SmashFly’s Thom Kenney warns Big Data Belfast conference about the dangers of AI, ML and Big Data
President and CEO of SmashFly Technologies, Thom Kenney, delivered the opening keynote address at the Big Data Belfast conference at ICC Belfast, the newly rebranded Waterfront conference centre. Attendance at the event was up 50% from last year, making it the largest big data event in Ireland.
Kenney challenged the audience to consider what they mean by big data and take seriously the bias and discrimination that is present in automatic decision-making.
He referred back to 1997 and the software context of when the term ‘big data’ was first coined. Back then it referred to leveraging data that could not exist inside a traditional relational database system. Today, the scale of data, the computing power to hand, and the processing that’s being executed have all hugely increased.
“Do you really have AI just because you have a machine learning algorithm running on top of big data?” asked Kenney. He argued that there’s a blurring of understanding as terms like artificial intelligence, machine learning, and big data are bandied around. Often an organisation just has a good algorithm and they’re teaching it how to learn.
Kenney identified some dangers that can be introduced when big data is combined with AI. There can be a monetization trap that prejudices business decisions. There’s also the risk that mistakes will be relearned and reinforced within the system, perpetuating poor quality outputs. An implied bias can also creep into its learning, influenced by staff opinions and beliefs.
There is also the danger of wanting to solve a problem – indeed any problem – by using the big data that has been gathered but no one has the slightest idea what to do with!
SmashFly delivers software solutions to help some of the world’s largest companies run more effective recruitment campaigns. In July it announced the creation of 70 new jobs in Belfast.
Kenney addressed the need to be conscious of discrimination in decision-making algorithms. Given the volume of applications, it’s common practice for technology companies and recruiters to automatically sift CVs. Traditional biases have been found to continue in this new automated world, however.
Amazon has recently admitted that its machine learning algorithm perpetuated the tech industry’s bias against women. Analysing past data and decisions to teach a system how to be successful will not eliminate built-in misogynistic patterns. Firms need to be cognisant of discrimination discovery techniques and pattern mining and be ready to make adjustments.
Kenney finished by explaining the UK’s role as a leader in ethical machine learning and AI technology with the launch of the government’s Centre for Data Ethics and Innovation, which is working with international partners to build a common understanding of how AI can be deployed in ways that are safe and ethical.
He outlined an example of a for-profit US weather service that uses free data produced by the government. The commercial company accurately predicted the path of a tornado but restricted its warning to paying customers. The government weather service didn’t utilise its own data to perform the same analysis. Could lives have been saved if the insight gleaned from data produced by the government for the benefit of all had not been restricted to the monetised few?
