\n<\/p>\n
Last Friday, AI startup OpenAI filed a new application to trademark products<\/a> associated with its brand \u2014 \u201cOpenAI\u201d \u2014 with the U.S. Patent and Trademark Office (USPTO). Normally, this wouldn\u2019t be newsworthy. Companies file for trademarks all the time. But in the application, OpenAI hints at new product lines both nearer-term and of a more speculative nature.<\/p>\n For example, the filing lists hardware including headphones, goggles, glasses, remotes, laptop and phone cases, smartwatches, smart jewelry, and virtual and augmented reality headsets \u201cfor AI-assisted interaction, simulation, and training.\u201d<\/p>\n OpenAI last year confirmed<\/a> that it was working with former Apple designer Jony Ive on a hardware project, and the startup\u2019s CEO, Sam Altman, on Sunday told Korean outlet The Elec<\/a> that OpenAI hopes to develop AI-powered consumer hardware \u201cthrough partnerships with multiple companies.\u201d But in that same interview, Altman cautioned that it\u2019d take \u201cseveral years\u201d to complete even a prototype AI device.<\/p>\n OpenAI\u2019s trademark application also mentions robots \u2014 specifically \u201cuser-programmable humanoid robots\u201d and \u201chumanoid robots having communication and learning functions for assisting and entertaining people.\u201d <\/p>\n OpenAI recently began hiring for a new robotics team led by Caitlin Kalinowski, who\u00a0joined<\/a>\u00a0the startup to lead hardware last November from Meta\u2019s AR glasses division. According to job listings<\/a> and reporting from The Information<\/a>, OpenAI is looking to test robots \u2014 possibly humanoid in form \u2014 powered by custom sensors and AI that can operate with human-like intelligence in real-world settings.<\/p>\n Further down in its filing, OpenAI makes references to custom AI chips and services for \u201cleveraging quantum computing resources to optimize AI model performance.\u201d<\/p>\n It\u2019s long been rumored that OpenAI is creating custom chips to run its AI models. The company has a division focused on co-designing chip components<\/a>, and reports suggest<\/a> OpenAI aims to bring a custom chip to market with semiconductor manufacturers Broadcom and TSMC as early as 2026.<\/p>\n OpenAI\u2019s plans for quantum computing \u2014 assuming it has any set in stone \u2014 are murkier. But last year, the startup added to its technical team a former quantum systems architect at quantum computing startup PsiQuantum.<\/p>\n As The Register noted in a piece from March<\/a>, quantum computing has the potential to drastically improve the efficiency of training AI models thanks to its ability to perform vast numbers of calculations simultaneously. With the computing costs of AI showing no signs of coming down anytime soon, perhaps OpenAI sees a future in training models on hardware far architecturally different than the machines on which it relies today.<\/p>\n Then again, trademark applications are often intentionally written to be broad and not necessarily indicative of a company\u2019s product roadmap. OpenAI\u2019s may well reveal domains the startup is exploring \u2014 or at least, has considered exploring. But exactly when \u2014 or whether \u2014 any of the tech mentioned in the filing makes it to market is anyone\u2019s guess.<\/p>\n TechCrunch has an AI-focused newsletter!\u00a0Sign up here<\/a>\u00a0to get it in your inbox every Wednesday.<\/em><\/p>\n<\/div>\n