An interview with:

David Jones
Non-Executive Board Member
Ofcom and Qualifications Wales

 

___________________________________________________________________________________________________________________________

Wales Tech Week in October, created by Technology Connected, has been designed to help professionals from all sectors keep pace with rapid technological developments.

Less than a year since the launch of ChatGPT and other generative artificial intelligence models, this is an important discussion topic for the summit. Kate Rimmington, from Technology Connected, talks to software expert David Jones, who is non-executive board member of Ofcom and Qualifications Wales, to get his take on the latest developments.

KR: David, AI is all over the news – give us a quick overview.

DJ: AI isn’t the same as normal software. We’re used to software following rules, often quite complex rules in programs like Excel or Facebook – and whilst the software is powerful, it’s easy to understand how Facebook interacts with photos or how Excel interacts with numbers to give its results.

AI is entirely different. AI doesn’t really follow rules at all. It looks at previous sets of data and predicts outcomes. If, say, you wanted to examine a photo of a dog and determine its breed, that’s pretty easy for a lot of dog lovers. But if you try and write down the rules around what makes one dog a corgi and another a chihuahua, then it gets a little trickier. They are both small dogs, but the Corgi is stockier and has shorter legs etc. But that rule on leg size is not enough.

AI learns to spot if a photo is a corgi or a chihuahua based on its “training” on thousands, maybe millions of photos of Corgis and Chihuahuas and looking for patterns.

KR: One of terms we see on AI is “Large Language Models”…what does this mean?

DJ: A Large Language Model is a fancy name for a database of existing written text which can be used to predict which words and sentences can be used to express a view or make a point – ChatGPT is an example. In the dog example, it’s more a Large Dog-Photo Model.

KR: So how has AI suddenly become such a big deal. Is it all just hype?

DJ: It does feel like an overnight sensation, but it’s really been progressing slowly for decades and it’s good for all sorts of different kinds of computing problems.

The recent breakthroughs have been the sudden developments of better computing technology for these Large Language Models, plus the vast power of new cloud computing.

I think that the principle of Large Language Models is here to stay, but the way that AI will impact so many areas of life – that’s something we can’t even imagine yet.

KR: Some people have expressed extreme-sounding concerns it could develop into some kind of apocalyptic force?

DJ: On the end-of-the-world scenarios, I really don’t think this is a helpful way of looking at things. But I can understand where the point comes from. Go back to that comparison with traditional software, versus AI. Traditional software has rules and processes – and that’s all logical. It may take some time, but a skilled programmer can usually determine what any piece of software does, how it does it and why it works. AI really is different, it’s not transparent in the same way. But that doesn’t mean that AI will take over nuclear weapons and destroy the world, or choose to set all the UK traffic lights to red and create traffic jams (or set them to green and make cars crash).

KR: How worried should we be, as some have predicted, that it will eliminate huge numbers of jobs?

On jobs, this is where it gets interesting. Many managerial jobs involve writing, with reports and email. Chat GPT, using its Large Language Models, can cut 90% of the time for many writing tasks, so it’s dramatic and important. Whoever is producing the work still needs to think and review the output, but it has the potential to be a huge time saver.

The people using ChatGPT are not shouting about it – why would they? And it’s important to understand that I’m not just talking about people who are professional writers. It’s everyone.

KR: And in education, don’t ChatGPT, Google Bard etc allow for extensive plagiarism?

DJ: It does, yes. It’s the equivalent of using a calculator for arithmetic. But like a calculator, education policy makers and regulators need to find a way to adapt. Because the kids and learners have already found ways to adapt and are using it on an industrial scale.

KR: Any thoughts on the Creative industries?

DJ: We know that Wales has strength in this sector. I’d suggest that US entrepreneur and investor Marc Andreessen’s brilliant AI essay, has it right. He said: “The creative arts will enter a golden age, as AI-augmented artists, musicians, writers, and filmmakers gain the ability to realise their visions far faster and at greater scale than ever before.”

While all writers and image creators will be vulnerable to AI’s disruptive influence, it has potential to significantly boost creative output.

KR: Wales Tech Week is aimed at people in organisations of all sizes and sectors – do you have any advice for individual professionals and entrepreneurs on using Large Language Models like ChatGPT?

DJ: It reminds me of how the internet developed. I was CEO of a software company, Travelink, about 15 years ago as the Internet was taking off. In those early days, there was almost a view from some that the Internet was a finished product – it would be great to buy books or scheduled flights, but that would be about it. Now we know that almost all businesses need to use internet technologies, smart phones and apps. It’s not an optional extra. None of the success of, say, Instagram was predicted. I think it’ll be the same for AI but it’s a mugs game to try and predict where this will end up.

Having said that, policy makers, regulators and venture capitalists are trying to predict the future of AI. But for 99% of professionals, I’d suggest that your time is much better spent on doing what I saw back 15 years ago, which was to dive in and grasp the technology as it exists today. Back then it meant CEOs being active on the Internet, buying stuff and using Paypal etc. Today’s equivalent is ChatGPT and many other cheap and accessible AI products. So have a look into “prompt engineering” – essentially learning the best way to give it instructions – and have a go.

Remember that Large Language Models such as ChatGPT use statistical prediction, so the information they produce may not be 100% reliable. Whatever they output should always be treated as a draft and “sense-checked”. They are incredible tools – but ultimately you are still in charge.

Wales Tech Week takes place 16-18 October 2023 at the ICC Newport. Among the speakers are Artificial Intelligence expert Danilo McGarry and Microsoft Chief Security Advisor Sarah Armstrong-Smith.

The summit features expert discussion panels, an exhibition, demo zone, Start Up Alley, international link ups and more. Explore the packed agenda and register for your free ticket at walestechweek.com.