LLMs Info

LLMs hold the key to generative AI, but some are more suited than others to specific tasks. Here's a guide to the five most powerful and how to use them. Aug 8th, 2023 12:03pm by Charles Humble

Modern Large Language Models (LLMs) are pre-trained on a large corpus of self-supervised textual data and are then tuned to human preferences via techniques such as reinforcement learning with human feedback (RLHF).LLMs have seen rapid advances over the last decade or so, particularly since the development of GPT (generative pre-trained transformer) in 2012. Google’s BERT, introduced in 2018, represented a significant advance in capability and architecture and was followed by OpenAI’s release of GPT-3 in 2022 and GPT-4 this year. At the same time, while open-sourcing AI models is controversial given the potential for abuse in everything from generating spam and disinformation to misuse in synthetic biology, we have also seen a number of open-source alternatives in the last few months, such as the recently introduced Llama 2 from Meta.

From sources across the web

1. LLaMA is a family of autoregressive large language models, released by Meta AI starting in February 2023. For the first version of LLaMA, four model sizes were trained: 7, 13, 33, and 65 billion parameters.  Wikipedia
2. GPT-4 is a large multimodal language model developed by OpenAI. It is the fourth generation in the GPT series of models and is known for its advanced reasoning, complex instruction understanding, and creativity. GPT-4 can generate human-like text, translate languages, write different kinds of creative content, and even generate computer code. It is also capable of answering complex questions and providing detailed explanations. GPT-4's Potential GPT-4 has the potential to revolutionize many industries, including customer service, education, and healthcare. It can be used to create personalized experiences, automate tasks, and provide real-time assistance. 

3. BLOOM (BigScience Large Open-science Open-access Multilingual Language Model) LLM is a highly sophisticated text generation model that can produce coherent text across 46 natural languages and 13 programming languages. Developed by BigScience, it's a product of a collaborative effort involving volunteers and organizations, leveraging industrial-scale computational resources and open science principles. The model excels in generating human-like text and can perform text tasks it hasn't been explicitly trained for by interpreting them as text generation tasks. BLOOM is built on a transformer-based architecture, specifically a modified version of the Megatron-LM GPT2, with a decoder-only setup. It boasts 176,247,271,424 parameters, demonstrating its massive capacity for learning and language understanding. The model was trained on the Jean Zay Public Supercomputer, utilizing 384 A100 80GB GPUs, highlighting the immense computational power required for its development. The training data, known as ROOTS, is a composite of data from various sources in 59 languages, emphasizing the model's extensive multilingual capabilities. One of BLOOM's key features is its open-access nature, with the model, code base, and training data distributed under free licenses. This initiative aims to democratize access to large language models (LLMs), making powerful AI tools more accessible to a broader audience. Over 1,000 researchers from more than 70 countries and 250+ institutions contributed to BLOOM's development, underscoring the global collaborative effort behind this project. The BLOOM LLM stands out as a monumental achievement in the AI landscape, showcasing the power of collaboration, open science, and advanced computational resources. Its ability to generate coherent text in a vast array of languages, along with its open-access model, positions BLOOM as a pivotal tool in the advancement of AI applications and research worldwide.

4. Cohere is a leading AI platform for enterprises that helps build powerful, secure applications that search, understand meaning, and converse in text. source. Cohere's large language models (LLMs) enable applications to perform various tasks, including natural language processing, question answering, code generation, and more. source. Cohere's AI solutions are used in various industries, including healthcare, hospitality, and real estate, to improve customer experiences, streamline operations, and drive innovation.

5. Falcon
6. Guanaco
7. OpenLLaMA
8. StableLM
9. Alpaca
10. MPT
11. T5
12. Gemini
13. BERT
14. Claude v1
15. Dolly
16. Google
17. LaMDA
18. Llama by Meta AI
19. Orca
20. PaLM
21. RedPajama
22. Vicuña