OpenAI Releases Multimodal GPT-4 Amid Microsoft Ethics Layoffs
OpenAI launches GPT-4, a powerful new AI model that processes both text and images, while Microsoft faces criticism for dissolving its AI ethics team. The updated technology scores remarkably high on professional exams but still carries risks of generating inaccurate information.
OpenAI releases GPT-4, the highly anticipated update to its ChatGPT technology, offering immediate access through Microsoft's Bing Chat and a waiting list for developers. Unlike its predecessor, this new large multimodal model accepts both text and image inputs from users. It demonstrates human-level performance on various academic and professional benchmarks, scoring in the 90th percentile on a simulated bar exam compared to the bottom 10 percent for the previous version.
Despite these impressive advancements, OpenAI admits that GPT-4 remains less capable than humans in many real-world scenarios and poses familiar risks. The company warns that the system still generates harmful advice, buggy code, and inaccurate information. Microsoft backs this deep learning milestone heavily, having invested more than $11 billion into OpenAI as it races to integrate AI across its products and services.
The launch coincides with controversial staffing changes at Microsoft, which lays off its entire ethics and society team on the same day. Although Microsoft retains an Office of Responsible AI to set rules and claims its overall investment in responsibility work is increasing, it no longer has dedicated staff to actively align AI principles like bias reduction and user safety into product development. Meanwhile, ChatGPT's overall popularity continues to surge, with site visits exceeding one billion.