Nordic enterprises are actively seeking innovative strategies to integrate ethical and responsible practices for the use of generative AI within their business models, while also ensuring profitability. They face a challenging journey ahead. Report from a webinar as experienced by Nordic Scrutinizer.
‘Last year has been crazy. I’ve never seen a technology elicit such a strong reaction among customers and clients,’ said Microsoft’s Director for Azure Data & AI in Norway and Denmark, Jon Jahren. He appeared weary behind his glasses, which he donned in the manner of an architect. Representing a major player in AI with Microsoft Copilot, formerly known as Bing, he sensed that people are weary of discussing trust, yet he welcomed the discussion on AI risks, even though the focus is more on what the technology can achieve. We will soon return to Jon Jahren after presenting a bit of context.
In 2023, generative artificial intelligence (AI) became accessible to internet users globally, often at no cost. This groundbreaking development meant that anyone could interact with sophisticated software by using ‘prompts’ or plain language instructions to create intricate texts or visual content—tasks that previously required the expertise of a trained professional. The advent of this technology took the world by storm.
‘ Consumer rights need to be in the center of EU’s AI Act.’
user in the chat of the webinar
One year on, technological advancements are accelerating at an unprecedented pace, prompting professionals and businesses alike to engage in critical discussions on leveraging artificial intelligence for societal end economical benefit while circumventing its potential drawbacks.
While AI researchers engage in discussions about the timeline and categories of jobs that will become redundant for humans in the coming years, and also consider the potential threats that could lead to humanity’s demise, organizations in the Nordic countries are actively seeking ethical and responsible approaches to employing AI.
The AI Sandbox
Nordic Innovation, a body of the Nordic Council of Ministers, spearheads the ‘Nordic Ethical AI Sandbox’ program, which engages a consortium of Nordic companies and specialists. The initiative aims to shape the creation of guidelines and recommendations for responsible artificial intelligence. At the heart of the program is the belief that ethics, along with trust, human rights, and cooperation, are integral to the shared values of the Nordic region. Therefore, the program endeavors to integrate these core values with the dynamic realms of technology and innovation. The aim is to release the findings in the ‘2024 Ethical AI Playbook’ come spring.
On February 15th, Nordic Innovation hosted a webinar on the learnings of the most crucial factors for deploying trustworthy AI in the Nordic countries—a region that combines a population of 27 million and the world’s 11th largest economy. The webinar was chaired by the consulting firm Accenture, with representatives from telecom enterprises Ericsson and Telenor, the consulting firm Saidot.ai, and the software company Microsoft.
Part of the discussion was based on takeaways from an earlier industry workshop with participants from companies: Factiverse (Norway, AI), Innsikt (Norway, Marketing), Equinor (Norway, Oil), Íslandsbanki (Iceland, Bank), Aidn (Norway, Health), Microsoft (US, IT), Apheris (Germany, IT), Saidot (Finland, AI), and Volvo (Sweden, Cars). Here the participants, according to moderator Andrea Risberg from Accenture, had stressed their many expectations for the new technology but also expressed a lot of concerns regarding where to start and how to mitigate risks with their AI projects.
Inherient AI risks
During the workshop, there was initial uncertainty regarding whether companies would need to devise entirely new business practices to integrate AI responsibly. However, it was ultimately concluded that they should enhance their existing practices instead. Central to the business discourse were pivotal questions: Is the investment justified, can we achieve the anticipated improvements in efficiency, and do we possess the capability to execute an AI initiative that yields dependable results?
There was a consensus that companies should prudently initiate an internal project and assess their current governance frameworks to determine which policies and guidelines require reinforcement. Equally crucial is for the company to define an acceptable risk threshold, acknowledging that implementing AI comes with inherent risks.
Panelists at the webinar: Ieva Martinkenaite (Telenor) , Jon Jahren (Microsoft), Milap Patel (Ericsson), Veera Siivonen (Saidot)
In the panel discussion at the webinar the importance of starting with a pilot project with clear accountability was stressed by head of Research and Innovation Ieva Martinkenaite from Telenor: ‘We have been utilizing machine learning for many years. The emerging generative AI will be significantly more impactful and is evolving rapidly. To be successful with this new technology, we need cross-functional teams and the engagement of top leaders. The discipline of data governance is becoming increasingly essential – determining where the data is and who owns it.’
Third party dependencies
Milap Patel, the Head of AI Products at Ericsson, also supported her perspective, emphasizing the vast potential of AI technology. He highlighted the importance of meticulously selecting use cases and ensuring robust accountability measures. Patel further remarked on the necessity for companies to adapt, noting that they cannot operate their legacy systems or utilize data in the same manner they historically have, due to the rapid pace of technological advancement. However, he did not elaborate on the practical implications of this change, and the moderator did unfortunately not probe further.
AI consultant Veera Siivonen of Saidot did raise concerns, emphasizing that the idea of ‘AI Governance’ is not well-understood by the majority. She advises proceeding with caution, suggesting that AI initiatives be approached with the same meticulous quality management seen in the healthcare sector. Drawing parallels to medical device regulations, which are inherently designed to cause no harm and include safety measures in case of failure, could serve as a valuable model for AI project development.
She discovered that third-party dependencies, such as OpenAI, Google, Microsoft, etc., come with significant risks and are a major concern: ‘They continuously update the software, and those updates can alter the foundational elements of the use cases. It’s crucial to stay informed about how the models are developing as part of the process. However, clients often overlook this, despite the necessity of doing so.’
AI bias, consumers and black boxes
Ieva Martinkenaite from Telenor further emphasized the complexities of relying on third-party dependencies in AI, highlighting its immense potential to improve customer experience while also acknowledging the inherent risks concerning privacy and discrimination. She posed a critical question: “How do you manage these ‘black boxes’ and justify their choices?” Martinkenaite advocates for a prudent, value-driven, risk-based approach to AI implementation, suggesting that starting on a small scale is preferable to launching large-scale projects.
In the chat connected to the webinar program manager Sophia Ivarsson from Vinnova, Sweden’s innovation agency, noted that the agency had funded multiple projects that aim to use AI to promote gender equality. It’s a wellknown example that programs often will translate the English term ‘nurse’ using a female-gendered word, and render ‘doctor’ as a male noun. Also in the chat, Martinkenaite referred to the EU’s first regulation on artificial intelligence – the AI Act – and reported on intense and informative discussions in Norway that involved multiple actors. She suggested that industry players should come together and co-create responsible AI solutions building on the AI Act. She concluded that to address bias, modern data governance must be established from the outset. A reference to a report from June 2023 on Generative AI threatens consumer rights was made in the chat by another user, who advocated for consumer rights to be in the center of the act.
All eyes at the webinar were now focused on the representative from a third party—the AI Director from Microsoft, Jon Jahren. Seated casually and speaking in a soft voice, he acknowledged the rapid pace of change brought about by new technologies. He announced the upcoming launch of another new AI tool set to arrive later this month. He then summarized the challenging scenario that lawmakers, consumer organizations, and businesses face when trying to regulate the development of AI. Generative artificial intelligence may be multilingual, but AI is monocultural. Since all current AI originates from one place in the world, it is inherently biased.
