Google Brain
An AI research team under Google, now merged into Google DeepMind, that proposed speculative decoding techniques in 2022–2023 nearly simultaneously with the DeepMind team.
Core Facts
BF16 was originally designed by the Google Brain team for TPUs, retaining the same 8-bit exponent as FP32 while reducing the mantissa from 23 bits to 7 bits.
The Chain-of-Thought prompting paper published in 2022 by Jason Wei et al. of Google Brain was the first to systematically demonstrate that adding step-by-step reasoning examples significantly improves the accuracy of large models such as PaLM on mathematical reasoning tasks
The ReAct framework was proposed by a Princeton University and Google Brain team in a 2022 paper.
Timeline (last 90 days)
The name Gemini was explained by Jeff Dean as referring to the two merged teams (Google Brain and DeepMind), layered with the NASA Gemini program and the Gemini myth.
TypeSafe's management team is: CEO Diogo Almeida, CTO Erik Gafni, and COO Sasha Sheng, who previously worked at OpenAI, Google Brain and Meta's FAIR research group respectively.
TypeSafe AI's CEO Diogo Almeida previously worked at Google Brain and OpenAI, was one of the core authors of the InstructGPT paper, and contributed to work on RLHF, ChatGPT, and GPT-4.
Diogo previously worked with Jeremy Howard and at Google Brain before joining OpenAI.
Google Brain's NASNet required running on hundreds of GPUs for thousands of hours to find an architecture that performed well on specific tasks.
Andrew Ng is one of the pioneers of deep learning, a co-founder of Coursera, and an early core figure in the Google Brain project.
Character.AI was founded in 2021 by former Google Brain researchers Noam Shazeer and Daniel De Freitas.
Jeff Dean led the founding of the Google Brain team and drove the creation of deep learning frameworks such as TensorFlow.
Quoc Le is a core researcher at Google Brain with pioneering work in neural architecture search and sequence-to-sequence learning.
Chain-of-Thought reasoning was first proposed by Jason Wei and others from Google Brain in 2022.
12 more timeline events
All Facts (20)
BF16 was originally designed by the Google Brain team for TPUs, retaining the same 8-bit exponent as FP32 while reducing the mantissa from 23 bits to 7 bits.
90%VerifiedThe Chain-of-Thought prompting paper published in 2022 by Jason Wei et al. of Google Brain was the first to systematically demonstrate that adding step-by-step reasoning examples significantly improves the accuracy of large models such as PaLM on mathematical reasoning tasks
90%VerifiedThe ReAct framework was proposed by a Princeton University and Google Brain team in a 2022 paper.
90%VerifiedGoogle DeepMind was formed in 2023 through the merger of DeepMind, which was acquired in 2014, and Google's internal Google Brain team.
90%VerifiedChain-of-Thought (CoT) reasoning was proposed by the Google Brain team in 2022 and improves accuracy on complex tasks by guiding the model to expand its reasoning step by step.
90%VerifiedBF16 (Brain Floating Point 16) is a floating-point variant proposed by Google that retains the 8-bit exponent range of FP32 but reduces the mantissa to 7 bits.
90%VerifiedThe Chain-of-Thought prompting technique was formally proposed by the Google Brain team in the 2022 paper "Chain-of-Thought Prompting Elicits Reasoning in Large Language Models".
90%VerifiedThe Transformer architecture was proposed in 2017 by the Google Brain team in the paper "Attention Is All You Need"
90%VerifiedVision Transformer (ViT) was proposed by the Google Brain team in 2020, introducing the Transformer architecture into the field of computer vision.
90%VerifiedThe speculative decoding paradigm was proposed almost simultaneously by teams at Google Brain and DeepMind in 2023.
90%VerifiedChain-of-Thought technology was first systematically proposed by Jason Wei and others from the Google Brain team in a 2022 paper.
80%VerifiedAndrew Ng previously served as the founding director of the Google Brain project and as Chief Scientist at Baidu.
80%VerifiedTensorFlow was open-sourced and released by the Google Brain team in 2015.
80%VerifiedIn 2022, the Google Brain team first systematically documented the phenomenon of 'emergent abilities' in large language models in the paper 'Emergent Abilities of Large Language Models'
80%VerifiedAndrew Ng is a Stanford University professor and co-founder of Coursera; his machine learning course has had more than 5 million registrations since it launched in 2011.
75%VerifiedThe concept of emergent abilities was systematically elaborated by the Google Research team (Wei et al.) in the 2022 paper Emergent Abilities of Large Language Models, which examined more than 200 tasks.
70%VerifiedTransformer架构由Google在2017年的论文《Attention Is All You Need》中提出，其核心自注意力机制能捕捉序列中任意位置之间的依赖关系
70%VerifiedMoE was reactivated in the era of large language models by Google's Switch Transformer (2021) and GLaM.
70%VerifiedCompared to FP16, BF16 has the same 8-bit exponent width as FP32 and therefore a larger dynamic range, making it the mainstream precision choice for current large model training.
70%VerifiedEfficientNet是Google Brain团队于2019年提出的高效网络，通过复合缩放策略同时优化网络深度、宽度和分辨率
65%