In recent developments, OpenAI has unveiled a new experimental large language model (LLM) designed to enhance understanding of artificial intelligence mechanisms. This innovative model represents a significant leap forward in AI transparency, addressing a common concern in the field: the “black box” nature of current LLMs. Researchers have long grappled with the unpredictability of these models, which can produce unexpected outputs or “hallucinations.” By shedding light on how LLMs operate, OpenAI’s new model could help the AI community build more reliable systems and better assess their suitability for critical applications.
Meanwhile, Google DeepMind is making strides in the realm of robotics and gaming with its new agent, SIMA 2. This advanced agent operates within 3D environments, showcasing capabilities that move it closer to becoming a general-purpose AI. Built on the foundation of the Gemini LLM, SIMA 2 demonstrates enhanced problem-solving skills and adaptability, which are essential for real-world robotics applications.
On a different front, the UK government has announced an ambitious initiative to phase out animal testing in scientific research. By the end of next year, testing for skin irritants on animals will cease. Furthermore, by 2027, researchers are expected to eliminate the use of mice in Botox strength tests, and significant reductions in testing on dogs and primates are anticipated by 2030. This shift is supported by recent technological advancements that offer alternative methods for evaluating the safety and efficacy of medical treatments without reliance on animal subjects.
Source: The Download: how AI really works, and phasing out animal testing via MIT Technology Review