Chinese company Baidu has applied for a patent for an artificial intelligence (AI) technology designed to translate animal sounds into human language, as reported by the South China Morning Post on May 7, 2025. This innovative technology aims to bridge the communication gap between humans and animals by interpreting various forms of animal vocalizations and behaviors.
Baidu reportedly filed the patent application in December 2024, but the details were not disclosed by the China National Intellectual Property Administration until May 6, 2025. The patent outlines a sophisticated method for collecting data from animals, which includes analyzing vocalizations, body language, behavioral shifts, and other biological signals to assess their emotional states. These assessments are then translated into a designated human language.
The technology draws on multiple branches of AI, including machine learning, deep learning, and natural language processing. This multifaceted approach allows the system to learn from a variety of inputs and improve its accuracy over time. According to the application, if the system encounters a sound that does not match its historical data, researchers will manually label the input to retrain the model, ensuring continuous improvement and adaptation.
While the filing reveals Baidu's significant technical ambitions, it does not specify when or if the company plans to commercialize this groundbreaking technology. You Yunting, a senior partner at Shanghai Debund Law Firm, noted that the patent review process can take one to three years, or even longer in more complex cases. This uncertainty raises questions about the timeline for potential consumer access to the technology.
In a parallel development in the AI sector, Fastino AI announced the launch of its new Task-Specific Language Models (TLMs) on May 6, 2025. This innovative AI model architecture promises to deliver 99X faster inference than traditional large language models (LLMs) and was trained using less than $100,000 in low-end gaming GPUs.
Fastino's recent funding round raised $17.5 million, led by Khosla Ventures, which is known for being an early investor in OpenAI. This brings Fastino's total funding to $25 million. The funding round also included participation from Insight Partners, Valor Equity Partners, and notable angel investors such as Scott Johnston, the former CEO of Docker, and Lukas Biewald, the CEO of Weights & Biases.
Developers can now access the TLM API, which features a free tier allowing up to 10,000 requests per month. The TLMs are designed for specific tasks, with initial models focusing on summarization, function calling, text to JSON conversion, personally identifiable information (PII) redaction, text classification, profanity censoring, and information extraction.
Ash Lewis, CEO and co-founder of Fastino, explained the motivation behind creating TLMs: "We started this company after our last startup went viral and our infrastructure costs went through the roof. At one point, we were spending more on language models than on our entire team. That made it clear: general-purpose LLMs are overkill for most tasks. So we set out to build models that worked for devs." He emphasized that Fastino's models are not only faster and more accurate but also cost a fraction to train compared to traditional models.
Fastino's TLMs can operate on low-end hardware, such as CPUs or gaming GPUs, and deliver market-leading accuracy, achieving inference speeds that are 99.67 times faster than existing LLMs. This efficiency is particularly appealing to enterprises, which often seek performance on a narrow set of tasks.
Jon Chu, a partner at Khosla Ventures, remarked that Fastino’s technology allows enterprises to create a model tailored to their specific needs, achieving better-than-frontier model performance while maintaining portability and speed. This adaptability opens up new use cases for generative models that were previously impractical.
Fastino is also breaking away from industry norms by offering a flat monthly subscription model that eliminates per-token fees. This approach allows developers to access the complete suite of TLMs without worrying about unpredictable costs. For enterprise customers, Fastino's TLMs can be deployed within a customer’s Virtual Private Cloud (VPC), on-premise data center, or at the edge, ensuring control over sensitive information while leveraging advanced AI capabilities.
George Hurn-Maloney, COO and co-founder of Fastino, stated, "AI developers don't need an LLM trained on trillions of irrelevant data points – they need the right model for their task. That’s why we’re making highly accurate, lightweight models with the first-ever flat monthly pricing – and a free tier so devs can integrate the right model into their workflow without compromise."
Both Baidu and Fastino are pushing the boundaries of AI technology, albeit in different realms. Baidu's efforts to decode animal communication could revolutionize how humans interact with the animal kingdom, while Fastino's TLMs aim to optimize the efficiency and cost-effectiveness of AI applications across various industries. As these technologies evolve, they hold the potential to transform our understanding of both the natural world and the digital landscape.