OpenAI's GPT-5 Launch Fails: Users Demand GPT-4o Return Amid Glitches
OpenAI’s highly anticipated GPT-5 model, intended as a significant leap forward in artificial intelligence, has instead faced a tumultuous launch, provoking widespread user discontent and calls for the return of older, more familiar versions. The initial rollout was marred by a series of technical glitches, notably a crashing autoswitcher that misrouted queries, making GPT-5 appear less capable than its predecessors. Compounding these issues were restrictive rate limits for users and a highly publicized “mega chart screwup” during a live demonstration, which CEO Sam Altman later acknowledged on Reddit.
In response to the backlash, which Altman likened to “villagers gathering outside of Dr. Frankenstein’s castle,” OpenAI has moved quickly to implement fixes, double rate limits for Plus subscribers, and promise greater transparency and customization in future updates. A key point of contention for users was the abrupt removal of GPT-4o, a model cherished for its distinct personality and emotional intelligence. Users flooded online forums demanding its reinstatement, leading Altman to admit that OpenAI had underestimated the community’s attachment to 4o. He has committed to restoring GPT-4o for paid users while the company continues to refine GPT-5, highlighting that raw intelligence alone isn’t always what the user base values most in an AI assistant.
Beyond OpenAI’s internal challenges, the broader AI landscape continues to see remarkable developments, particularly in the financial sector. Leopold Aschenbrenner, a former OpenAI researcher who was part of its superalignment team, has reportedly raised over $1.5 billion for his AI-focused hedge fund, ‘Situational Awareness.’ This impressive feat comes despite Aschenbrenner having no prior professional investing experience and having been controversially fired from OpenAI in April 2024 amid accusations of leaking sensitive information. The fund, named after a viral essay he published detailing his predictions for Artificial General Intelligence (AGI), has posted an astonishing 47% return in the first half of 2025, significantly outpacing the S&P 500. Aschenbrenner’s strategy involves focusing on AI-tangential investments, including companies in the semiconductor, infrastructure, and power sectors, positioned to benefit from the ongoing AI boom. This success underscores a shifting dynamic in the hedge fund industry, where deep insight into emerging technology may now be valued as highly as traditional financial acumen.
In another groundbreaking application, Google and NASA are collaborating on the development of an AI medical assistant for astronauts. Dubbed the Crew Medical Officer Digital Assistant (CMO-DA), this system is designed to diagnose and treat astronauts during deep-space missions, where communication delays with Earth make real-time human medical consultation impractical. Running on Google Cloud’s Vertex AI platform and utilizing open-source models like Llama 3 and Mistral-3 Small, CMO-DA has demonstrated up to 88% accuracy in diagnosing injuries during initial tests. The system aims to address critical gaps in space healthcare, such as the impossibility of immediate evacuation and communication latency. NASA plans to enhance CMO-DA with capabilities like ultrasound imaging, biometric data integration, and training on space-specific health conditions. This initiative not only promises a significant upgrade in astronaut medical care but also holds the potential to drive advancements in remote healthcare on Earth, benefiting underserved and isolated communities. The prospect of an AI as the primary medical expert aboard future space flights moves us closer to a new era of autonomous support in extreme environments.
The rapid pace of AI innovation extends across various domains. xAI recently made its next-generation Grok 4 freely available globally for a limited period, introducing a new ‘long press’ feature to transform images into video with Grok Imagine. In the competitive world of AI chess, OpenAI’s o3 model dominated the Kaggle AI chess tournament, securing every game against formidable rivals including DeepSeek R1, Grok-4, and Gemini 2.5 Pro. Meanwhile, Roblox open-sourced Sentinel, an AI model designed to enhance child safety by filtering inappropriate chat messages on its platform. Microsoft also unveiled Copilot 3D, a tool that converts images into usable 3D models with a single click, streamlining integration into games, animation, and virtual/augmented reality applications. These diverse advancements collectively illustrate the profound and pervasive impact of AI, reshaping industries from entertainment to engineering and beyond.