Ex-Google AI Researcher Starts Robotics Firm ‘Integral AI’ in Tokyo
An AI startup originating from Silicon Valley is looking to Japan to demonstrate how AI can transform one of the largest industrial robot supply chains globally. Integral AI Inc., a company established five years ago by former Google researchers Jad Tarifi and Nima Asgharbeygi, focuses on developing AI models designed for automated systems, including robots and self-driving cars. The company has collaborated with auto parts manufacturer Denso Corp. since 2021 to facilitate the teaching of new skills to industrial robots through observation of demonstrations. The 15-person startup is engaging in preliminary talks with Toyota Motor Corp., Sony Group Corp., Honda Motor Co., Nissan Motor Co., and Mitsui Chemicals Inc. to present their vision of how artificial intelligence can enhance manufacturing processes. The subsequent phase involves a human operator providing a robot with a language prompt, such as “make a coffee,” allowing the robot to autonomously learn the process, as informed.
Japan hosts numerous leading industrial robot manufacturers, such as Fanuc Corp. and Yaskawa Electric Corp., while SoftBank Group Corp. is acquiring the robotics division of ABB Ltd. The country is home to factory automation providers like Mitsubishi Electric Corp. and Kawasaki Heavy Industries Ltd., with Japanese companies accounting for an estimated 29 percent of the global supply of industrial robots, as reported. Integral has a role to play because “Japan is strong in robotics, but they’re not strong in AI and computing,” Tarifi stated. The 42-year-old, who initiated Google’s first generative AI team in 2013, is among a rising group of AI doctorate holders who regard the functions of the brain’s neocortex as essential for developing AI architecture and algorithms that replicate the learning processes of a child.
Tarifi aims to develop AI models capable of distilling information using minimal data while processing new information without inadvertently erasing previous data — a crucial aspect of learning. “Such models would enable companies to push ahead in physical AI and handle sophisticated tasks such as designing new batteries, discovering materials and drugs or powering humanoid robots,” Tarifi said. Ultimately, the goal is to assist companies in instructing robots to construct new robots. “They might build a cooking robot, they might build a cleaning robot, or they might build a factory robot that builds an iPhone,” he stated. The capacity for self-learning may ultimately liberate machines from the necessity of updates. According to Tarifi, existing large language models, including OpenAI’s ChatGPT and Google’s Gemini, necessitate additional human-guided training, which may constrain flexibility, efficiency, and reliability.
Confronting life-and-death scenarios as a child in war-ravaged Lebanon, Tarifi understood from a young age that “to actually have meaningful impact with AI, you need to affect the physical world, not just the digital world,” as stated in a 2024 interview — author of Conversations with the Future: 21 Visions for the 21st Century — on the Singularity.FM podcast. Integral, having raised approximately $5.5 million to date, is now pursuing around $10 million in a new funding round to enhance its model and prepare for public release. “That’s pocket change compared with what big tech companies are spending on AI, but adequate for developing an algorithm,” Tarifi said. The company aims to achieve greater scale following the anticipated launch of Integral’s Genesis model later this year. “The company’s claims are extremely bold,” Danaylov stated in a recent interview. “But when you cannot afford to use or recreate the paradigm, you have no other option but to invent a new one.”
