It is thought that Antropic has began to attract $5 billion (about 6.6 trillion won) of investment to develop a next-generation artificial intelligence (AI) model that surpasses Open AI.
TechCrunch, citing an internal report on Antropyc on the sixth (local time), said the corporate will raise as much as $5 billion over the subsequent two years to create a big language model (LLM) that’s 10x more powerful than probably the most powerful AI models currently available inside 4 years. It was reported that they were planning to release a
The LLM under development by Entropic is ‘Claude-Next’, which suggests the successor model of ‘Claude’. Along with excellent performance, the aim is to significantly reduce hallucinations and malfunctions that transcend the guardrail through ‘Constitutional AI’.
Constitutional AI is a term that appeared in a paper at the top of last yr. It refers to a technique wherein AI improves itself by utilizing techniques corresponding to supervised learning or reinforcement learning, somewhat than correcting AI problems through human feedback. Here, the role of humans is to offer only basic rules or principles, so the word ‘structure’ was added as a metaphor. Claude became famous due to undeniable fact that he gave more accurate answers than OpenAI’s ‘GPT’ or Google’s ‘Lambda’.
As well as, Claude-Next requires computing power including tens of hundreds of GPUs since it will need to have computational power that exceeds the present highest-performance AI model by several times. For that reason, it’s estimated that $1 billion (roughly 1.32 trillion won) will likely be needed over the subsequent 18 months.
To this end, it plans to draw as much as $5 billion through Series C investment. In line with The Information, an expert media, it has already succeeded in attracting $1.3 billion (roughly 1.71 trillion won) in early March.
Claude-Next also performs the identical functions as ChatGPT, corresponding to creating documents and pictures and helping with research. Nonetheless, the report argued that “constitutional AI is far less prone to generate harmful outputs, easier to interact with, and easier to regulate”.
“This model could have an effect on automation in lots of areas of the economy,” he predicted. As well as, the conclusion is that “if this model actually appears, it is going to be difficult for other firms to catch up by 2025-2026.”
Antropic gave an official response to this, saying, “We plan to announce additional products and can give you the chance to discuss them soon.”
The corporate opened in 2021 with all OpenAI employees, including research vice chairman Dario Amorey and policy director Jack Clark. After concentrating on research for a yr and a half, Claude was released at the top of last yr, followed by a closed test, followed by a industrial version in March. Currently, 15 partners are using Claude.
It received an investment of 300 million dollars (about 400 billion won) from Google, and in addition signed a contract to make use of computing and cloud. As a result of this point and the performance of Claude, it is taken into account a rival of OpenAI contracted with Microsoft.
Reporter Lim Dae-jun ydj@aitimes.com
