A new Turing Test has been proposed by a leading AI expert, which involves creating an online store to determine whether an AI is truly intelligent. The Turing Test was first proposed by British mathematician and computer pioneer Alan Turing in 1950, and aims to evaluate a machine’s ability to demonstrate human-like intelligence. In the original test, a human evaluator engages in text-based conversations with an unknown entity, either a machine or a human, to determine which is which. If the evaluator cannot distinguish between the two, the machine is considered intelligent. However, modern AI systems such as ChatGPT, which are based on Large Language Models, can now pass the Turing Test regularly, rendering it outdated.
Mustafa Suleyman, co-founder of Google-owned AI research lab DeepMind, has proposed a new Turing Test in which an AI is given $100,000 to create an e-commerce business and generate $1 million in revenue. The AI would need to create a business plan, find suitable products, and sell them online. Suleyman believes that the current focus of the AI industry on developing a general artificial intelligence (AGI) that possesses or even surpasses human cognitive abilities is misguided. Instead, he suggests that the goal should be to develop a powerful artificial intelligence (ACI) that can autonomously perform complex tasks. He believes that this is a more realistic goal to achieve within the next two years.
While modern AI systems can pass the Turing Test, Suleyman argues that it cannot determine whether an AI truly understands abstract concepts and can handle long-term planning. By creating an online store, the new Turing Test would provide a more practical evaluation of an AI’s capabilities. Suleyman’s proposal highlights the need to shift the focus of AI development towards creating systems that can perform specific tasks autonomously, rather than attempting to replicate human intelligence.
