Well, welcome everyone. This is fantastic. And we are going to be talking about the business transformation of AI. But before we get into it, just a pause. I was on the stage talking about AI last year and the conversation has shifted. It is no longer the purview of nerds. Everyone wants to talk about this. And every day we see headlines that are one day positive. This is the future. The next day, this is going to destroy the human race. Before we get to the challenges and the risks, let's just get from each of your perspectives where you see the transformation here and the potential. Let's start with you, Nigel.
Anna, as you said, AI has moved pretty significantly. We had deep learning and neural nets and now we're on to generative AI. The thing that is, I think, a bit of a game changer with this is that everyday people can now interact with technology in natural language without them knowing that they're actually building models that are helping things get automated in a way that we've just never seen before. So while there's a lot of talk about the end of the world, I feel like we have to deal with things like biases and ethics and understanding how we keep companies' data and IP protected and people's data and IP protected. And the opportunity for transformation in the context of organizations is with practical things that can now get accelerated like risk and fraud in banking or trying to identify molecules in pharmaceuticals to advance clinical trials that have been held back because lots and lots of processing couldn't be done fast enough. So for me, the opportunity I feel like in the here and now is just gotten significantly more accelerated in a way I think that will really move the conversation of AI on. And I don't think we're done with the development of the technology either. I think this is the beginning of that wave. Think first websites from the internet back in 95. Fortunately, I can really barely remember that I was just so young. Nigel, we're going to move on to eventually how you advise businesses when it comes to this. But Peter, for Siemens, what is the transformation here? It's quite niche, I feel like, for industry.
Well, it has specific requirements because if you want to run the train, if you want to run on a car, you better make sure that these things are safe. So therefore, you have extraordinary requirements with regards to precision and that these things work flawlessly. But I do agree with Nigel that generally speaking, AI, which was previously for the nerds, right, more in the hardcore R&D developments, now has become truly mainstream, which is the reason why everybody's talking around it. But let me be clear, we've been using AI since more than 20 years. We are building CNNs since a long time. And it is almost based in most of the products that we're using today. So it's going to be gradually transitioning into what you really can do. But to us, this is definitely no news.
Philip, paint us a picture. Business transformation.
So, and I can't agree with you more, first of all, around that AI has come front and center. I used to joke, I think I was the most boring guy at a cocktail party talking about epistemology. Like, what do we actually know? Like, what is knowledge? And then like, let's talk about algorithms to try and quantify that. No one wanted to hear about it. Now I think everyone wants to. It's really captured the imagination because this generation is probably the most approachable generation of technology we've ever seen. You know, in businesses of all kinds, I think the number one thing that people are really using this technology for is to connect to their customers. When I think about customer service organizations, Orange is here and they're talking about their use of it in the call center. You really are able to answer a question in the language that the customer has asked it in, as opposed to having to type in and find what the answer is in the language of the company. You can answer it in the language of the customer. And so in that call center in customer service, first and foremost, 100%. The other area is really in marketing. Organizations, every organization I talk to says, if I could just connect to that audience better, or I could just connect to that demographic better, I could just connect with this type of a message better. And it's hard to repetitively be able to create that information. Carrefour is a great example that's here as well, where they're actually creating marketing content by audience that they want to try to connect to. And so I see this generation as putting a conversational interface, the interface that we were designed to have, and just having a conversation with our information. That's what I think a lot of organizations see as the opportunity for this generation.
It is tricky, though, I think, for a number of businesses to know how to adopt AI. Some businesses are, frankly, putting their head in the sand and just banning employees from using it. Big fear about proprietary information being shared. But Nigel, how do you advise companies that come to you and say, we want to get into AI, we want to do it fast.
It's a journey. For sure. But you have to start by experimentation. One of the biggest challenges with putting something like chat GPT out into the world and companies not having a way to test information in a secure sandbox is everybody throws your confidential information to chat GPT. The model is getting smarter, but it's getting smarter off your data. So the first thing we do with a lot of clients is establish, we've got open AI and Azure cloud environments that we establish proprietary sandboxes so you can actually start training the model on your data that stays within your walls. So we partner with the likes of Car4 on their digital transformation. And if you're starting with a company like that, you want to start by establishing sandblocks where people can experiment. But the other thing that we've just talked about, AI is not something that is going to drive business transformation or technological transformation in isolation. You still need the other skills like product, like experience and how you design things and engineering and being really clear about what is the use case from a strategic perspective that you want to get out. We use this acronym called SPEED, strategy, product, experience, engineering, data and AI. And if you, it's like fingers in a hand, if you go with a strong thumb or a strong index finger, it doesn't matter if it doesn't connect to the pinky, you still can't move and lift and shift stuff. And that's where a lot of companies get slowed down because they say, yeah, we have engineering, we have experience, we have AI, we've been doing it for 20 years, but they're not connected in a way that actually is allowing them to move forward. And today, as Philip just said, customers are using natural language to talk in a way that they can push you harder than they've ever had. So accepting marketing messages that are generic, they don't want that. In a quick service restaurant, getting the order wrong because you couldn't understand what they were saying, they don't like that. And companies who start to figure out these use cases in the here and now will make significant progress. So experimentation is where we advise our clients to start.
And also, I think with general sort of AI platforms, that doesn't really work for a lot of businesses. I've been really enjoying some terrible use cases of AI. For Siemens, it must be quite interesting because you have such specific use cases for AI. Is it ever hard to figure out what platform you need to use?
Really, it is. I mean, first off, you are right. It is at the very heart of what we do. So if you think about, I mentioned trains, right? We are able to run trams autonomously. I mean, we're talking about self-driving cars, but you can do it already on the railroad because think about it, it's going forward, backwards. So it's much easier in that sense to start there. We talk about energy transition. We talk about green energy. As we move in that green energy space, you need to have AI that actually really stabilize the grids, which we built already into it. Or think about factories, right? We think about picking, pick and place. So things in the bare house today, it's very repetitive. It's very hard for workers to do. Now we can train robots really, really easily now based on AI to actually do it very, very precisely. And so forth, therefore liberating the, of course, the factory workers. So all of this is happening already as we speak. Now, the nice thing about generative AI is two things. Number one, your language. So for our customer service, it's great. For marketing, it's great. For actually tearing down cultural differences when it comes down to languages so that people do understand. You have a service ticket. It's automatically then translated, then sorted, and then actually put back into the database. Perfect. So that really works well. And now what we're experimenting with is can be used for coding because we get 24,000 software engineers. And think about the potential if you were able now to use it for code generation. So there we're not quite sure yet. We look at it as it can do wonderful documentation. It can do wonderful validation and testing. But for code generation, we definitely need to find platforms as Nigel was explaining where we may have to reinforce it in terms of the learning so that we get this right, which probably will be more proprietary than actually as a major platform for everyone to use.
Philip, we were talking on the phone last week about generative AI and running through some fairly amusing cases of it going really wrong. Hallucinations, love a weird chat with a chatbot, inaccurate information. Do we think this is teething problems? Is it the problem of the AI? Is it the problem of us using it wrong?
Yeah, I think these algorithms are like confident toddlers. They'll just bl
Peter, is the answer policy and regulation? I mean, we're both Europeans. The EU is a leader when it comes to tech regulation, possibly not tech. Is this the answer?
Well, right now, definitely we see a lot of papers being published. And then, as you said just yesterday, the EU AI Act coming through after GDPR, after, of course, the EU Data Act, everything that we're discussing. We need to be careful, in particular given the pace of technology. Because think about how long it takes legislature to actually come up with a law that then already is going to be outdated by the time the technology really has advanced. So we need to make sure, of course, that the consumer is safe. Safety is absolutely non-discussable. So that's fine. But then we still need to hold people accountable and companies accountable that if they use AI, you are on the hook. If we at Siemens, we provide you services that are AI-based and are not working, guess what? It is Siemens' accountability and liability to make sure that this is the case. So we need to be a little bit careful because if you are competing on a global stage and we're discussing about how we Europeans actually will see more tech in Europe, and I don't believe that regulation will be per se the answer. It's great that we are leading that, but how about we really create these jobs, these products, platforms that we can pull across the globe because we're in competition with China, we're in competition with the United States. And that, I think, we have to be really careful about that.
You get the last word. So is the world going to end? Should we pause AI? And is the answer regulation? You've got two and a half minutes.
Two and a half minutes. Well, first of all, I want to compare this to blockchain and Bitcoin. It was invented back in 2010. And think about we're finally arriving at some of the regulations related to that technology. We regulated and we managed that technology, I would tell you, poorly for way too long. We are thrilled at Google. We invented the transformer algorithm back in 2017, which underlies a lot of this core technology. And in 2018, we built a set of principles inside of Google, things like being beholden to humans, not reinforcing unfair bias, privacy by design. And we knew then what the opportunity and also what the risks were of this. And so we implemented principles. We're very much in favor of regulation. Like I would say smart regulation, but regulation that doesn't stifle innovation, agreed 100%. But we're really thrilled with the whole world this early on in the development of this technology. I would tell you it's only been released to the world in about the past year. And already the whole world is talking about using this responsibly. And so, no, I don't. I have more faith that if you can write a line of code that does something bad, somebody can write a line of code to actually protect you against it. And I'm glad that we're having the conversation so the technologists actually have guidance on how to do it.
And I think also what's so positive is that the pioneers of this technology are actually leading the conversation and are speaking to governments. And there is a conversation being had. So hopefully AI can be all about the potential in the future and less about the risks. Perhaps when we're back here next year, the conversation will have moved on, I'm sure, once again. Thank you very much to all three of you. That was wonderful. Thank you. Thank you very much.