Experts gather in Abu Dhabi to answer the question: can we trust AI?

How much, if at all, can we trust AI? That is one of the main topics being discussed and debated among some of the world's foremost computer science experts at the Adia Lab symposium in Abu Dhabi.

“It's not a fault of the technology, it's a fault of how people use it,” said Alex Pentland, one of the experts taking part at the three-day event organised by Abu Dhabi’s independent institute dedicated to research in data and computer sciences.

"People assume that the answers that come from AI can automatically be trusted in some way," he added.

“Trust it [AI] as a way of suggesting possibilities,” the director of MIT's human dynamics laboratory and Adia Lab advisory board member explained.

AI, he said, should be treated in the same realm as one of the many social networks people use to communicate and occasionally seek out advice.

Mr Pentland, who also leads the World Economic Forum's big data and personal data initiatives, stressed the need for AI researchers and developers to ultimately make data control and auditing of AI systems a priority, whereas today, it is too often an afterthought.

Also addressing the topic of AI trustworthiness at the symposium was cryptographer, Turing award winner and UC Berkeley computer science professor Shafi Goldwasser.

“The fact is, the last few years have proved that this is the biggest scientific revolution of my time,” Prof Goldwasser.

“It's amazing what these systems have been able to do in a short amount of time,” she added.

Prof Goldwasser, said that the trust issues many have when it comes to AI comes from AI developers who often cannot explain and struggle to explain exactly how AI works.

“Should we really trust models we don't understand and don't control?” she asked. “It's a good question.”

Prof Goldwasser suggested that AI models should eventually include a “verification algorithm”, to help increase overall confidence in large language models, the main backbone that makes AI possible.

“The biggest challenge is safety,” she said.

“The power of AI is clear, and the desire to adopt is clear,” Prof Goldwasser added, pushing for continuing discussions among governments to establish guardrails and sensible regulations for AI, along with the co-operation of technology firms and AI researchers.

Also addressing the symposium on the topic of trust and AI was Carme Artigas, co-chair of the UN's AI advisory body.

“It's ubiquitous,” she said, noting the importance of the topic regarding the burgeoning AI field, adding that it will continue to impact just about every aspect of life.

Ms Artigas also alluded to the mystery as to how various iterations of AI continue to learn, often to the surprise of those who develop AI tools.

“It can continue to evolve without human agency,” she explained, pushing for AI safeguards through the use of oversight boards, market incentives, transparency and regulation.

She said despite initial efforts and progress made by the UN, there's still more work that needs to be done to correct an “AI awareness deficit” around the world, and that there needs to be a continued pursuit to make sure there's inclusion across several regions and demographics, noting that much of the attention of AI has been focused on how it affects the Global North.

“Remember everything we know about AI has just been trained with data from the Global North, talent and computing capacity,” she said. “This not sustainable for trust.

Although research on the topic of artificial intelligence has been continuing since the 1960s, recent advancements in computer processing power, coupled with iterations of AI solutions such as OpenAI's ChatGPT in 2022 led to an explosion of interest, investments and start ups in the AI tech ecosystem.

Along with that interest has come concern from technology leaders, government regulators and elected officials who see both the potential for AI-driven solutions as well as the downsides that could come to fruition in the form of labour market disruption, along with hypothetical scenarios where unchecked AI could become too powerful and uncontrollable.

Those worries have led to the rise of various artificial intelligence safety summits, executive orders to establish AI guardrails as well as the creation of AI ministerial positions in various governments.

In the US in particular, a recent Pew poll indicates that although more citizens than ever are using AI tools, those same citizens grown more anxious and concerned about the long-term effects of AI.

“52 per cent are more concerned than excited about AI in daily live,” a highlight from the Pew poll explained, which added that only 10 per cent are more excited than concerned.

The Adia Lab symposium, now in its second year, is focusing on the topics of sustainability, innovation and the idea of trustworthy AI throughout the three day event.

“We’ve had tremendous interest in this,” said Horst Simon, world renowned computer scientist and director of Adia Lab in front of a packed auditorium inside Abu Dhabi Global Market.

“That shows how far we’ve come,” he added, explaining that since its inauguration in 2022 by the Abu Dhabi Investment Authority, Adia Lab has sought to focus on innovation at the intersection of data, AI and computational science, while also advancing global collaboration through impactful research,

“The goal of Adia Lab is to build an international network and to bring capabilities to Abu Dhabi,” he said.

In 2024, Adia Lab expanded offerings through increased fellowships, research projects, and new collaborations with universities and research institutions around the world.

It also set up a European headquarters in Spain.