复旦大学计算机科学技术学院教授、上海市数据科学重点实验室主任肖仰华推测，ChatGPT 或许使用了大小模型结合的技术，多模型协同可以很大程度上解决成本问题。“让大模型去解决复杂问题，但普通用户一般都是简单的问题，用小模型就可以了。” 他形容，ChatGPT 的 AI 技术已形成一种 “飞轮效应”，迭代和优化正推动技术进入一个自我强化的快速发展阶段行业或将变成 “只有第一，没有第二” 的局面。“我们的文化中，聪明人太多，总想抄近路。有时候就是要坚持，要笨一点，才能做出真正的大杀器。我们与 OpenAI 的差距正在拉大，不是在缩小。” 肖教授指出，还有很多如机器人、环境交互技术、医疗大模型、科学大模型等其他方向，都是大趋势。企业可以在这些新领域找到自己的优势。肖仰华指出，AI 技术的进步可能导致人类智力的相对退化。
Recently, Professor Xiaoyanghua, from the School of Computer Science and Technology at Fudan University and Director of the Shanghai Key Laboratory of Data Science, speculated that ChatGPT may have used a combination of large and small models to address cost issues. “Let the large model solve complex problems, but ordinary users usually have simple questions that can be answered with a small model,” he described. The AI technology of ChatGPT has formed a “flywheel effect”, where iteration and optimization are driving the technology into a self-reinforcing rapid development stage. The industry may become a situation where there is only one leader without any competitors. “In our culture, there are too many smart people who always want shortcuts. Sometimes we need to persist and be a little foolish in order to create truly powerful tools. The gap between us and OpenAI is widening rather than narrowing,” Professor Xiao pointed out. He also mentioned that there are other emerging trends such as robotics, interactive environmental technologies, large-scale medical models, scientific models, etc., where companies can find their own advantages. Professor Xiaoyanghua indicated that advancements in AI technology could potentially lead to relative decline in human intelligence.
微信扫一扫