Google LLC today announced a $300 million investment in Anthropic, a firm that researches and develops “reliable and controlled” artificial intelligence systems, as part of the search giant’s entrance into generative AI, one of today’s trendiest fields of technology.
Google will gain a 10% share in Anthropic as part of the deal. In addition, the corporation will purchase cloud computing capabilities through a “preferred provider” arrangement with Google Cloud.
Google Cloud also wants to develop its own large-scale, application-specific AI-accelerator tensor processing unit and graphics processing unit clusters, which Anthropic will utilize to train and deploy its AI models.
Generative AI refers to artificial intelligence algorithms that can generate text, graphics, and other media from plain-language user prompts. Following the launch of OpenAI LLC’s ChatGPT last year, which is capable of answering questions nearly like a person, the technology has won the hearts and minds of millions of people and produced headlines.
Anthropic was created in January 2021 by a group of researchers, engineers, policy experts, and others who have worked on AI advances such as the massive language model GPT-3, which drives ChatGPT. There is also human feedback reinforcement learning, which allows machine learning models to quickly train to become more conversational based on positive or negative input.
The company’s focus is on AI safety, with the goal of developing more predictable, reliable, and steerable AI systems. The collaborative team hopes to minimize bias in the AI model and solve some of the present limits of conversational AI.
“At Google, we believe it is critical to explore AI boldly and ethically,” said James Manyika, the company’s senior vice president of technology and society. “We are dedicated to creating and delivering helpful and beneficial applications while adhering to responsible principles based on human values and safety.”
Google will have access to Anthropic’s flagship conversational assistant AI product, Claude, as a result of this collaboration. It’s an intelligent agent chatbot that understands and responds to natural language, similar to how conversational AI models like ChatGPT work, but it’s only available in closed beta through a Slack integration right now.
Anthropic refers to the process of developing Claude as “Constitutional AI,” a learning model for AI that employs both reinforcement learning and supervised learning to evaluate its behavior in order to create a more “harmless AI assistant.” According to the research report on the issue, it employs self-improvement to detect and eliminate harmful outputs.
As generative and conversational AI become more common, this will become increasingly crucial. AI depends greatly on the type of data it’s trained on and how dependent people are on it for their day-to-day use. Machine learning algorithms employed in smart agents in the past demonstrated significant racial and gender prejudices, repeated slurs, and promoted falsehoods. Even with new breakthroughs like ChatGPT, this is still a possibility, and OpenAI needed to implement safeguards to fight against it.
The conversational AI space is heating up, with Microsoft Corp. investing $10 billion in OpenAI, the maker of ChatGPT, and adding its service to Azure, as well as incorporating GPT-3.5 into its Sales Viva platform. Baidu Inc., the Chinese search engine, said this week intentions to develop its own ChatGPT competitor in Chinese markets based on the machine-learning model Ernie. Character AI, a neural network AI developed by former Google engineers who worked on the company’s internal LaMDA AI system, could pose additional competition.
