Co-founder of DeepMind proposes "New Turing Test" based on AI's ability to generate wealth

2023-06-21

With the hype around artificial intelligence (AI) leading venture capital firms to eagerly invest in any new company that associates "A" and "I" with its name, there is a need for a new definition of AI. In the world of large-scale language models today, the Turing test fails to define true intelligence. A prominent figure in the field of AI has proposed a better idea, one that can only come from the pursuit of wealth in the tech industry: the best way to determine if AI is intelligent is by its ability to quickly accumulate wealth. Mustafa Suleyman, a leading AI developer who was previously a co-founder of DeepMind and is now owned by Google, is reportedly publishing a book discussing the intelligence of AI based on its ability to accumulate cash. According to a report by Bloomberg, in his upcoming book "The Coming Wave: Technology, Power, and the Greatest Dilemma of the 21st Century," Suleyman argues that AI research needs to focus on short-term development rather than the unrealistic concept of artificial general intelligence (AGI). What is a good example? His "modern Turing test" would give AI $100,000, and researchers would wait for the initial investment of AI to reach $1 million. You see, only true intelligence can "make the team rise," and those who are good at making the team rise are obviously the smartest. This proposal sounds more like a get-rich-quick scheme based on AI, which you can find on the lower end of YouTube and TikTok financial influencers. AI certainly has the ability to design products more effectively than having genuine imagination, but modern systems are still far from giving Suleyman a substantial bank account. Earlier this year, several AI agents appeared in the spotlight of AI. They essentially act as multiple versions of chatbots working together to accomplish specific tasks. Users have even created agents that can perform tasks such as ordering pizza over the phone. In theory, programs like Auto-GPT can develop business plans for new products, but current large-scale language models definitely lack the ability to conceive and accurately implement truly desired products. AI is essentially iterative and does not truly understand the content generated in any real-world context. Its output is based on its training data, so any "novel" innovation it creates is based on existing examples that best fulfill the original prompts. AI evangelists often argue that everything created by humans is essentially iterative, but unlike AI, human creations come from an understanding of basic human needs, or at least simple empathy. Of course, AI can create a product that may even sell well. But will it be a good product? Will it truly help people? Does it really matter if the ultimate goal is "making me a million dollars"? Do we need empathy in AI? Suleyman, who previously led AI development at DeepMind but is now the CEO and founder of startup Inflection AI, has a highly capable AI chatbot called Pi that attempts to demonstrate empathy. It is also more restrictive compared to other chatbots like ChatGPT. It cannot research information on the internet or generate code, and always emphasizes that it cannot replace a real therapist or human companion. In our last conversation, Suleyman stated that the narrow definition of AI capabilities by the company's AI is the chatbot's biggest selling point. He firmly stated that the development team does not want to create a product with harmful "maximized engagement" for users. To give a charitable interpretation, this DeepMind alumnus may have used this test as an example, but if that's the case, why is making money his top priority? From Google to Microsoft to large tech companies like OpenAI, they are all competing to sell AI products to users, without truly taking the time to consider the risks or potential impacts of the systems. Requests from company executives to slow down development from the broader research community have been scoffed at because, so far, they have proven that their primary task is to profit for shareholders rather than the entire society. In the conversation, Suleyman stated that those who lose their jobs due to AI need "compensation" from governments around the world. If that's the case, why can't we get some consideration from tech companies that are skyrocketing these AI models? In an era where chatbots have become so sophisticated that they can mimic human voices, AI can easily appear intelligent but fails to understand any of the content it speaks. Despite some doomsayers and claims from figures like Sam Altman and Greg Brockman, leaders of OpenAI, that we are far from achieving AGI levels. AI has made significant progress, but in many ways, it still falls short of the capabilities envisioned by early computer scientists. Natural language has long been the goal of these "Turing-like computers." But whether computers can achieve true "intelligence" as understood by humans remains a subject of public debate. On the other hand, former Google employee Blake Lemoine claims that the company's chatbot model is actually intelligent because it sounds human. Defining intelligence as a binary ability to earn money is more helpful than simply imagining it as a friend just because AI says it is. Source: https://gizmodo.com/deepmind-suleyman-new-turing-test-make-money-1850557322

RECENT AI NEWS

RECENT AI TOOLS