AI Miscellany 6: From AI Native to +AI

Original link: https://pt.plus/notes-on-ai-6/

AI Miscellany 6: From AI Native to +AI

The article Generative AI’s Act Two published by Sonya Huang of Sequoia Capital is a follow-up to Generative AI: A Creative New World published in September last year. In just the past year, the article has been updated to a second edition, and the title has also added “Second screen”.

Act 2

At this time last year, GPT as the world knows it today—that is, versions 3.5 and later—had not yet been released. In that article, Huang proposed four waves of AI:

  • The First Wave: Small Models Dominate (Before 2015) 5 years ago, small models were considered the “state of the art” for understanding language. These small models excel at analytical tasks and can be used for everything from delivery time predictions to fraud classification. However, they are not expressive enough for general generation tasks. Generating human-level writing or code remains a dream.
  • Wave 2: The Race to Scale (2015-present) A landmark paper from Google Research (Attention is All You Need) describes a new neural network architecture for natural language understanding, called Transformer, that can generate High-quality language models that are both more parallelizable and require significantly less training time. These models are few-shot learners that can be tailored to a specific domain with relative ease.
  • Wave 3: Better, Faster, Cheaper (2022 and Beyond) Computing becomes even more affordable. New technologies such as diffusion models can reduce the cost of training and running inference. The research community continues to develop better algorithms and larger models. Developer access expands from closed beta to public beta or, in some cases, to open source.
  • Wave 4: Killer Apps Emerge (Now) As the platform layer consolidates, models continue to get better/faster/cheaper, model access tends to be free and open source, the time has come for a burst of creativity in the application layer Mature.

A year later, Huang announced that generative AI had entered Act 2, the second act. In the structure of traditional drama, the entire drama is divided into separate acts. The second act refers to the second major part or segment of the drama, which usually represents a major development or upgrade of the plot and character relationships in the drama. Huang explains it this way:

This article is reproduced from: https://pt.plus/notes-on-ai-6/
This site is for inclusion only, and the copyright belongs to the original author.