Семейство моделей RIO от West Lake Xinchen — это интеллектуальные ИИ-агенты, способные выходить за рамки простой генерации текста благодаря доступу к поисковым системам. Модели поддерживают двуязычный режим и обладают параметрами до 100 миллиардов, что позволяет им оперировать актуальными внешними данными.
According to him, the RIO series now has more than a dozen members according to the size of the model parameters, ranging from more than 10 billion to more than 100 billion parameters, supports bilingualism in Chinese and English, and in addition to memory, has mastered the ability to call cognitive tools such as search engines, knowledge graphs, and external data. In the middle layer, West Lake Xinchen provides models that go deep into various fields (such as marketing, etc.). “ The large model has greatly improved its ability to process unstructured data. In the past, data that had been deposited within the enterprise but was not fully utilized will be It shows more value under the empowerment of large models, so we started to contact a large number of B-end customers last year We are also constantly communicating with various industries and jointly exploring scene opportunities. We have currently carried out in-depth cooperation with many customers in the marketing and social fields” Lan Zhenzhong said, “There are many flexible cooperation methods with the B-side, and calling API is one of them, the customized model is also a kind." "After testing, the current ability of the RIO model is equivalent to the level of GPT-3.5", Lan Zhenzhong said."