Q: What is AI? 

A: There is no simple answer! AI is a collection of technologies and techniques that aim to reproduce the intelligence that we see in animals and people inside machines, particularly computers. 

Q: How does AI work? 

A: There’s no single way in which AI works – it’s a range of technologies. The simplest perhaps are expert systems which are rule-based, and you have a whole load of rules which can be thought of as ‘if this is true or if this is false, then do this…’. They’re widely used in medical diagnosis and fault finding in complex engineering systems, for example. An alternative approach is to mimic the way that simple animals work, such as insects, so a lot of robots will walk and move and behave and interact with the world based on the way we know insects do. The most complex way, and the one that is picking up all modern technologies, such as machine learning, is based on something called a ‘neural net’, which is a model of how the brain works, in a very simplistic way. It’s based on training huge amounts of data into a software system – the neural net –  which mimics a human brain and it learns by picking up this training, repeating it and testing it, and then, if it works, fine! If it doesn’t work, you train it some more! 

Q: What is AI actually capable of doing? 

A: You’re using AI every day if you’re on social media. AI is classifying the information you see. It tells you what you want to see, and it hides things you don’t want to see. If you go shopping on a big site, such as Amazon, it gives you the recommendations of what they think you’d like to buy. It’s being used by companies like Spotify and Apple Music to show you music you may not have seen before, because it knows what your music tastes are. It’s used for classifying, so it can be used to identify cats from dogs, to give a simple example. It can identify things such as tumour cells in a medical scan; it can identify galaxies in photographs or from space; it may be used for identifying faces in a crowd which can be used for crime detection; it can be used for identifying photos in the crowd for your social media feed.  

The big burst of information of interest in the last few years has been in what's called ‘large language models’, such as Chat GPT, which are capable of generating rich interesting fluent text in almost any language, and translating between languages, summarising languages, rephrasing, restructuring and doing things that humans are basically really good at. A related field is image generation, so there are a lot new of new technologies around for fixing photographs, or creating completely new images that no one’s ever seen before. 

Q: Is there any way of us knowing what is AI generated?  

A: It’s getting tricky. AI is developing incredibly quickly in this area. Huge amounts of money, time, research and computing power are being put into making better AI systems, but there's a sort of ‘artificiality’ at the moment to both the text that is generated by AI, and the sound that AI can generate or the images it produces. So, if you read AI-generated text, it's very fluent, it's usually grammatically correct, but it's boring! It doesn't take strong opinions because it's been programmed not to be controversial. If you look at the images, they have a sort of ‘waxy’ appearance to them, but that's rapidly disappearing as the technology improves. The quick way of seeing if an image is AI-generated is to look at things like hands – AI really struggles with hands, faces, the details that we as humans pick up on, so that's what gives the strange artificially. But if you don’t think an image is real, go check somewhere else – does it exist anywhere else on the web? Look on a search engine, or a reputable news source. Don’t rely on a picture being real just because it looks real the first glance. 

Q: How close are we to the fully self-driving car and how much of that is down to AI? 

A: It’s an idea that’s been around probably since the car was invented, but it’s becoming possible thanks to AI. Modern cars already have a lot of AI in them, so things such as lane guidance, making sure you're following the road, maintaining distance, reading road signs, things like that, are all done with AI technology, but it’s still you actually controlling the car. There are lots of companies experimenting with it and there are some cars on the road that have, more or less, self-driving capabilities. It’s incredibly complex because roads are complex. Most of the cars on the road won’t be being driven by computer, so the computer has to try and work out what it’s like to be among human beings. There are people crossing the road and they don't always follow the rules, and neither do other drivers. It’s not here yet, but it’s perhaps 5 or 10 years away. The biggest issue will be things such as the law and insurance, rather than technology. 

Q: Does that mean AI works better when operating with other AIs, rather than with humans? 

A: In theory, yes, AI would work better with other AIs because it could communicate with them at incredibly high speeds across a wireless network, or some other form of networking. It’s rather that humans are incredibly unpredictable. I’m not saying AI is entirely predictable, but humans are incredibly unpredictable things! 

Q: What role does AI play in social media?  

A: The two are intrinsically linked. AI curates what you see, so when you go onto your social media feed and you get recommendations for posts that are interesting, or people who are interesting, the reason for that is that AI has been monitoring what you’ve been clicking on in the past, and what people who have similar interests to you have been clicking on, and it makes recommendations. AI is also working behind the scenes in a not-so-good way with ‘bots’, which are basically software programs that generate content that will create clicks and therefore money for the people who operate them. They can also be used to influence behaviour. Political or extremist behaviours can be driven by malicious AIs and the social media companies try to crack down on these things. But the software is always evolving and there’s a lot of money and power behind the people who do create these bots. 

Q: How close are we to the intelligent machines we see in movies?

A: There is a field called ‘artificial general intelligence’, which is a study of technology that doesn’t exist yet, and artificial general intelligence is a machine that will be as intelligent or more intelligent than a human being and can do the full range of tasks a human could do. So, to give some examples from the movies. The most famous are the Terminators, which are bad! Then there’s also the very famous killer computer from 2001, Hal 9000. But we are not close to them yet in the sense that we have not created a truly intelligent machine. AI can produce images and text and music which looks superficially real, but it doesn’t yet understand what’s actually being generated. If you watch 2001 – it’s a very long movie, but it’s a very good movie – you’ll see that Hal 9000 has conversations with its users, it can play chess, it can lip read and, in the end [spoiler alert!], it murders people because it has intelligence and it realises it has its own priorities, which differ from the people. We’re not there yet though. There are some interesting things going on, and some people say we might be 5 or 10 years away, and some people say we will never get there.  

Q: What is your own personal favourite example of the best use of AI that you are using?  

A: I do take photographs, and I sometimes travel a long way to take them, and you get there and the weather sucks! There are a whole range of packages now that will let you use AI to do quite sophisticated changes to photographs. So you can could replace the sky if you came on a cloudy day to a blue sunny sky, you can put reflections in the water, you can even move the light around.  

Q: Are there any ethical concerns around using AI?

A: There are. It goes from things such as the ethics of the huge amount of energy that is consumed by training and using artificial intelligence. There are ethical concerns about the data that goes into the training. Some of it might be copyright and not actually being allowed to be used by by the companies creating artificial intelligence. So there are a number of legal cases underway particularly in the United States about books, movies, pictures and music that’s been incorporated into AI without permission. There are ethical considerations about the range of data that goes into the training sets. Sometimes certain groups of people are over represented or under represented in the training set which means that what comes out when the AI is used may not be accurate. So there have been cases such as Black people being unfairly associated with criminality and women and ethnic minorities also being under represented in the images that come out. There are also ethical issues about how you use it particularly if you’re using it at work or in school. Are you actually creating your answers in your assessment or is the AI, and by using it are you missing out on some useful skills that you will need in future lives?

Q: Can AI take my job?

A: It’s too early to say, although it’s certainly going to change a lot of jobs. One thing AI is really good at is going through boring tasks, so if you’re working in a career where you have to summarise vast amounts of information, AI is really good at that. For example, going through medical results looking for possible cancer cells and things like that, humans are not good at that because we get bored. It’s really murky data. AI can be as good if not better than humans, so some roles are certainly going to either disappear or change as AI comes along. Perhaps some of the boring work we currently do will done by AI which is good for all of us and obviously new careers are going to appear in and around AI applications as the AI companies get bigger and bigger. And there are more applications for it so some jobs might disappear, some will change and there will be new jobs coming along but it’s really early days and I don’t think anyone can honestly say what the future’s going to hold.

Q: Can you give one pro and one con about AI?

A: One pro of AI in 2024 is it can do some of the tasks I find really boring. Or if I get stuck getting started with a long piece of work where I’ve got ideas but I don’t know how to structure them I can I throw it into a program such as Chat GPT. It’s also really good at looking at large chunks of text and summarising it sometimes I’ve got very boring text which goes on for pages and I need to know what the key information in it is. It’s great for that.

A con of using it right now is that it’s still really flaky. You can’t guarantee that what it’s telling you is the truth, particularly without doing a lot more research. So sometimes you think it’s going to be a time saver but in reality it isn’t.