GamesBeat is thrilled to partner with Lil Snack to bring customized games exclusively for our audience! As fellow gamers, we know this is an exciting new way to interact with the GamesBeat content you already enjoy. Start playing games here.
Justin Davis, after selling Spectrum Labs, is back with a new AI startup called Nurdle. Today, the company emerges from stealth mode with support from the same backers and aims to revolutionize AI deployment for enterprises.
Nurdle has secured funding from key investors like Greycroft, Intel Capital, and Twilio Ventures. The company has surfaced following the acquisition of Spectrum Labs by ActiveFence. Nurdle was initially incubated within Spectrum Labs.
Utilizing cutting-edge technology, Nurdle’s goal is to make AI deployments for enterprises faster, cheaper, and more accurate by offering custom language models. This new venture is an extension of a promising project within Spectrum Labs aimed at refining custom Large Language Models (LLMs) using proprietary “lookalike” data technology.
The company employs a unique mix of human-generated data precision and the scalability of synthetic data to make custom AI applications more accessible for businesses.
Davis, CEO of Nurdle, mentioned that Spectrum Labs served as the testbed for Nurdle’s innovative AI solutions, initially developed for content moderation. The technology, which focuses on human behavior for natural language understanding, has now been extended to improve AI performance in various communication-focused applications.
Nurdle’s methodology in content creation, data labeling, and quality assurance has shown impressive results. It has reduced the time required by data scientists by five to ten times, achieving 92% accuracy for human-generated and labeled data, all at just 5% of the usual cost. This efficiency has significantly sped up production timelines.
Before Nurdle’s launch, Spectrum Labs, known for AI-driven text moderation, was last valued at $146 million, with clients such as Riot Games, Grindr, and The Meet Group. The acquisition by ActiveFence was an all-cash deal, underscoring the importance of this move in the AI space.
Justin Davis explained that during his time at Spectrum Labs, they developed a second LLM while attempting to create better datasets. These datasets proved highly effective, achieving 92% accuracy, which led to the realization that this technology could benefit many other companies.
Josh Newman, CTO of Nurdle, stated that general LLM applications like ChatGPT often fail specific business needs due to their generalized data, leading to inaccuracies. Nurdle addresses this issue by generating high-quality, domain-specific synthetic datasets.
NurdleGPT is a second-generation synthetic data generator focused on creating domain-specific unstructured text data, useful for training language models, and can assist with various types of text like chat logs, call transcripts, emails, and articles.
Nurdle is offering workshops to product managers, data scientists, and AI/ML engineers to share insights and solutions, aiming to democratize AI applications. For more information, visit Nurdle’s platform.
Nurdle’s technology empowers product teams by providing advanced AI solutions, making deployments faster, more affordable, and more straightforward. The NurdleGPT technology allows the creation of specialized “lookalike” datasets, enhancing the accessibility and practicality of AI applications across industries. Interested parties can test their data for free at Nurdle.ai.
Justin Davis noted the importance of ensuring AI products are privacy-safe and secure for children, expressing confidence in their intellectual property and the need for a supportive environment to develop these technologies.
The creative process at Nurdle allows game companies and other businesses to fine-tune datasets to develop better-performing AI systems. By generating synthetic data that adheres to privacy regulations, Nurdle helps train AI models more accurately.
Nurdle’s methods involve a process known as retrieval augmented generation (RAG), which ensures the AI refers to specific data before answering questions, thus preventing errors or “hallucinations.” This system requires ample datasets to benchmark and test accuracy effectively.
With 18 team members and financial backing, Nurdle is well-positioned to address the growing AI needs across various sectors, from gaming to healthcare and beyond. The company’s mission extends to creating better datasets for AI development, enabling diverse industries to enhance their operations using advanced AI technologies.
ActiveFence, involved in vetting Nurdle’s tech, continues to work on trust and safety solutions, reflecting the game industry’s ongoing need for AI and datasets.
Nurdle has conducted pilot testing and is now prepared to launch as a separate entity. The company is engaging with early partners and customers across various sectors, ready to expand its innovative solutions.
GamesBeat remains engaged with its audience, ready to introduce more exciting collaborations and developments in the gaming world. Stay updated with the latest news by subscribing to our daily newsletter.