Let’s chat about Generative AI

The popularity of ChatGPT, a chatbot released by OpenAI in late 2022, paves the way for generative artificial intelligence (AI) to increasingly be a part of mainstream conversation. Early use cases for generative AI are in text, image and code generation. The technology is still at the experimental stage, with potential for integration in creative tasks, writes Nick Patience, research director, AI applications and platforms at 451 Research.

Machine learning

Generative AI refers to AI models that are taught on large training sets aimed to predict the probability of a sequence of words and to generate text. Some of the earliest examples were large language models that emerged from Google’s Bert in 2018. They can perform functions such as summarizations, language translations and poetry writing. The more recent models can generate text or images from text descriptions. Some can even generate software codes or sequences of proteins.

Underlying models of generative AI

A few big tech firms, such as Meta Platforms and Google, are working on generative AI models as part of their business. Some of the underlying models owned by commercial firms include AI21 Labs, Aleph Alpha, Anthropic, Cohere AI, Google’s DeepMind Gopher, Hugging Face Bloom (backed by the French government), Meta OPT-175B, OpenAI GPT-3, Codex and DALL.E.

Generative AI categorization

One way to categorize companies in the generative AI space is based on the applications of the technology:

Text generators – Text generators produce text when prompted by instructions or questions. Organizations in this field include Compose AI, Copysmith, Grammarly, Jasper.AI, Kaizan, Lavender, NovelAI, Textio, TypeWise and Wordtune.

The results are not always accurate — misinformation is a risk, while potential for abuse is high in academic settings, for instance. Early examples of use cases in text generation include general-purpose writing assistants and automated marketing copy writers.

Image generators – Image generators produce images when prompted by a text description. Organizations in this field include DALL.E, Mage.Space, Midjourney, OpenArt, Rosebud.AI and Stable Diffusion.

Code generator – Code generators produce software code from natural language questions. Organizations in this field include A12SQL, Debuild, GitHub Copilot and MutableAI.

Overcoming Obstacles to Scaling AI

Overcoming Obstacles to Scaling AI


The increasing interest in ChatGPT has opened many potential use cases — and potential pitfalls — for generative AI. It is still an experimental tool, but text generation, in particular, holds the potential to be integrated into many applications, and could have major implications in areas such as chatbots and enterprise search.

Want insights on AI trends delivered to your inbox? Join the 451 Alliance.