasons, but there's no nuance in AI that would actually pick this up.
No. Even if you were to try and code your algorithm so that it took into account this is a woman that's being examined to apply for a job, for instance, and you wanted to make it gender neutral and remove and strip out the fact that she had taken some time off from maternity leave, it would still be able to work out that it's a woman through other ways. We've seen this in all sorts of other areas, and that's called proxy discrimination. There are many other ways to figure out somebody's female other than just simply a photo or their name or anything that would be an obvious characteristic to you and I. Ai can do this in a way that's really subtle. The way that they build these models is going to have to be very careful and open to investigation. They're going to have to be transparent and auditable. Open source.
Exactly. Yeah, alarm bells ringing for employers on discrimination law. Let's talk about these thousands of new materials that Google Deep Mind is looking at. I remember going up to Manchester not many years ago to talk about graphene, a new atom thick, super strong material that took years to develop. These technologies now are coming much, much quicker. You think about lithium batteries, for instance.
They used to take, on average, just to give the audience an idea, 10-20 years to develop these types of materials. It's super time-consuming and costly. If Google Deep Mind has truly created something where people are going to be able to cut that time and get us new materials to market faster so we get new computer chips, new semiconductors, new materials to help for sustainability, green tech, climate tech. This could be a real game-changer for us, and they've made it publicly.
Available to our researchers. How is it doing that? Is it putting different pools together, the research teams that wouldn't necessarily work together?
Well, I'm delighted you ask. The tool is called Genome, and this stands for what? It stands for Graph Networks for material exploration, excuse me, and it's a graph neural network model. We're using deep tech, deep AI, deep neural networks. They are able to make connections between atoms and between the components that will create these materials in ways that humans would struggle to do. They can do it faster. They can do it at scale in a new novel ways that might not even occur to us.
Extraordinary. It's really extraordinary. Let's talk about this story in The Sun, voices from the Past. So it can decipher. I don't know if you're going to tell me that within a year, we'll have discovered the Lost Ark, right? We're going to go through all these Mesopoietamian and ancient texts, and we're going to find the secrets because for the first time, we can translate them at speed.
It's.
Incredible.
Well, if you think about how long it took Champollion to decipher the Rosetta code, and it so exhausted him that he actually had a nervous breakdown and collapsed and needed five days to recover. This would allow us to go through all of these cuniforms, which are 5,000 years old, 12 different languages. There's a million of them. We'll probably have them all deciphered within a year. We're going to have a glimpse into human history at the dawn of time that we've.
Never had before. I'm guessing that ancient scribes and chisellars also made typos. Would it be able to see through that?
It probably would actually start to notice typos. It might even help you to figure out who were the different scribes, because different humans have different ways of making letters, just like your handwriting is different to mine. They might even be making the codes a little bit different. So there's probably-.
So it literally would be like Google translate. You'd have the Mesopotamian text and then the translation to the side. That is extraordinary. I say that so much on this program because this is moving so fast. We have just a minute to discuss this last story, sports illustrated. I'm talking of mysterious texts. These are authors that didn't exist. They said that what they cut cost to create writers who have personalities... Yeah, they came up- And each week, they produce material?
They came up by... I can't believe we're even discussing this, but it is true. They came up with fake headshots, fake names, and fake biographies down to their fake hobbies and interests. And then they wrote articles, but then they got busted. So far, I think the sports-illustrated swimsuit issue is still an actual woman on the cover, but we may see that also become a woman. Right.
There's no hope for me, really, is there? No. The journalists are in that seven %, aren't they?
I disagree. I actually think we're going to have a backlash against all of this, and people are going to want real human beings with real senses of humor and beautiful, glorious human error.
Well, there's plenty of that on this program. Apologies to everyone for the sound quality and the problem with the mics. When we get AI, it'll be much clearer, won't it? They won't be those technological problems. Stephanie, thank you very much indeed. That's it from AI Decoded this week. We're here every Thursday to explain all things AI, so I hope you'll dip in same time next week. Thank you for watching.