AI giants' "AGI predictions" have suddenly increased. Is this just hype?

Recently, a striking phenomenon has emerged in the field of artificial intelligence: several renowned researchers from major AI companies have predicted that superintelligent AI systems will appear in the near future.

These views suggest that AGI will be achieved within the next 1-2 years and will profoundly change society. This radical prediction has raised questions: Is this just hype?

In response, Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania, recently pointed out that it is necessary to remain cautious about the statements of these industry insiders. They may make bold predictions for motives such as raising funds or inflating stock valuations. Additionally, current AI technology itself has limitations. Even if AGI is realized, the speed at which human society adopts and adapts to this technology may be overestimated.

Nevertheless, Mollick also noted that some recent advancements in the AI field are indeed remarkable. There has been significant progress in narrow intelligence and modal capabilities. What we should be most concerned about is not whether the timelines of various AI giants are correct, but whether we are prepared for the impending transformation.

Mollick: Stay Cautious About Bold Predictions for AGI

Some well-known researchers from major AI companies have recently made statements predicting the imminent emergence of superintelligent AI.

Professor Mollick expressed caution, believing that these bold predictions may be motivated by fundraising and stock valuation inflation. He pointed out that technologists are not prophets, and many technological predictions in history have proven to be overly optimistic.

Moreover, current AI technology itself has limitations and remains unstable. This "uneven capability" is a core characteristic of current AI systems, which is difficult to eliminate:

“Although today’s large language models are impressive, they are essentially unstable tools—performing excellently on certain tasks but making mistakes on seemingly simpler ones.”

Mollick believes that even if we assume researchers are correct about achieving AGI within the next year or two, they may overestimate the speed at which humans will adopt and adapt to this technology.

“Organizational change takes a long time, and changes in work, life, and educational systems are even slower.”

The Future Tide: Dual Considerations of Technology and Society

Nevertheless, some recent advancements in the AI field are indeed remarkable. OpenAI's o3 model has defeated human experts in several tests for the first time, including the graduate-level question-and-answer test (GPQA) that Google could not solve, cutting-edge mathematical problems, and the ARC-AGI fluid intelligence test.

As AI capabilities improve, the practicality of AI agents in specialized or niche fields is becoming increasingly evident. Mollick mentioned Google's Gemini and Deep Research project, which can write a 17-page research report in just a few minutes. After personally testing it, Professor Mollick believes that while it is not as good as the best humans, it is better than many reports written by humans

"The AI agents in specialized or niche fields are truly products, not future possibilities. There are already many coding agents, and you can use experimental open-source agents for scientific and financial research."

In addition to the rise of agents, the progress of AI in multimodal capabilities is also remarkable. With technological advancements, AI's contextual memory and image generation capabilities have significantly improved.

Mollick cites examples where multiple image creation tools can accurately generate complex scenes like "a otter using wifi on an airplane," which was difficult to achieve not long ago. The advancements in video generation are equally significant.

The transformation has arrived, are we ready?

In light of all this, we can't help but wonder: should we take seriously the claims from major AI companies about the impending flood of intelligence?

Mollick believes we must take the current technological advancements seriously. What we should be most concerned about is not whether the timelines of various AI giants are correct, but whether we are prepared for the upcoming transformation.

"This is not just about the technology itself; it’s about how we choose to shape and deploy it. These questions are not ones that AI developers can or should answer alone."

He calls for not only AI developers but also organizational leaders, employees, and other stakeholders to start thinking about how to shape and deploy AI technology to ensure it becomes a force for progress rather than destruction.

"The time to start these discussions is not after the water level begins to rise—but now."