Responsible AI

Responsible AI is one of the most important issues for the future of humanity. I’ve often heard variations of the message “We can do good while doing well with AI”, at every AI conference I’ve attended this year. But what exactly is Responsible AI?

The development of the field of Artificial Intelligence dates back to the 1950s. To ensure AI safety, three main rules were initially proposed:

1. Don’t connect it to the internet.

2. Don’t allow direct human interaction.

3. Don’t automate the training process.

These guidelines aimed to control AI system’s influence and avoid unintended consequences. All three rules have been broken over the past two years, accelerated by the release of ChatGPT to the public. So, how can we ensure responsible AI now?

In the European Union, the approach is straightforward: they rolled out the EU AI Act which, though not perfect, provides a solid starting point. In the USA, however, there is no federal regulation yet, so it is up to each of us to ensure the ethical development, implementation and use of artificial intelligence.

With the rapid pace at which AI is evolving and its broad scope, attempting to “regulate AI” effectively presents significant challenges. A more effective approach might be to regulate the application of AI technologies within specific industries or sectors.

A good example of this approach is California’s passage of a bill to combat deepfake election content. From a company standpoint, Stability AI has also taken the lead by designing its text-to-image tools to exclude celebrities and political figures from any inferences, helping to prevent deepfakes. If you try to force a prompt that includes these figures, the tool generates a caricature instead.

When it comes to transparency and explainability, many AI companies adopt an open-source approach, making their models, data and/or research publicly accessible for anyone to use, modify and distribute. Hugging Face serves as a valuable hub for open-source tools and resources that support AI development.

These examples show that we don’t need to wait for global consensus. Responsible AI starts with each of us, one good decision at a time. The promise of AI is "making the world a better place"… let’s all contribute to ensuring it is a safe one too! I look forward to receiving your comments… #BetterAI for a #BetterWorld #EverydAI VxM AI