Could AI Ever Build Itself? Scientists Map the Road to the Last Human-Built AI

As artificial intelligence development costs skyrocket, researchers are exploring whether AI systems could eventually design and build themselves through recursive self-improvement. This concept, sometimes called the 'last AI built by humans,' could address the economic and technical barriers to creating ever-more-advanced models. Leading AI labs and academic teams are now mapping theoretical pathways and practical milestones toward autonomous AI development. Key challenges include ensuring safety, maintaining human oversight, and overcoming current limitations in AI reasoning and hardware constraints. If achieved, self-improving AI could accelerate innovation but also raises profound questions about control and the future role of human engineers. The road is long, but early research suggests the idea is moving from science fiction toward serious scientific inquiry.