可之科技2017年创办于哈佛大学创新实验室,在麻省理工学院孵化,核心技 术团队集纳一批来自MIT 、布朗大学、卡耐基梅隆大学等全球顶尖学府及曾任职于 Google等知名机构,拥有多年行业经验的人工智能专家。
2021年,核心团队回国发展。可之科技针对以 OpenAI GPT 为代表的通用大 语言模型计算不精确、逻辑不严谨、无法在线迭代的全球性难题,通过独创的技 术路径,融合符号体系和概率体系的优点,研发 Symbolic Pretrained Transformer ( SPT )技术,让模型做到精确、可靠;提出了交互式强化学习 Interactive Reinforcement Learning( IRL)的模型迭代方式,让模型可以在预训练数据有限的 情况下,通过在线学习实现自我进化。基于差异化的SPT技术和IRL技术,解决了大 模型在多产业落地应用的难题。
在教育考试领域,可之SPT大模型自2021年起连续四年服务我国十余省中、高 考,实现了比人更准、更快的AI阅卷服务,保障考试公平性与科学性,并得到了国 际龙头Prometric的认可。
此外,在能源、交通航空、装备制造、安全等不同场景,可之与龙头合作,解 决行业核心痛点,实现大模型产品化应用。
2021年,可之科技从数千家企业中脱颖而出,入选世界经济论坛科技先锋榜 单,成为该年度入选的8家中国企业之一;凭借独特的大模型技术和扎实的产业应 用;2022年获HICOOL全球创新创业大赛最高奖; 2023年获得“直通乌镇”全球互 联网大赛最高奖,大模型及数字人赛道第一名;2024年荣获ITEC全球创业大赛最高 奖;斩获2024年第十三届中国创新创业大赛全国行业赛第二名佳绩;入选2024年世 界互联网大会“互联网之光”博览会十大首发成果及“新光”系列十大“新耀”场 景称号。
Learnable.ai was founded in 2017 at the Harvard University Innovation Lab and later incubated at the MIT. Its core technical team consists of artificial intelligence experts from top global institutions, as well as industry veterans with experience at leading organizations. Learnable.ai addresses major global challenges in LLMs, including inaccurate computation, flawed logic, and the inability to perform real- time online updates—common issues in mainstream models like GPT. Learnable's technologies are being applied in sectors such as education, energy, transportation and aviation, equipment manufacturing, and public security. The company collaborates with industry leaders to tackle core challenges and drive the practical deployment of LLMs across these domains. Learnable's achievements have been widely recognized. In 2021, it was selected as one of the World Economic Forum's Technology Pioneers, one of only eight companies from China to be included that year.
地址:北京市朝阳区将台乡酒仙桥北路乙10号院3号楼星地中心 C座3层307室
34/5000 Room 307, 3rd Floor, Block C, Xingdi Center,Building 3, No.10, North Jiuxianqiao Road, Jiatai Township, Chaoyang District, Beijing
电话:18141792667