As negotiations on the AI Act in the European Parliament move closer to an agreement, one issue dominates the news for weeks in a row: Generative AI or powerful artificial intelligence. Call it a hype or a trend, in any case it’s fascinating users, non-users and legislators. ChatGPT, a generative AI system capable of generating complex text, is the fastest-growing consumer application in history, according to a UBS study. All of this public attention and the ongoing political debate have now culminated in the publication of a letter co-signed by many key-MEPs on the AI Act across the political spectrum calling for significant political attention and an ever-increasing sense of responsibility from developers working on AI. While these discussions are extremely important in the face of the fast-paced developments in this field, the growing concerns linked to the technology threaten to overshadow the current debate on regulating AI. Let’s not confuse things…
General purpose AI > Generative AI > ChatGPT
Fundamentally, General Purpose AI (GPAI) systems are AI systems that are use-case independent. They can be used in a wide variety of situations performing a broad range of tasks and have more than one application. One subset of GPAI systems are systems which can create, generate, or develop images, text, codes and so on. These systems are referred to as Generative AI (GAI). It is crucial to differentiate between GPAI and GAI systems due to significant differences in capabilities and possible risk-levels of their application.
Some important conclusions to take with us in the debate:
- General Purpose AI is not the same as Generative AI, and certainly not the same as ChatGPT. It’s much more than that.
- Any risks related to ChatGPT, and other generative AI do not necessarily apply to other GPAI system
- Many General Purpose AI applications provide a lot of value at a very low risk
- Generative AI applications also provide a lot of value, but may come with a higher risk
It’s time to dispel the noise: not all GPAI is high risk
We all know Italy temporarily banned ChatGPT on the basis of concerns about the system’s compliance with GDPR. The Spanish Data Protection Agency (AEPD) announced it also opened, on its own initiative, an investigation into the American company OpenAI, the creator of the conversational chatbot, for a possible breach of the regulations governing data protection. The European Data Protection Board (EDPB) plans to launch a dedicated task force to investigate ChatGPT and whether the technology is compliant with the EU’s General Data Protection Regulation (GDPR).
The problem we’re facing now is that the hype, the confusion and the lack of understanding is leading to a situation where the EU is asked to create restrictive legislation for this upcoming technology due to fear over a small subset of applications. Adding General Purpose AI altogether to the category of “high risk” AI systems, would be disproportionate and would hinder innovation in Europe. Commissioner Vestager underlined this on twitter, by stating “No matter which tech we use, we have to continue to advance our freedoms and protect our rights, That’s why we don’t regulate AI technologies, we regulate the uses of AI. Let’s not throw away in a few years what has taken decades to build”.
- The AI Act should recognise the special role of generative AI systems compared to the rest of GPAI systems and should not create blanket requirements for the technology which would be largely disproportionate.
- The AI Act should serve as a horizontal piece of legislation, complementing the EU’s existing legislative framework on digital. As such, any sectoral questions relating to AI, GPAI or generative AI should be addressed through the appropriate existing legislations, such as GDPR in the case of data protection concerns, or through new vertical legislation.
We hope the ongoing discussions on the AI Act will take a step back from the current media-focus on ChatGPT. The end goal of the EU should be to find a clear, balanced and constructive approach to make sure the AI Act can realise its full potential and can contribute to the EU’s overarching framework of digital legislation.
The following quote can be attributed to Siada El Ramly, Director General DOT Europe
“Acknowledging the need to safeguard the health, safety and fundamental rights of people, we urge MEPs to recognize the distinction between GPAI and foundational models. We fear new rules applying to all GPAI would be disproportionate and hinder innovation in Europe as not all GPAI is high risk.”
For more information from DOT Europe contact:
+32(0) 472 26 83 02
Notes to the editor:
DOT Europe (previously called EDiMA) is the voice of the leading internet companies in Europe. DOT Europe’s mission is to develop ideas and support policy initiatives that foster an innovative, open and safe internet for Europe’s citizens and businesses. More information is available here: https://doteurope.eu/
DOT Europe represents 23 of the leading internet companies: Airbnb, Amazon, Apple, Discord, Dropbox, eBay, Etsy, Expedia Group, Google, Indeed, King, Meta, Microsoft, Mozilla, Nextdoor, OLX, Shopify, Snap, Spotify, TikTok, Twitter, Yahoo, Yelp.
Its members produce and manage a variety of products, services and applications including browsers, entertainment platforms, social networks, marketplaces and review sites. More information is available here: https://doteurope.eu/members