Swiss professionals foresee How do we navigate risks to technologies – and GenAI in particular – “Bias in AI comes from algorithms clear opportunities and harness opportunities? Switzerland must occupy a “sweet spot” producing results that may have a 2 bene昀椀ts, but remain wary of that safeguards innovation without sti昀氀ing prejudice against a given population and risks to privacy, over-reliance, Our AI framework workshop also it. (We will look at the topic of regulation in are the result of a certain polarization and data manipulation. highlighted risks around privacy and greater detail in Section 4 of this report.) of content creation. By bringing human critical thinking and judgment, we can Our survey reveals that respondents data protection, 昀氀agging a lack of broad Our AI framework workshop also train the AI on its biases and allow it to see customer service, healthcare, and and coherent standards within the looked in detail at the issue of bias and improve. Doing so will allow AI to correct enhanced decision-making as the greatest Swiss Federal Act of Data Protection hallucination in GenAI in particular: a natural bias in humans, enhance potential bene昀椀ciaries of generative AI and warning that the deletion of data the risks of producing discriminatory, processes, and, eventually, decision- adoption. A high proportion anticipates as a function of data privacy can lead to culturally inappropriate, or inaccurate making.” opportunities in the use of chatbots and breaks in the learning effect in GenAI. Our outputs because of biased input or 昀氀awed content creation as well as bene昀椀ts experts also stress the risks of unchecked pattern detection by GenAI. Here, our Anthony Corbaz, digitalswitzerland accruing from precision medicine, medical adoption, particularly among younger, experts highlighted the “dual relationship” diagnostics, and robotics. Conversely, “digitally native” generations, which is fast between AI and human-generated input many respondents are concerned about outpacing regulation in Switzerland. And data. While bias and discrimination in AI- “AI and new technologies are not a risks which include, but are not limited that’s not all. generated output is a critical concern, to, threats to privacy and cybersecurity, they stress the potential for “neutral” 昀椀nality or a fatality; people need to overdependence on the technology Swiss innovation and competitiveness GenAI itself to become a corrective tool manage the technology and make the itself, and manipulation of algorithms. in GenAI are at risk of stalling due to that can uncover, adjust for, and rectify best use of it. For continuing education While these concerns are common to a lack of data. Because Switzerland is human biases. instructors, this means acquiring new employees and entrepreneurs alike, the a relatively small country, producing skills and the opportunity to add more relatively small amounts of training data, value.” survey shows that organizations are more Understanding the risks, bene昀椀ts, and focused on the potential loss of control our reliance on foreign GenAI platforms is potential of GenAI to rectify its own Ofra Hazanov, EPFL and intellectual property infringement, relatively high. Regulation is needed, say shortcomings is perhaps contingent above while entrepreneurs cite transparency, our experts, along with incentive systems all on remaining educated and vigilant accountability, and ethical risks as key for data sharing, so that global viewpoints about its evolving capabilities. Our experts challenges. and culture differences like those of stress the need for organizations and Switzerland are fully and appropriately educators to be proactive in developing represented in AI. What could this their understanding and skillset in regulation look like? One possibility could tandem with innovation and to exercise be to forge a Swiss Data Act. Another critical thinking and common sense as we could be to align Swiss regulatory navigate risks today – and tomorrow. measures to the EU AI Act, bringing Swiss in昀氀uence to bear on US in昀氀uence. Above all, say our experts, regulating transversal
Charting the future: Switzerland's path to generative AI leadership in 2024 Page 9 Page 11