When most people think of the laws of AI, science fiction writer Isaac Asimov’s laws of robotics come to mind. He created these laws within his fictional world to create his imagined distant future where robots would be everywhere and what it would be like. More than 80 years later, we don’t have the intelligent robots he used in his stories yet, but we are getting closer each day.
Asimov’s laws started as science fiction, but now we need to set up some rules with the ever-present dominance of Artificial Intelligence in our lives. Not addressing AI while acting like it doesn’t affect our lives will only lead to problems. We can even see the first protests in the form of the recent Writers Guild of America strike. One of their main points is not allowing AI-written scripts. Not too long ago in Italy, the dubbing industry went on strike to make sure AI doesn’t replace their voices.
AI technology is years away from posing an actual threat to screen writers’ jobs. But the lack of legal bargaining requirements explains why the WGA is adamant about adding limits to studios’ use of AI in its next agreement. pic.twitter.com/koaG7tXFTB
— Bloomberg Law (@BLaw) May 8, 2023
Individual companies, like Google, set their own internal principles to address this issue. But no official law was ever written, until today. The European Union just drafted the AI Act which, once approved, will be the world’s first set of rules on AI.
What the draft entails
As a premise, most of what’s written in the AI Act will seem like logical and straightforward rules. However, these things are still not regulated today; even if it turns out to be immoral, they will still be legal. With the AI Act, all intrusive and potentially discriminatory uses of AI will be banned. The European Parliament press release gives a few examples of this, one being biometric identification. This type of identification based on biological features (such as facial recognition), may be banned from being used remotely in public spaces for real-time identification. It may also be banned for post-identification unless judicial authorization is given to prosecute a serious crime committed.
You might have heard that some countries restricted access to ChatGPT over privacy concerns. The press release directly calls out GPT and how models like it will be needing to pass additional transparency measures. The exception to this rule is open-source models, which GPT isn’t. The definition of “High-risk AI” was also expanded to include all models that could potentially cause harm to people’s health, safety, and fundamental rights. This also considers the possibility of influencing people during political campaigns.
The world is catching up to the future Asimov foresaw in his books. The AI Act is the first step that will allow us to be prepared when we reach that point. It’s limited and doesn’t directly address many issues like the replacement of writers and voice actors people have already protested about. But it’s a step, and every journey begins with a single step.
Photo credits: The feature image is from a previous event on the same subject and has been taken by Alexis Haulot. (© European Union 2022 – Source: EP)
Source: Alissa Wilkinson (Vox) / European Parliament / Nick Vivarelli (Variety)