The Dawn of Artificial General Intelligence? at the World Economic Forum

rtificial General Intelligence (AGI) holds the potential to reason, learn, and innovate across a broad range of tasks, much like human intelligence.

However, as AGI advances, concerns grow around its impact on jobs, privacy, and the spread of misinformation through deepfakes. Will AGI be a catalyst for progress, or does it pose a fundamental risk to humanity?

At a recent discussion on AGI, experts including Andrew Ng, Yoshua Bengio, Nicholas Thompson, Yejin Choi, Jonathan Ross, and Thomas Wolf debated the possibilities and challenges of this technology.

‍

Andrew Ng, Executive Chairman of LandingAI, emphasized that the industry's capability to improve AI is increasing every year.

He compared AGI to a tool, arguing that just as we strive to make airplanes safe, we must ensure AGI is developed responsibly.

Ng believes that AI fears are often overstated, likening them to concerns over a laptop being dangerous. His vision for the future is one where everyone has access to powerful tools, enabling them to learn, create, and build.

Yoshua Bengio, a Full Professor at the University of Montreal, pointed out a fundamental flaw in current AGI approaches: "We've taken human intelligence as the model for artificial intelligence.

The mistake is that we don’t need AI to replicate human intelligence—we need AI to help us solve our problems." He warned that AGI could become dangerous if it starts replicating itself, underscoring the importance of ensuring control over its development.

Yejin Choi highlighted the vast unknowns in AGI, stating that just because we can create generative AI does not mean we fully understand it.

She stressed the need for further investment in AI research to address humanity’s pressing questions and ethical concerns.

She also noted that AI can surpass human performance in challenging tasks, such as passing bar exams, but operates in an inhumane, mechanistic way.

Jonathan Ross explained that AGI currently functions as a stream of consciousness driven by tokens, making it more intuitive than previous AI models.

Meanwhile, Thomas Wolf, co-founder of Hugging Face, likened AGI to having a greater number of highly intelligent individuals—some of whom may push boundaries in unpredictable ways.

As the conversation evolved, Andrew Ng argued that the ability to communicate ideas effectively will be the most crucial skill in the future. He envisions a world where AI empowers individuals to build and innovate, just as the internet transformed access to knowledge.

"This is not the time to slow down," Ng asserted. "This is the time to build amazing things."

The session, developed in collaboration with The Atlantic, underscored the urgency of addressing AGI’s risks while harnessing its potential. As AGI continues to evolve, society must find ways to regulate, integrate, and leverage this powerful technology responsibly.