Generative AI Brings Cost of Creation Close to Zero, Andreessen Horowitz’s Martin Casado Says
‘I think it’s going to creep into our lives in ways we least expect it,’ Mr. Casado said at the WSJ CIO Network Summit
The value of ChatGPT-like technology comes from bringing the cost of producing images, text and other creative projects close to zero, according to Andreessen Horowitz General Partner Martin Casado.
With only a few prompts, generative AI technology—such as the giant language models underlying the viral ChatGPT chatbot—can enable companies to create sales and marketing materials from scratch quickly for a fraction of the price of using current software tools, and paying designers, photographers and copywriters, among other expenses, Mr. Casado said.
“That’s very rare in my 20 years of experience in doing just frontier tech, to have four or five orders of magnitude of improvement on something people care about,” Mr. Casado said Tuesday at The Wall Street Journal’s CIO Network Summit in Palo Alto, Calif.
On Tuesday, ChatGPT maker OpenAI released GPT-4, the startup’s latest AI language model trained on massive amounts of data, as well as human feedback, to generate natural language response to user prompts. The technology can also write computer code.
OpenAI said the tool is designed to be capable of advanced reasoning, leveraging “broader knowledge across a range of domains,” according to a statement. It also features more advanced capabilities to generate, edit and collaborate with users on creative projects, the company said.
Microsoft Corp. , which is investing billions of dollars in OpenAI, and other large enterprise-technology firms say they plan to add ChatGPT-like tools to their business-software products and services.
Yet many corporate technology chiefs have taken a wait-and-see approach to the technology, which has developed a reputation for producing false, misleading and unintelligible results—dubbed AI ‘hallucinations’.
Though ChatGPT, which is available free online, is considered a consumer app, OpenAI has encouraged companies and startups to build apps on top of its language models—in part by providing access to the underlying computer code for a fee.
For businesses, Mr. Casado said, there are “certain spaces where it’s clearly directly applicable,” such as summarizing documents or responding to customer queries. Many startups are racing to apply the technology to a wider set of enterprise use cases, he said.
But Mr. Casado said GPT-4 and other generative AI tools are less likely to be successful when retrofitted to an existing business model. Instead, he expects the technology to spark entirely new businesses and organizations.
“I think it’s going to creep into our lives in ways we least expect it,” Mr. Casado said.