At the end of her talk, Scottish AI Alliance Chair, Catriona Campbell analogously described the seemingly overnight proliferation of OpenAI to when Google essentially took control of the search engine market through it’s clean interface and accessibility.

It’s this democratisation of tech, as well as the relative ease-of-use of these powerful tools that now has Government’s and regulators back-pedalling to get a grip on a technology that’s developing faster than any legislation can reconcile in any kind of compelling way.

This all begs the question, how will Scotland navigate these choppy technological waters, and broadly speaking, how will it impact Scottish citizens and businesses?

Recently appointed chair of the Scottish AI Alliance, Catriona Campbell, took to the Traverse Theatre stage in Edinburgh at Burness Paull’s Tech & IP Conference to discusss these issues.

AI & The Trust Economy 

Trust has always been a fundamental component of evolution, and the same anthropological principles can apply to emerging technology – the less trust there is, the less adoption we subsequently see.

To illustrate this, Campbell cites the psychological concept of the trust framework, based on three components: ability, integrity and benevolence.

Campbell says: “Ability regards being able to complete a transaction, integrity, which shows that the transaction has been delivered properly and appropriately and with benevolence, the fact that the product or the solution is acting in your best interests.”

It’s here that Campbell then makes the important distinction between the most common kinds of AI because, inherently, they interact with this trust framework quite differently.

Campbell says: “Explainable AI is the way that an AI algorithm uses datasets, making decisions out of those datasets. Previously, all of the AI that had been designed and developed – we’re going back maybe three, four years ago – was mostly explainable.”

This was because, largely, you would know what data was going in to the AI model so there was a traceable or explainable element to any decision making or action from the AI.

This is beneficial because with this kind of AI, you can control what data you feed into the system, controlling for bias while maintaining an ethical standard.

OpenAI, on the other hand, works quite differently.

“Open AI has been training itself over the last few years, on a huge data corpus. It’s developers haven’s released much information about its data sources.”

While generative AI tools like GPT-4 generate “net new” info, it’s data sources are unexplained beyond the fact that we know that the model is trained on a snapshot of the internet in 2021. Beyond that though, OpenAI is fairly secretive around the specifics regarding the data that it uses.

It’s because of this that the issue of trust becomes so central when we talk about generative AI. If you’re leveraging a tool like GPT-4 for decision-making processes, you’re not only testing trust with regards to the quality of the decision made, but the data used and compiled to reach that decision – was the data obtained in a GDPR compliant way?

“If you’re not in control of the third party software that you’ve deployed to make those decisions, then that is a serious breach in my particular view.

“If we look at the trust model, new and highly complex systems are going to need super high risk assessors within an organisation, they’re going to have to understand their third party solutions that provide them with their AI. And they’re going to have to understand the knowledge that they are internally using in order to develop their AI.”

Is Scotland AI-ready?

In Campbell’s view, the best way to responsibly leverage AI with regards to the Scottish populace is through education and transparency.

To this effect, Campbell and the Scottish AI Alliance are rolling out Living with AI, an “educational series for anyone from the age of five, to 95, that will be available to every citizen in Scotland in June.”

With regards to Scottish business, Campbell strongly believes that adopting an educational and transparent approach to data and data-driven AI/ML will drive demand from customers.

She added: “For organisations to do this properly means analysing your data, ensuring that the third party AI you’re using is efficient, effective and is satisfying the needs of your customers.

“But it’s also about auditing yourself, you need to hold yourself accountable. Independent AI audit auditing, and I don’t mean by someone like me, but independently doing that within your organisation from your risk team. Independently auditing what you’re doing in each of the different departments is really, really important.”

Beyond doing this as an ethical exercise, Campbell feels there could be a legislative imperative to this, saying that at some point, there’s every chance the UK Government will demand audits like this.

Dovetailing the idea of independent AI auditing to the notions of trust she discussed earlier, Campbell says: “I think this is a really difficult area to retain trust when it’s such an incredibly complex system.”

Recommended

• Comment: Is Quiet Hiring the Solution to the Tech Skills Shortage?
• £50M Space Infrastructure Fund Launched by UK Government
• Glasgow Startup Ecosystem Report: Tech Sees Continued Growth

As far as Scotland goes, Campbell is clear on the organisations that now need to step up to catalyse AI regulation and education.

She says: “The Scottish AI lines are made up of Scottish government, the private and third sectors, as well as education and academia.

“As a group, we are trying to ensure that our citizens are armed to understand what AI means for them, with organisations such as the Scottish AI Alliance serving as a centralised one stop shop to go to for information, and that we are ensuring that there are standards.

“I am faithful that we’ll get to some kind of international regulation at some point. And we will stand ready to support that when it happens.”