Big language models expand the reach of AI in industry and business
This article is part of a VB Lab Insights series on AI sponsored by Microsoft and Nvidia.
In July 2022, the AI world and popular press around the world made a buzz with the release of DALL-E 2, the generative AI with 3.5 billion parameters developed by Open AI. Then came ChatGPT, the great interactive conversational language model developed and trained by OpenAI.
Until now, the flashy text-to-image conversion templates have been attracting media and industry attention. But the expansion of public trials of Open AI's new conversational chatbot in December 2022 brought to light another type of Large Language Model (LLM).
LLMs are learning algorithms that can recognize, abstract, translate, predict, and generate languages using very large sets of textual data, with little or no training supervision. They handle various tasks such as answering customer questions or recognizing and generating texts, sounds and images with high precision. Besides text-to-image, a growing range of other modalities include text-to-text, text-to-3D, text-to-video, digital biology and more.
Quietly expanding the impact of AIOver the past two years, LLM neural networks have quietly expanded the impact of AI in healthcare, gaming, finance, robotics, and other fields and functions, including the software development and machine learning in business.
"Large language models have proven to be flexible and capable, capable of answering probing questions, translating languages, understanding and summarizing documents, writing stories, and computing programs," says Bryan Catanzaro, vice president of applied research in deep learning at Nvidia. .
The arrival of ChatGPT marked the emergence of another type of LLM, the foundation of generative AI and transformative neural networks, the latter being increasingly touted as a revolutionary disrupter of the AI, including enterprise applications.
Beginning an influential 2017 research paper, the idea took off a year later with the release of the open-source software BERT (Bidirectional Encoder Representations from Transformer) and the GPT-3 model d 'Open AI. As mAs, these pre-trained models have grown in complexity and size - 10 times a year recently - as have their capabilities and popularity. Today, the largest PaLM 540B and Megatron 530B models in the world are LLMs.
Indeed, as one of the newest and most powerful classes of models, LLMs are increasingly replacing convolutional and recurrent models. A key breakthrough has been the combination of specialized AI hardware, scalable architectures, frameworks, customizable models, and automation with robust “AI-first” frameworks. This makes it possible to deploy and scale production-ready LLMs in a wide range of commercial and consumer enterprise applications across public and private clouds and via APIs.
LLMs can help companies codify intelligence through insights gained in multiple areas, says Catanzaro. This helps accelerate innovation that extends and unlocks the value of AI in ways previously only available on supercomputers.
This article is part of a VB Lab Insights series on AI sponsored by Microsoft and Nvidia.
In July 2022, the AI world and popular press around the world made a buzz with the release of DALL-E 2, the generative AI with 3.5 billion parameters developed by Open AI. Then came ChatGPT, the great interactive conversational language model developed and trained by OpenAI.
Until now, the flashy text-to-image conversion templates have been attracting media and industry attention. But the expansion of public trials of Open AI's new conversational chatbot in December 2022 brought to light another type of Large Language Model (LLM).
LLMs are learning algorithms that can recognize, abstract, translate, predict, and generate languages using very large sets of textual data, with little or no training supervision. They handle various tasks such as answering customer questions or recognizing and generating texts, sounds and images with high precision. Besides text-to-image, a growing range of other modalities include text-to-text, text-to-3D, text-to-video, digital biology and more.
Quietly expanding the impact of AIOver the past two years, LLM neural networks have quietly expanded the impact of AI in healthcare, gaming, finance, robotics, and other fields and functions, including the software development and machine learning in business.
"Large language models have proven to be flexible and capable, capable of answering probing questions, translating languages, understanding and summarizing documents, writing stories, and computing programs," says Bryan Catanzaro, vice president of applied research in deep learning at Nvidia. .
The arrival of ChatGPT marked the emergence of another type of LLM, the foundation of generative AI and transformative neural networks, the latter being increasingly touted as a revolutionary disrupter of the AI, including enterprise applications.
Beginning an influential 2017 research paper, the idea took off a year later with the release of the open-source software BERT (Bidirectional Encoder Representations from Transformer) and the GPT-3 model d 'Open AI. As mAs, these pre-trained models have grown in complexity and size - 10 times a year recently - as have their capabilities and popularity. Today, the largest PaLM 540B and Megatron 530B models in the world are LLMs.
Indeed, as one of the newest and most powerful classes of models, LLMs are increasingly replacing convolutional and recurrent models. A key breakthrough has been the combination of specialized AI hardware, scalable architectures, frameworks, customizable models, and automation with robust “AI-first” frameworks. This makes it possible to deploy and scale production-ready LLMs in a wide range of commercial and consumer enterprise applications across public and private clouds and via APIs.
LLMs can help companies codify intelligence through insights gained in multiple areas, says Catanzaro. This helps accelerate innovation that extends and unlocks the value of AI in ways previously only available on supercomputers.
What's Your Reaction?
