With a stacked roster of cross-sector speakers, spanning: energy, gaming, chemicals, Artificial Intelligence (AI), cybersecurity and VR, DIGIT North, the 10th Annual Tech Innovation Summit, held on Thursday 25 May at the P&J Live brought together some of the best and brightest in Scottish tech.

The interplay between data science, OpenAI, Machine Learning (ML) and how each of these can interact within business operations – with regards to parsing data, improving efficiences – is a hotly debated topic, with many evangelising the virtues of OpenAi and its potential to streamline many business operations.

Conversely, there’s no shortage of cautionary tales around the pitfalls of ‘unexplainable’ AI – yes, it sounds human and confident; unfortunately, it doesn’t mean it’s always right.

It’s a point that Colin Gray, Senior Lead Data Scientist at famed game developer Jagex, illustrates in a panel discussion featuring the day’s keynote speakers when he questions the ability of GPT-4 to answer technical queries with empirical voracity considering that it gives a different answer to the same question every time you ask it.

In his full keynote, Gray extolled the virtues of tried and tested data science methodology to deliver value to businesses, offering case studies through his work with Jagex on the wildly popular MMO Runescape.

What do Data Scientists do Well?

It can often be difficult for business to know where to start with data science, Gray himself concedes that he often feels “overwhelmed trying to keep up with the options and the modelling techniques.”

Gray thinks that one of the biggest problems he sees is – even when you have something resembling actionable datasets – putting things into production.

Talking about the best way to avoid this, Gray says: “Achieving a trusted method for data modelling is essential, it must be sustainable and maintainable. Go for the things that you understand and you can maintain.

“When it comes to levelling up for your business, my advice is always to be early and often. Get models out there that will help you build trust. But really importantly, you’ll learn more from a model in production than one stuck in testing or conception phases.”

To help achieve this, it’s best to have clear idea of your business goal and end product, what frequency should you model(s) work with, as well as software, data and deployment requirements.

“One of the two things that I see a lot of, especially in younger data scientist is that they don’t visualise things enough,” says Gray.

He adds: “They also don’t communicate with the business enough to keep that constant flow.”

What Should Your Goal be With Data Science?

“Always go back to what you’re trying to achieve and try to make a difference for your business, and be constantly engaged in your business.”

If you want to level up your business with data science, the primary focus of this is as a business outcome, not the data science itself – Gray advises to always stay focused on what you can achieve.

The best way to do this is to work backwards. Have an attainable business goal or functionality you want to implement based on data, and have your data scientists work on this – bearing in mind to just craft one model at a time as adding more will overly-complicate the process.

Recommended

• How Can Scotland Navigate the OpenAI Explosion?
• Survey: 58% of Firms Have Had a GDPR-related Data Breach
• BR-DGE and Volt Partner to Help Merchants Prevent Payment Fraud

Gray says: “There’s opportunities everywhere in your business, don’t think you have to start with big things.

“Little things across your business can have a big impact overall. Engagement with people is key to get rolling on this: employ a data scientist who’s a good communicator, but overall, really be realistic.

Companies like Microsoft and Google, they have huge resources, huge teams and decades of experience. If you’ve only got two or three people and a couple of data engineers, be realistic with what you can achieve, make sure you don’t overwhelm yourself and then actually build out these small, high value products.”