Artificial Intelligence (AI) can proceed at a breaknake speed, but we still have some “miracles” away from receiving AI that can actually rival human intelligence, Nathan Myvold, former Microsoft CTO and current CEO of intellectual enterprises said. Speaking at the Microsoft@50 event in Seattle, Myhrvold underlined why humanity is “three to five miracles away” from making AI that matches the complete cognitive abilities of humans.

Myhrvold, who helped shape Microsoft’s early technology roadmap after joining the company in 1986, said that while AI has made significant progress, yet it lacks a significant ability: the ability to generate completely new abstract concepts, assign the meaning of them, and authentically cause them. “We will eventually reach there,” he said, but insisted that prominent, unexpected scientific successes – “miracles” – are still necessary.

His comments come amid growing industry debate on the boundaries of today’s AI model. Myhrvold reported that AI is at a point of explosive capacity, like the early days of individual computing in the 1980s, but its trajectory is uncertain. “Two years ago, no one would have predicted how far we have come today,” he said, credited Microsoft’s early investment in Openi to put the forefront in this change.

The conversation echoes the recent comments of Mata’s chief AI scientific Yan Lakeun, who suggested in Davos that today’s big language models (LLMs) and generic AI systems are reaching the limits of their utility. Laken has predicted a new wave of AI architecture, within the next five years, with capabilities beyond the current model. They argue that today’s AI systems struggle with memory boundaries, lack of understanding of the physical world, and poor logic and planning abilities-all gaps that accept myhrvold equally.

While companies like Openai and Perplexity are avoiding the new “agent AI” system, which can demonstrate complex, multi-step tasks from booking travel to digital content-Myhrvold and Lecun only see them as the beginning. They believe that the real challenge lies in the creation of AI that can actually innovate, argument about novel concepts that humans naturally do.

Despite the ongoing apprehensions about AI, Myhrvold reduced the diastopian scenarios. He compared AI “Doomsde” to fictional villains such as Solar or Night King – frightening, but eventually unrealistic. “We enjoy the invention of demons, but AI is not yet one,” he said.

Yan Lakeun made a similar statement last year when he said that we reduce AI’s professionals and opposition. They believe that we are often more than how smart AI is. AI does not even have intelligence of our pets, let humans go alone, Lakeun says. On the other hand, he also believes that the opposition associated with AI – mostly the dangers that can come up with its development – are also underestimated. In fact, he said that he is “full BS”.

That said, Myhrvold warned of AI’s collateral challenges, especially around energy consumption. As the CEO of intellectual ventures and vice -president of Terapavar, he emphasized that the next generation nuclear energy would probably be important to meet the Energy demands operated in detail of AI. “Think of a world where the demand for energy increases five to ten times in this century,” he said.