OpenAI CEO Sam Altman acknowledged on Reddit that a dearth of computational resources poses a significant obstacle to the company’s ability to release products at its desired pace, citing a notable constraint in this regard.
He explained that all recent advancements in fashion had reached a high level of sophistication, responding to inquiries about the protracted development time for OpenAI’s subsequent AI models. “We also confront numerous limitations and daunting choices regarding how we allocated our computing resources towards many intriguing ideas.”
Counsel that OpenAI has faced significant challenges in securing adequate compute infrastructure to support the development and deployment of its generative models. This week, according to sources cited by Reuters, OpenAI has collaborated with Broadcom for several months on the development of an AI-powered semiconductor designed specifically for fashion applications, which could potentially hit the market as early as 2026.
Due to constrained resources, Altman noted that OpenAI’s lifelike conversational AI will not be acquiring its initially promised creative vision capabilities anytime soon, despite the initial hype in April. During their April press event, OpenAI showcased the ChatGPT app’s ability to run seamlessly on smartphones, responding to visual cues, such as recognizing clothing being worn, through the phone’s camera view.
According to Fortune, Apple deliberately hurried the demo’s release to preempt attention from Google’s I/O developer conference, which coincidentally took place that same week. Despite internal skepticism within OpenAI, few expected GPT-4 to be unveiled – a notion underscored by the secrecy surrounding the voice-only model of Superior Voice Mode, which remained under wraps for months.
According to Altman’s statement within the AMA, there is no definitive timeline for the main launch of OpenAI’s picture generator, DALL-E. Although we don’t have a launch plan yet, he pointed out that “We’re still working on improving the model, addressing security, impersonation, and different issues, and scaling up our computing resources,” said Kevin Weil, OpenAI’s chief product officer, who also joined the AMA in the meantime.
Sora’s development has been marred by technical issues, leaving it lagging behind rivals like Luma, Runway, and others. The innovative Data system, unveiled in February, was criticized for requiring over 10 minutes of processing time to produce a single 1-minute video clip.
In October, one of several co-leads on the Sora project was Tim Brooks, a professional at Google.
Later during the AMA, Altman noted that OpenAI allows “NSFW” content in ChatGPT, stating that they sometimes permit it (“we completely believe in treating adult users like adults”), with a top priority on refining its series of reasoning models and their subsequent iterations. OpenAI is set to unveil a plethora of innovative options at its annual DevDay conference in London, including advancements in picture understanding capabilities.
“We’re excited to share that several outstanding titles will be released in the coming year,” Although we won’t call it GPT-5, nothing.
