In two weeks, Open AI will come up with its latest AI model which goes by the name Strawberry. It would be larger and more complex than ChatGPT, introducing entirely new experiences in solving complex problems using deep web search capabilities interconnected with advanced reasoning. This announcement of Strawberry means the far-reaching efforts the company has been putting into widening the functionality of its AI systems will soon be a reality.
Codenamed Strawberry, the new OpenAI model is supposed to bring even more reasoning-focused AI into ChatGPT. It's reportedly much more contemplative in answering questions compared to earlier versions. According to a report from The Information, answers with the new AI model are thought through before being given, and responses come out longer and more considerate. Strawberry will only answer text prompts and will not read images like the existing multimodal version of ChatGPT does.
The operation of Strawberry will embed multistep tasks that are complicated in nature. These range from solving complex algebra to devising long-term marketing strategies. Its subtle responses are expected to set the model apart from other AI models. The responses have been designed to take anywhere from 10 to 20 seconds, somewhat above the time it takes for most conversational AIs. This is due to its advanced computation and much more contemplative processing.
On September 5th, OpenAI said it has surpassed 1 million paying users on its business products from 600,000 in April. That's a sign the company's chief executive, Sam Altman, has at least somewhat succeeded in convincing companies to adopt ChatGPT. OpenAI is also weighing new premium subscription tiers for its newest models, called Strawberry and Orion, for as much as US$2,000 a month, far more than the US$20 ChatGPT Plus subscription. Read more
The advanced functionality of Strawberry is exciting, but testers have already remarked on the slower pace of the responses. Its high computational needs may translate into rate limits or higher costs of use. The word is that users will be rate-limited on how many messages they can send in an hour, with faster, premium tiers.
Projects like Strawberry and GPT-Next keep it at the bleeding edge. According to Tadao Nagasaki, the CEO of OpenAI Japan, GPT-Next can be as powerful as 100 times GPT-4 is today. The next-generation model will further develop the new refinements taken from Project Strawberry, bringing in more architectural performance for even better processing capability.