It was found that generative artificial intelligence (AI) startups are pushing for diversification of the essential models they use for development.

It’s analyzed that the mixture of wariness against counting on a particular model and the demand to cut back the price required for developing a customized model worked together.

Reuters reported on the twenty ninth (local time) that as OpenAI couldn’t handle the influx of demand and servers went down often, startups that used it were searching for alternatives and turning their eyes elsewhere.

In keeping with the report, Tome, a startup developing an AI storytelling format, is providing a service that creates slides by composing a story when text is entered based on OpenAI’s ‘GPT-3’. We began exploring ways to make use of AI models.

Antropic’s language model, which known as a rival of Open AI, was added to the text generation tool, and the image generation tool is planned to be modified from Open AI’s ‘Dali’ to Stability AI’s ‘Stable Diffusion’.

“The aim of the experiment is to seek out the model that may best increase the speed and quality of every story generation,” CEO Keith Faris Tom explained.

Jasper, which develops writing tools, also introduced language models from OpenAI’s rivals Coher and Antropic. Hyperlight, which developed one other copywriting app, is responding to varied requests from users by introducing several language models.

It is understood that these corporations are using the open AI model to generate long texts, and aside from that, using models reminiscent of Coher to extend speed while reducing costs.

“OpenAI servers often go down,” said Srinath Sridar, CEO of Reg.AI, which generates marketing text. “We wish to offer a greater experience to our customers and lower costs through the use of various models.”

George Matthew, an AI investor at Insight Partners, said that just as many service providers appeared up to now web age, multiple basic model providers are needed for healthy ecosystem functioning within the generative AI era.

Reuters introduced that there may be an industry consensus that AI developers and investors should reduce their reliance on a single model to offer more reliable services, reduce costs, and utilize quite a lot of specialized models.

Jeong Byeong-il, member jbi@aitimes.com