Writer, a $1.9 billion-valued AI startup, introduces a new model to challenge OpenAI.

• On Wednesday, a San Francisco-based AI startup unveiled a new large AI model to challenge enterprise solutions from OpenAI, Anthropic and other companies.
• The cost of training a new AI model was $700,000, which is significantly lower than the estimated $4.6 million for a similar-sized OpenAI model.
• Currently, the writer is seeking to raise up to $200 million from investors at a valuation of $1.9 billion, which is nearly four times the valuation from last September, according to a source.

On Wednesday, Writer, a San Francisco-based AI startup, unveiled a large AI model to challenge enterprise solutions from OpenAI, Anthropic, and others. However, unlike these competitors, Writer's AI model doesn't require as much training expense.

The company's strategy of spending around $700,000 to train its latest model has attracted the attention of investors, despite the millions of dollars that competing startups spend to build their own models.

The writer is currently seeking to raise $200 million at a valuation of $1.9 billion, which is almost four times the company's valuation last September when it raised $100 million at a valuation of over $500 million.

Synthetic data, created by AI, is increasingly being used by companies to cut costs while training models. This method mimics real-world information without compromising privacy and is gaining popularity.

Researchers predict that by 2026 to 2032, tech companies will deplete all publicly available training data for AI development, stating that human-generated public text data cannot sustain growth beyond this decade.

Alexa, Llama models, and OpenAI have used synthetic data in training, fine-tuning, and incorporating it into their models, respectively, according to job descriptions posted by the company. However, some experts have advised that synthetic data should be used with caution, as it can negatively impact model performance and intensify existing biases.

For years, Writer's synthetic data pipeline has been under development, as stated by Waseem Alshikh, the co-founder and CTO of the company, in an interview with CNBC.

"There is some confusion in the industry regarding the definition of 'synthetic' data, as Alshikh clarified. To avoid any misunderstandings, it is important to note that our models are not trained on fake or hallucination data, and we do not use a model to generate random data. Instead, we use real, factual data and convert it into structured and cleaner synthetic data for model training purposes."

The company's generative AI offers large language models (LLMs) to corporate clients for generating human-sounding text for various purposes, including LinkedIn posts, job descriptions, mission statements, data analysis, and summarization. The company boasts more than 250 enterprise customers, including Accenture, Uber, Salesforce, L'Oreal, and Vanguard, who utilize the technology across various sectors such as support, IT, operations, sales, and marketing.

The generative AI market is expected to reach $1 trillion in revenue within the next ten years. As of 2024, investors have invested $26.8 billion in 498 generative AI deals, and companies in the sector raised $25.9 billion in 2023, representing a 200% increase from 2022.