Tenzai co-founder and CEO Ariel Zeitlin said he rejects recursive self-improvement, calling a rapid takeoff in which one company’s model quickly iterates and leaves all competitors behind “basically impossible.” According to Sina Finance, he expects AI development to slow rather than accelerate, partly because AI capabilities under development are approaching the level of human experts, making it increasingly difficult for them to evaluate AI outputs. Zeitlin also said each dollar invested will yield progressively smaller improvements in AI capabilities, which is one reason companies such as Anthropic and OpenAI need huge amounts of funding. If a company spends ten times as much to achieve only a doubling in model capabilities, measuring technical progress will become difficult, he said. He added that breakthroughs will become increasingly difficult and that this trend will continue.