Generative AI start-up Imbue announced on its official website that it has received US$200 million (approximately 1.45 billion yuan) in financing, with a valuation of more than US$1 billion. This time, Nvidia, Astera Institute, Cruise CEO Kyle Vogt, Notion co-founder Simon Last and others participated in the investment.
Imbue did not launch a C-side ChatGPT product like OpenAI, but focused on general AI agents. Imbue hopes to build a tool that can help people automatically perform repetitive and tedious business processes on computers, such as automatically uploading and downloading data; automatically designing PPT and other operations, all completed by AI. Human workers can use their time on work with higher business value.
To put it simply, the tool Imbue hopes to build is a bit like the AutoGPT that became popular around the world some time ago, or an AGI (artificial general intelligence) product, with the addition of coding and reasoning capabilities.
Currently, Imbue has not released any products and is in the technology research and development stage. However, it still received huge amounts of financing, and we have to marvel at the money-attracting ability of the generative AI track.
According to the data, Imbue was founded in April 2021, formerly known as General Intelligence, and is headquartered in San Francisco, California, USA. On October 21, 2021, it received US$20 million in financing from Tom Brown, OpenAI’s GPT-3 main development leader.
Kanjun Qiu, co-founder and CEO of Imbue, graduated from MIT with a master's degree and is one of the few female serially successful entrepreneurs in Silicon Valley.
According to Imbue, if you want to build a general AI agent, you first need to train a basic large model. Currently, Imbue is pre-training a basic model with more than 100 billion parameters, and is using new financing to invest in 10,000 H100 GPU computing clusters to improve the model's reasoning and output capabilities.
Imbue's product plan is to first train an AI agent that can automatically perform coding, for example, automatically develop a website based on user needs to check the performance and stability of the AI agent. More agents are then trained for prototyping to achieve general purpose.
Then the product prototype is pre-trained through massive codes. Imbue believes that the code can improve the agent's reasoning ability. Because the code is a very objective and simple testing environment, it either passes the test or reports an error, and there are almost no ambiguities.
In addition, Imbue hopes to build AI agents with coding capabilities, which can expand its application scope and capabilities and simplify or automate many complex and tedious businesses.
Imbue said that the product is currently in the technology research and development stage and will be open to all users in the future. It can quickly create AI agents suitable for various fields such as programming, sales, personnel, finance, and finance.
The material of this article comes from the official website of Imbue. If there is any infringement, please contact us to delete it.
END