Archive — US — techcrunch
Thinking Machines amps up its bet against one-size-fits-all AI with its firstopen model, Inkling
Thinking Machines Lab has launched Inkling, an open-weight AI model with 975 billion parameters that organizations can adapt for their specific needs. Unlike traditional models from competitors, Inkling allows customization and is positioned as a starting point for enterprises. The model, which generates calibrated text outputs, has been trained on various types of data and aims to outperform rigid, closed models.
Key points
- Inkling is an open-weight AI model released by Thinking Machines Lab, enabling external modifications.
- With 975 billion total parameters, Inkling operates on about 41 billion for specific tasks, keeping costs lower.
- The model was trained on 45 trillion tokens across text, image, audio, and video, but primarily outputs text.
- Inkling is designed for organizations to fine-tune, emphasizing that customer expertise is crucial for customization.
- Thinking Machines positions Inkling as well-rounded, not as the best overall model, focusing on adaptability over dominance.
- Recent research suggests customized AI can outperform centralized models while costing significantly less.
- Thinking Machines claims to have developed Inkling in about nine months, far quicker than competitors like OpenAI.
- The company has not disclosed its funding strategies or revenue model, emphasizing that Tinker, its customization platform, is crucial for monetization.
Why it matters
- The trend towards open-source and customizable AI models reflects a wider desire for flexibility in AI deployment.
- Concerns over proprietary models' limitations are gaining traction, exemplified by Microsoft and Hugging Face's warnings.
- Inkling's approach resonates with broader IT trends that favor private or open-source models for production AI over costly proprietary solutions.
- The ongoing shift to customizable AI can be linked to the rise in data privacy and control among enterprises seeking to maintain competitive advantages.
Overall, Inkling represents a strategic move away from one-size-fits-all AI, catering to organizations seeking tailored solutions, and reflects a significant trend towards open and adaptable AI models in the technological landscape.