Experts in the field of Artificial Intelligence (AI) are asking European Union (E.U.) leaders to monitor the use of tools like ChatGPT. The E.U. had just passed an AI Act that was proposed with regards to safety and transparency for certain “high-risk” AI tools, such as those used in education or law enforcement. However, the version of the draft that was approved, though wider in scope, does not cover so-called “general purpose” AI, such as OpenAI’s popular chatbot.
Now, with many companies investing in this technology to be used in everyday products, a group of prominent AI scholars have asked the E.U. to consider tools such as ChatGPT to be placed under the “high-risk” category. This is because other AI models may be more dangerous if left unregulated.
The request was signed by prominent figures that include former Google AI ethicist, Timnit Gebru and Mozilla Foundation President, Mark Surman. It calls for a “expansive” approach to what AI tools the E.U. will cover in their proposed rules. The warning made by the scholars is that AI models such as ChatGPT, DALL-E 2, and Bard are only the tip of the iceberg.
The AI Now Institute has also been vocal about this topic, particularly Amba Kak and Sarah Myers West, both former advisors to Federal Trade Commission Chair Lina Khan who had written a report in 2021, calling for greater scrutiny of AI harms. The E.U. may be the first to come up with an omnibus framework for AI, setting a global precedent.
Any regulations, the scholars wrote, must include regulating AI tools throughout the product cycle. This means holding AI developers accountable for the data choices they make. Additionally, they asked the European leaders to do away with legal disclaimers that may be used to dodge the rules applied to developers.
Alex Hanna, the director of the Distributed AI Research Institute, also expressed her thoughts. She said that not placing tools such as ChatGPT, Microsoft’s Bard, and others like them in the “high-risk” category would show that the E.U. is not taking the risks these models pose seriously.
Two of the lead European lawmakers for the AI Act proposed in February that AI systems generating complex text without human oversight should be included in the “high-risk” list, in an attempt to stop ChatGPT from spreading disinformation. However, according to reports, the suggestion was met with skepticism by right-leaning political groups in the E.U., creating a potential stumbling block for its adoption.
Though the United States is only starting to look into AI-specific regulations, E.U. Commissioner for Internal Market Thierry Breton told Reuters that their goal is to address the issues presented by chatbots such as ChatGPT. He stated that while they offer many opportunities, it is important to have a “solid regulatory framework to ensure trustworthy AI based on high-quality data.”
OpenAI is an American company founded in 2015 that researches and develops Artificial Intelligence, specifically reinforcement learning and unsupervised machine learning algorithms. They have developed their own chatbot, GPT-3, which has been used in various applications such as natural language processing and text-to-speech. Microsoft’s AI-based chatbot, Bard, is another popular example, having been used by brands for conversational advertising and marketing.
Timnit Gebru is a former Google AI ethicist and research scientist, who has done extensive research in the field of Artificial Intelligence, particularly in phenomena such as algorithmic bias, computer vision and machine learning. Mark Surman is the head of the Mozilla Foundation and has a background in software engineering, web development and digital activism.
