开源人工智能

开源人工智能（英語：Open-source artificial intelligence）是将开源实践应用于人工智能资源开发的做法。许多开源人工智能产品都是基于大公司共享的开源软件,并对现有工具和技术进行改进。^[1] 流行的开源人工智能项目类别包括大型语言模型、机器翻译工具和聊天机器人。^[2] 对于开发开源人工智能资源的软件开发人员来说，他们必须信任他们在开发中使用的各种其他开源软件组件。^[3]

大型语言模型[编辑]

开源大型语言基础模型比较
模型	开发机构	参数计数	上下文窗口	许可
LLaMA^[4]	Meta AI	7B, 13B, 33B, 65B	2048	——
LLaMA 2^[5]^[6]	Meta AI	7B, 13B, 70B	4k	Custom Meta License
Mistral 7B^[7]	Mistral AI	7 billion	8k^[8]	Apache 2.0
GPT-J^[9]	EleutherAI	6 billion	2048	Apache 2.0
Pythia^[10]	EluetherAI	70 million - 12 billion	——	Apache 2.0 (Pythia-6.9B only)^[11]

参见[编辑]

參考資料[编辑]

^ Heaven, Will Douglas. The open-source AI boom is built on Big Tech's handouts. How long will it last?. MIT Technology Review. May 12, 2023 [2023-12-15]. （原始内容存档于2024-01-05）（英语）.
^ Castelvecchi, Davide. Open-source AI chatbots are booming — what does this mean for researchers?. Nature. 29 June 2023, 618 (7967): 891–892. doi:10.1038/d41586-023-01970-6.
^ Thummadi, Babu Veeresh. Artificial Intelligence (AI) Capabilities, Trust and Open Source Software Team Performance. Lecture Notes in Computer Science 12896. 2021: 629–640. ISBN 978-3-030-85446-1. doi:10.1007/978-3-030-85447-8_52.
^ Introducing LLaMA: A foundational, 65-billion-parameter language model. 2023-09-11 [2023-10-03]. （原始内容存档于2023-09-11）.
^ meta-llama/Llama-2-70b-chat-hf · Hugging Face. huggingface.co. [2023-10-03]. （原始内容存档于2023-12-30）.
^ Llama 2 - Meta AI. ai.meta.com. [2023-10-03]. （原始内容存档于2024-01-05）（英语）.
^ mistralai/Mistral-7B-v0.1 · Hugging Face. huggingface.co. [2023-10-03]. （原始内容存档于2023-12-29）.
^ AI, Mistral. Mistral 7B. mistral.ai. 2023-09-27 [2023-10-03]. （原始内容存档于2024-01-06）（美国英语）.
^ EleutherAI/gpt-j-6b · Hugging Face. huggingface.co. 2023-05-03 [2023-10-03]. （原始内容存档于2023-03-12）.
^ Biderman, Stella; Schoelkopf, Hailey; Anthony, Quentin; Bradley, Herbie; O'Brien, Kyle; Hallahan, Eric; Mohammad Aflah Khan; Purohit, Shivanshu; USVSN Sai Prashanth; Raff, Edward; Skowron, Aviya; Sutawika, Lintang; Oskar van der Wal. [2304.01373] Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling. 2023-10-03. arXiv:2304.01373  [cs.CL].
^ EleutherAI/pythia-6.9b · Hugging Face. huggingface.co. 2023-05-03 [2023-10-03]. （原始内容存档于2023-12-15）.

[Heaven_2023-1] Heaven, Will Douglas. The open-source AI boom is built on Big Tech's handouts. How long will it last?. MIT Technology Review. May 12, 2023 [2023-12-15]. （原始内容存档于2024-01-05）（英语）.

[Castelvecchi_2023-2] Castelvecchi, Davide. Open-source AI chatbots are booming — what does this mean for researchers?. Nature. 29 June 2023, 618 (7967): 891–892. doi:10.1038/d41586-023-01970-6.

[Thummadi_2021-3] Thummadi, Babu Veeresh. Artificial Intelligence (AI) Capabilities, Trust and Open Source Software Team Performance. Lecture Notes in Computer Science 12896. 2021: 629–640. ISBN 978-3-030-85446-1. doi:10.1007/978-3-030-85447-8_52.

[:0-4] Introducing LLaMA: A foundational, 65-billion-parameter language model. 2023-09-11 [2023-10-03]. （原始内容存档于2023-09-11）.

[5] ta-llama/Llama-2-70b-chat-hf · Hugging Face. huggingface.co. [2023-10-03]. （原始内容存档于2023-12-30）.

[6] Llama 2 - Meta AI. ai.meta.com. [2023-10-03]. （原始内容存档于2024-01-05）（英语）.

[7] stralai/Mistral-7B-v0.1 · Hugging Face. huggingface.co. [2023-10-03]. （原始内容存档于2023-12-29）.

[8] AI, Mistral. Mistral 7B. mistral.ai. 2023-09-27 [2023-10-03]. （原始内容存档于2024-01-06）（美国英语）.

[9] EleutherAI/gpt-j-6b · Hugging Face. huggingface.co. 2023-05-03 [2023-10-03]. （原始内容存档于2023-03-12）.

[10] Biderman, Stella; Schoelkopf, Hailey; Anthony, Quentin; Bradley, Herbie; O'Brien, Kyle; Hallahan, Eric; Mohammad Aflah Khan; Purohit, Shivanshu; USVSN Sai Prashanth; Raff, Edward; Skowron, Aviya; Sutawika, Lintang; Oskar van der Wal. [2304.01373] Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling. 2023-10-03. arXiv:2304.01373  [cs.CL].

[11] EleutherAI/pythia-6.9b · Hugging Face. huggingface.co. 2023-05-03 [2023-10-03]. （原始内容存档于2023-12-15）.

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]