Generative AI is an artificial intelligence technology that uses deep learning algorithms to produce new content from existing data.
AI Tools such as AI Writing for content generation, AI Art and AI Video are taking the world by storm in 2023, with new applications launching daily.
Most notable is CHAT-GPT from OPEN AI. Read our articles or follow our social media channels to learn more.
Chinchilla AI is a 70B-parameter model developed by DeepMind which outperforms Gopher, GPT-3, Jurassic-1 and Megatron-Turing NLG across a large range of benchmarks[1]. However, it has since been surpassed by Google’s 540B-parameter PaLM model[3].
Chinchilla shares the same dataset and architecture as Gopher and shows similar behavior regarding bias and toxicity[1].
ChatGPT is built on the GPT-3 model with 175 billion parameters[2], making it significantly smaller than both Chinchilla and PaLM. ChatGPT alternatives such as ChatGenie, ChatSonic and WriteCream can provide more advanced features than ChatGPT[2].
It is unclear whether Chinchilla AI is smarter than ChatGPT as this depends on the task at hand.
Chinchilla, a 70 billion parameter model developed by DeepMind (owned by Google)[1], has been shown to outperform Gopher (280 billion parameters), GPT-3 (175 billion parameters), Jurassic-1 (178 billion parameters) and Megatron-Turing NLG (530 billion parameters)[2][3][5].
It was trained using the same compute budget as Gopher but with 4 times more data[1]. Chinchilla was able to achieve a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, a 7% improvement over previous models[1].
This performance is impressive not only in terms of magnitude of improvement but also because the model is smaller than all large language models[2].
The research team posits that current large language models are significantly undertrained and proposes three approaches for estimating the optimal tradeoff between model size and number of training tokens[5]
Reviews
There are no reviews yet.