Businesses should think long and hard about whether AI is the best solution for their problems and invest in auditing to avoid unfair, biased and inefficient results, according to Eticas Tech CEO Gemma Galdón-Clavell.
“There’s so many examples of moments when I’m like, ‘Why? Why would you think that AI is useful for this?’,” says leading AI auditor.
Did you know that according to Eurostat, 8% of EU companies used artificial intelligence (AI) technologies in 2021? Out of these, 53% had bought ready-to-use commercial AI systems and intelligence software.
Denmark had the most companies using AI, accounting for about 24%, with Romania and Serbia having the least, at around 1% each.
Most businesses used artificial intelligence in the communication and information sector. The companies which are for AI believe that it helps in data analysis, content creation and language translation amongst others. However, firms against AI mostly felt that it was irrelevant to them, expensive, not reliable and increased privacy risks.
In this episode of The Big Question, Angela Barnes met with Gemma Galdón-Clavell, the CEO of AI company Eticas Tech to discuss whether AI use is ethical and where we need to avoid using it.
What is AI good at?
According to Clavell, “AI is so many things at the same time. Understanding what it is that AI does is crucial to making the most of it.” This can make it difficult to pinpoint what the technology can be most useful for, especially if many businesses want to use it for several different things at once.
As such, businesses need to actually think about the problems they want to solve, and then consider whether AI is the best solution. Essentially, AI is a data process that is best at providing solutions for problems which have massive amounts of historical data.
This includes online shopping and streaming recommendations, based on search history, customer viewing data, ratings and location amongst other things. AI is also used for global positioning systems (GPS), which can access years, if not decades of location and movement history from millions of users across several cities. Chatbots are another example, as well as predictive texting.
However, for unique problems, or ones which do not have access to a lot of historical data, “AI is not your friend”, says Clavell.
Gemma also warns, that’s not just about how to use AI but of the importance of choosing good quality AI software.
“If we just buy an AI package and implement it and trust that things are going to go well, we’re going to be making a mistake. And that is really expensive,” highlights Clavell.
Is AI use ethical?
One of the biggest concerns with AI is that it uses such an immense amount of personal and behavioural data about how people live, work and relate to others.
Not only does that lead to huge security and privacy concerns, but this can also mean that AI conclusions can often be skewed or unfairly biased, as well as inefficient.
The reliance on historical data can also be a problem in many scenarios, causing unfair exclusions and biases, often without humans realising that they are happening.
As Clavell explains, “If at a bank, the past data reflects that your ideal client is a man with a stable job. AI tends to reproduce that. And so, on day one you will be discriminating against women, on day ten, you will only be giving mortgages to men with stable jobs.”
However, one of the key solutions to identify and mitigate discrimination is auditing AI packages and services, much like clinical trials for vaccines.
She points out that failing to do so would be like, “getting a vaccine that didn’t go through clinical trials, buying a car without a seatbelt, buying a house without the paperwork.”
At the end of the day, “we do not want discriminatory AI,” says Clavell.
How can we make AI better?
Clavell outlines that one of the key ways to make AI better is by not just theorising about having audits and other precautionary measures and governance in place, but also implementing them. As such, while Europe is very good at coming up with regulations, like the Digital Services Act, the US is far better at implementing them.
“Quite interestingly, even with the General Data Protection Regulation (GDPR) that was passed in 2016, the company that’s made the most money out of that regulatory change is a company based in Texas, […] that is now worth 5 billion,” says Gemma.
“I fear that with the AI Act, the EU will lead the way in regulation but the part of the market that will make the most of the opportunities created by the regulation will not be in Europe, or a lot of things would need to change for me to see Europe playing a role in leading not just at the regulatory level but also at the market level.”
The milestone AI Act has recently been launched in the EU, aiming to make it a pioneer in AI regulation. The act will also try to set a solid legal framework for AI regulations across all the EU’s member countries, which will make sure that the technologies are safe and do not violate privacy laws and values. However, at the same time, the act will also try to draw more AI investment into the continent.
According to Statista, the EU had about 6,000 AI companies in 2023, which was still significantly behind the UK, at 9,000 and the US at around 15,000.
Other ways to make AI better are to feed it better and fresher data, which in turn, will also allow it to take more factors and variables into consideration. In the absence of real data, synthetic data or hypothetical cases can also be used to simulate scenarios and provide more context to AI tools.
Improvements to the algorithm can also be made to ensure that companies are not filtering out the wrong clients or employees. This can be done by removing often unnecessary data parameters.
However, Clavell warns that until we put auditing into legislation “as a society, we will not have guarantees that the AI around us is good quality AI.”
The Big Question is a series from Euronews Business where we sit down with industry leaders and experts to discuss some of the most important topics on today’s agenda.
Watch the video above to see the full episode with Gemma Galdón-Clavell now.