According to Mustafa Suleyman, a co-founder of Google’s AI research lab DeepMind, AI chatbots like ChatGPT should undergo a “modern Turing test” to assess their human-like intelligence.
In this proposed test, their capability to transform $100,000 into $1 million would serve as a benchmark.
Suleyman, who previously held the position of Head of Applied AI at DeepMind and now serves as the CEO and co-founder of Inflection AI, is set to publish a book titled “The Coming Wave: Technology, Power, and the Twenty-first Century’s Greatest Dilemma.” The book delves into the complexities and challenges posed by technology in the modern era.
According to a report from Bloomberg, Mustafa Suleyman challenges the significance of the traditional Turing test in his book. Suleyman expresses uncertainty regarding whether this test truly represents a meaningful milestone in assessing machine intelligence.
Introduced by Alan Turing in the 1950s, the Turing test aims to determine if a machine possesses human-level intelligence. Human evaluators engage in conversations and attempt to discern whether they are interacting with a human or a machine. If the machine successfully convinces the evaluators that it is human, it is deemed to have passed the test.
Suleyman suggests a shift from comparing AI’s intelligence to that of humans and instead proposes a concept called “artificial capable intelligence” (ACI). ACI involves assigning AI bots with specific short-term goals and tasks that they can accomplish with minimal human intervention.
To demonstrate ACI, Suleyman proposes a new Turing test where an AI bot is provided with a $100,000 seed investment and challenged to generate $1 million in return.
The bot would need to conduct research on a potential e-commerce business idea, formulate a product plan, locate a manufacturer, and successfully sell the product.
Suleyman expresses optimism that AI will achieve this milestone within the next two years.