过去几年，AI领域仿佛被一条名为“Scaling Law”的法则所统治。人们坚信，只要模型足够大、数据足够多、算力足够强，AI的性能就能一路攀升，无所不能。OpenAI的GPT系列、谷歌的PaLM等模型的成功，似乎完美印证了这一点。

很多人认为，Scaling Law 正在面临收益递减，因此继续扩大计算规模训练模型的做法正在被质疑。最近的观察给出了不一样的结论。研究发现，哪怕模型在「单步任务」上的准确率提升越来越慢，这些小小的进步叠加起来，也能让模型完成的任务长度实现「指数级增长」，而这一点可能在现实中更有经济价值。

BEIJING, Sept. 12 (Xinhua) -- A new law was adopted by Chinese lawmakers on Friday to regulate the response to public health emergencies and enhance the country's capacity to address such situations.

Step by step, China has demonstrated its commitment to delivering on its word and taking action, adding fresh impetus to the ...

Wang and Tonga's Minister for Police Paula Piukala co-chaired the ministerial dialogue. The heads of delegations from Tonga, Fiji, Solomon Islands, Kiribati, Vanuatu, Samoa, Nauru and Papua New Guinea ...

BEIJING, Sept. 8 (Xinhua) -- Chinese lawmakers have begun reviewing a draft revision to the Enterprise Bankruptcy Law, as part of efforts to improve the market exit system. The draft revision to the ...