With the dramatic developments within OpenAI and warnings of threats to humanity, the risks associated with Artificial General Intelligence (AGI) have risen to prominence on the societal agenda. At the same time, it is imperative for companies to embrace AI to maintain and enhance competitiveness. How does one proceed slowly, safely, and ethically? Zimply met with Anna Felländer, founder and CEO of Anch.AI, to gain her perspective on responsible AI.

In theory, AGI can become more powerful and intelligent than humans, circumvent its own restrictions, and create its own agendas. It can unintentionally or intentionally harm us through digital systems that govern society, such as drones, robots, cyberattacks, and viruses.

- We are in a historic phase where OpenAI's product, ChatGPT, has become a symbol and embodiment of the power of this new technology, where both the tremendous benefits and the risks have come to the forefront, says Anna.

In the midst of an experiment

As early as 2016, Anna and her research team at KTH identified the potential dangers of the rapidly growing and unregulated development of AI. Unintentional discrimination, privacy infringements, disinformation, and manipulation are some examples. These issues arise, in part, due to a lack of oversight and control of risks from an integrated technical, legal, and business perspective, according to Anna.

- The language model GPT-4 has now taken AI to the next level, where users of ChatGPT are part of an experiment conducted by a private entity without clear guidelines and transparency regarding shortcomings. Both the value and risks have increased even more with the rapid development over the past year, argues Anna.

Crucial to Accelerate Guidelines

According to Anna, it is now crucial to shift focus to what we can do here and now to steer the development in the right direction, where guidelines need to catch up with progress.

- For instance, asking ChatGPT about successful traits of a CEO for a job application generates responses based on norms that have historically been successful for leadership positions; that is, highly educated white males," says Anna.

AI Act on the Horizon

In 2024, the EU is expected to approve a new AI regulation; the AI Act. This means that AI solutions will be categorized into risk levels ranging from low to unacceptable risks, with sanctions of up to 6 percent of revenue. In the high-risk category, organizations are required, among other things, to analyze risk from an integrated technical, legal, and business perspective.
Anna, who welcomes the new law, believes it will lead to people being able to trust new innovations and create long-term business models where customer trust is the focal point."

 

The AI Act will be the first AI regulation in the world, and now the USA and the UK have also opened up to regulation. It ensures the preservation of our democracy, human rights, and integrity," says Anna.

Methodology for AI Risk Control

Ensuring responsible AI and complying with the AI Act requires education, cross-functional governance with clear responsibilities, and reporting by organizations.

- With anch.AI's methodology and platform, organizations gain insight and control over their AI risks from an integrated technical, legal, and business perspective, says Anna.

Start by Embracing Low-Risk AI

So, how should one approach AI, and where does one begin? Anna advocates starting with the low-hanging fruit – areas with low risk of ethical and legal violations that don't necessarily require the most advanced AI technology. The focus should be on thinking business-driven and using AI as a tool to free up time and energy for critical thinking and analysis. For example, to perform repetitive and time-consuming tasks such as collecting data and compiling reports.

anch.AI and Zimply in Partnership

For Zimply, Anna Felländer and anch.AI's unique methodology and platform are valuable tools in a partnership that aids Zimply in integrating AI responsibly into both corporate strategy and AI assistants. This involves integrating the technical, legal, and business perspectives to ensure that AI assistants meet ethical and legal requirements.

- It's not just a technical issue but also about calculating the gains and how AI assistants are designed and built responsibly. At Zimply, we strongly believe in collaborating with forward-thinking companies like anch.AI to build a future where technology is a tool for positive change, says Fred Alpsten, CEO of Zimply.