Written by 8:35 pm AI, Latest news, Uncategorized

### AI IQ Test Reveals Surprising Results: Researchers Discover Artificial Intelligence’s Lackluster Performance

Artificial intelligence programs still struggle with basic problem-solving skills that people excel…

There has been significant discussion surrounding Artificial General Intelligence (AGI) recently, which represents a coveted milestone in AI development. AGI envisions a future where AI algorithms can perform the majority of tasks currently carried out by humans. Proponents of this concept suggest that the advent of AGI could lead to profound societal transformations, potentially paving the way for a “post-work” era where humans can relax while robots handle the heavy workload. Speculations suggest that OpenAI’s recent internal turmoil might have been influenced by advancements in AGI, particularly the “Q” program, as insiders claim it triggered a significant power struggle.

Contrary to these notions, recent insights from Yann LeCun, Meta’s leading AI scientist, challenge the idea that artificial intelligence will achieve general-purpose functionality anytime soon. LeCun’s latest publication argues that AI still lags behind humans in crucial cognitive aspects.

In a collaborative effort involving researchers from various AI enterprises such as Hugging Face and AutoGPT, a study was conducted to evaluate AI’s general reasoning capabilities compared to human cognition. The research introduced a set of questions designed to be straightforward for humans yet challenging for advanced AI systems. These inquiries were presented to both humans and a plugin-enhanced version of GPT-4, OpenAI’s latest large language model. The assessment aimed to gauge AI’s proficiency in handling real-world questions that necessitate fundamental skills like reasoning, multi-modality comprehension, web navigation, and overall tool utilization.

The outcomes revealed that despite excelling in tasks that pose difficulties for humans, large language models struggled with the complex problem-solving scenarios presented in the study. Even with additional tools, GPT-4 achieved only modest success rates, significantly lower than human participants across various tasks.

LeCun’s stance diverges from the optimistic outlook of some AI experts who anticipate the imminent emergence of AGI. He emphasizes the importance of AI systems possessing internal models of the world to predict outcomes, enabling reasoning and planning capabilities. LeCun critiqued current Auto-Regressive Large Language Models (LLMs) for lacking this essential ability, asserting that they fall short of human-level intelligence due to their limited understanding of the physical world and absence of planning skills.

Visited 1 times, 1 visit(s) today
Last modified: February 9, 2024
Close Search Window