Anthropic, an artificial intelligence (AI) research startup that plans to raise up to $5 billion in funding over the next two years and develop a more powerful model to compete with OpenAI, was founded in 2020 by former OpenAI researchers.
Among the company’s goals are interesting goals such as applying artificial intelligence to nearly a dozen large industries. However, one of the most important missions for us is to produce a bot that is much more advanced than the artificial intelligence chatbots such as GPT-3 and 4 developed by OpenAI.
It will be 10 times more capable than today’s most powerful AIs
Anthropic’s presentation for the Series C funding round includes the company’s long-term goals and strategies. In the presentation, it is stated that Anthropic will need to spend 1 billion dollars in the next 18 months to train an artificial intelligence model, which it calls the “frontier model”, while it is claimed that the border model will be 10 times more capable than today’s most powerful artificial intelligence.
The company says this model is a next-generation algorithm for AI self-learning. The temporary name of the model is “Claude-Next”. Anthropic developed an AI training technique it calls “constitutional AI” to ensure that the boundary model is compatible with human intentions. Constitutional AI uses a simple set of principles to enable systems to answer questions and perform tasks.
Anthropic estimates that the boundary model will require 10^25 FLOPs (floating point operations). That means it’s several times larger than today’s largest models. Of course, how long this calculation takes depends on the speed and scale of the system doing the calculation. Anthropic states that it uses clusters with tens of thousands of GPUs in its presentation.
The main motivation is to develop an AI that is more reliable and responsible than OpenAI’s models.
The motivation behind Anthropic’s development of the boundary model is to compete with OpenAI. OpenAI’s GPT-3 is considered a groundbreaking model in natural language processing. But OpenAI’s models also have some issues. For example, GPT-3 can sometimes produce inaccurate or biased information or conflict with human values.
Anthropic aims to develop more reliable and responsible models of OpenAI. The vision of the company is to use artificial intelligence in many areas such as health, education, finance, e-commerce, gaming and social media. In the presentation, giant names such as Google, Amazon, Facebook, Microsoft, Netflix, Spotify, Uber are among the potential customers and partners of the company.
It is not yet clear when Anthropic will complete the frontier model or how it will be used. But given the company’s ambitious plans and the backers behind it, it seems certain it will. What are you thinking? Please do not forget to share your thoughts with us in the comments.
40049
