Are SMEs adequately equipped to harness the full potential of tools like ChatGPT? Here are the insights from Sophie Hundertmark, a researcher and author of two books on the topic.
Many large enterprises have already taken steps to incorporate generative artificial intelligence tools into their operations. However, small and medium-sized companies often struggle to find effective ways to capitalize on these new technologies. SMEs frequently lack the know-how to utilize these innovations optimally or assess associated risks. Sophie Hundertmark, a scientific collaborator at Lucerne University of Applied Sciences and Arts and a doctoral candidate at the University of Fribourg, specializes in researching text robots, artificial intelligence (AI), and digital transformation. With two books on the subject under her belt, she sheds light on the challenges posed by these technological advancements for SMEs in Switzerland.
Until recently, AI was primarily used for deploying conversational robots, or 'chatbots'. Generative AI tools such as ChatGPT, launched by OpenAI in 2022, are opening up avenues for numerous additional uses. What features are most valued by Swiss businesses?
Sophie Hundertmark: The most common use remains text generation, particularly for drafting emails and press releases. Generative AI software can also be utilized for data analysis, such as comparing two financial reports or creating a summary from them.
Also worth mentioning is ongoing employee training: with artificial intelligence, it is now possible to create compliance training, for example. AI can be used to enhance existing training software or develop a customized module based on data provided by the programmer.
The automatic generation of images using tools like DALL-E is also gaining momentum, although it only applies to a certain type of businesses.
Are Swiss SMEs adequately prepared to capitalize on the potential of these tools?
Hundertmark: No, I believe they aren't sufficiently familiar with them as yet. Most SMEs not only overlook the opportunities provided by generative artificial intelligence but also the risks it entails. They lack a profound understanding of how to use this type of AI appropriately.
How can they improve?
Hundertmark: It's crucial for Swiss SMEs to undergo training. This not only facilitates knowledge transfer but also fosters the courage to experiment with different integrative AI tools like ChatGPT. It's through hands-on experience that SMEs can learn and improve over time.
Why is training so important?
Hundertmark: Training is a necessary step towards integrating generative artificial intelligence into the operational processes of businesses. At Lucerne University of Applied Sciences and Arts, we've developed a module for this purpose. Various profiles, including business leaders, executives, and sometimes even non-managerial staff, attend our training.
Generative AIs occasionally make mistakes, reproduce biases, or violate copyrights. How can we safeguard against these shortcomings?
Hundertmark: Errors and inaccuracies produced by generative AIs are among the major challenges in AI. However, it's important to acknowledge that humans also make mistakes! This issue isn't new; it's about being aware of it and minimizing risks and implications as much as possible. For example, considering that automatic text generators are used to provide content in various areas of expertise, it's reasonable to expect that employees responsible for this content, who understand the intricacies of their profession, will notice any potential errors. As for biases and copyright risks, they can be mitigated by refining the dataset.
Data protection is another important concern. What are the best practices for addressing it?
Hundertmark: The best solution would be to engage a service provider based in Switzerland. The new data protection law guarantees a higher level of confidentiality compared to overseas, as data will be processed and hosted within Swiss territory. Therefore, I would recommend using tools such as SwissGPT (launched in August 2023 by the Zurich-based company AlpineAI).
