Writer uses synthetic data to lower costs while training its AI models.
Writer.com’s AI Studio WRITER AI STUDIO The fastest way to build AI applications Build AI applications and workflows that are fully integrated with Writer’s full-stack generative AI platform. Screenshot from Writer.com official website
A new AI startup is taking on AI giants OpenAI and Anthropic.
The San Francisco-based startup, called Writer, is a new AI model in the game that has a more “cost-effective” approach than other giants.
According to CNBC, the company paid for all expenses, including data, GPUs, and manpower, with just $700,000. Developing similar AI models usually costs millions of dollars. Writer is currently raising up to $200 million from investors, which would bring its valuation to $1.9 billion. By 2023, after raising $100 million, its value will exceed $500 million.
Now, the AI startup is cost-effective because it uses synthetic data, AI-generated information designed to mimic real-world data while protecting users’ privacy. It has become a popular way to train AI models as industry concerns over confidentiality grow and the risk of publicly available training data running out in the coming years.
Amazon uses synthetic data for Alexa, while Meta uses it to train its Llama model. Recently, OpenAI also noticed this trend and started using it for its models.
Is Synthetic Data Better?
For users, it is a better option for companies to switch to synthetic data because it protects their privacy. However, VCPost mentioned that experts believe this will lead to a decline in model performance and increase bias between AI models.
For Writer, their approach is to continuously improve synthetic data. CTO and co-founder Alshikh clarified that they do not use fake or random data to train AI models. They take real and factual information and convert it into synthetic data to make it “cleaner” and easier for model training.
Thus, Writer’s generative AI will allow users to use their LLM for multiple purposes of human-like text. In addition to generating text, it can also analyze, summarize, and develop custom AI applications. Currently, Writer has over 250 enterprise customers who prefer their approach.