At the moment, today's AI models are judged to be at level one out of five. But artificial general intelligence could be closer than you think. Credit: Vitor Miranda / Shutterstock OpenAI, the company behind the popular AI chatbot Chat GPT, has now developed an evaluation scale to assess how closely AI models can approach human levels of intelligence, according to a Bloomberg report. The scale has a total of five levels. The higher the level, the closer the AI model is judged to be to human intelligence. Today’s large-scale language models are currently judged to be at level one; that corresponds to basic intelligence, but not a more advanced problem-solving ability. Level two means that the system has a basic problem-solving ability that should be comparable to a human with a PhD. Level three means the system can act as a representative for the user. Level four means that the system can create new innovations. Finally, level five involves the step to achieve artificial general intelligence (AGI), an AI system can perform the work of entire organizations. OpenAI has previously defined AGI as a highly automated system that can outperform humans on the majority of economically valuable tasks. OpenAI’s evaluation scale is considered preliminary and could be adjusted in the future. More OpenAI news: OpenAI is working on new reasoning AI technology OpenAI reportedly stopped staffers from warning about security risks OpenAI whistleblowers seek SEC probe into ‘restrictive’ NDAs with staffers OpenAI: Musk’s control issues at heart of ongoing rift OpenAI models still available in China via Azure cloud despite company ban Related content opinion Agentic RAG AI — more marketing hype than tech advance CIOs are so desperate to stop generative AI hallucinations they’ll believe anything. Unfortunately, Agentic RAG isn’t new and its abilities are exaggerated. By Evan Schuman Aug 16, 2024 5 mins Technology Industry Generative AI Emerging Technology news Researchers tackle AI fact-checking failures with new LLM training technique Deductive Closure Training (DCT) looks to address the problems of LLM bias, misleading information, and outright contradiction. By John E. Dunn Aug 15, 2024 4 mins Generative AI IBM Technology Industry news MIT delivers database containing 700+ risks associated with AI Called the AI Risk Repository, the goal, its creators say, is to provide an accessible and updatable overview of risk landscape. By Paul Barker Aug 15, 2024 1 min Generative AI Security news brief Hollywood unions OK AI-cloned voices in commercials But companies must first obtain consent from the actor for any ad that uses the digital voice copy. By Viktor Eriksson Aug 15, 2024 1 min Generative AI Technology Industry Podcasts Videos Resources Events SUBSCRIBE TO OUR NEWSLETTER From our editors straight to your inbox Get started by entering your email address below. Please enter a valid email address Subscribe