AI Errors Continue Piling Up As OpenAI's Ex-CTO Mira Murati Launches An AI Startup

OpenAI is known as the pioneer and undisputed leader of the generative artificial intelligence (GenAI) era, thanks to its powerful and mighty flagship AI chatbot ChatGPT.

OpenAI is also known for possessing a long list of former executives who’ve left the company to go on and form their own artificial intelligence (AI) companies.

One of the more notable figures on this list is Elon Musk, who was a part of the startup before it went big and became a household name.

Elon Musk went on to build xAI, an AI company that just recently released its newest family of AI models—Grok 3—a product that Musk calls “The smartest AI on Earth.” As per reviews, the chatbot lives up to the self-assigned title. Elon Musk is also on the warpath to acquire OpenAI now.

During OpenAI’s peak, the company lost both co-leads of OpenAI’s Superalignment endeavor, a division dedicated to developing safe AI, at the same time.

While Jan Leike soon joined Anthropic, Ilya Sutskever said he would reveal his next move later. It came in the form of creating a new AI startup called Safe Superintelligence Inc. or SSI, a company that went on to raise $1 billion in cash.

The next wave of resignations came in September 2024, when the company saw numerous key leaders quitting. This included Ermira “Mira” Murati, OpenAI's Chief Technology Officer since 2018, who even held the position of Chief Executive Officer for a little while when current showrunner Sam Altman was ousted.

Now, Mira Murati has made a move that’s all too familiar for OpenAI.

Through a post on X, Murati announced that she launched a new AI startup called Thinking Machines Lab.

“Our goal is simple, advance AI by making it broadly useful and understandable through solid foundations, open science, and practical applications,” reads her post.

The company will focus on helping people adapt AI systems suited to their specific needs, develop strong foundations for powerful AI systems, and foster a culture of open science aimed at benefiting the whole field. It will also look to build multimodal systems that work with people collaboratively, instead of fully autonomous AI systems, while building AI that works for everyone.

Thinking Machines Lab will see Murati assume the role of CEO, while Barret Zoph (former VP of Research (post-training) at OpenAI) will be CTO, and John Schulman (pioneer of deep reinforcement learning and creator of PPO, cofounder of OpenAI, co-lead of ChatGPT and OpenAI post-training team) will be the company’s Chief Scientist.

All in all, the startup possesses a team of around 30 that comprises scientists, engineers, and builders who hail from companies such as OpenAI, Meta, Mistral, Netflix, Google, Google DeepMind, Google Brain, Character.AI, Twitter, FAIR, HuggingFace, and more. Around 66% of the company comprises former OpenAI employees.

Thinking Machines Lab is also looking to hire more people to join them bridge key gaps that remain in AI systems and make such systems “more widely understood, customizable and generally capable.”


Despite how far AI has come, there still are a few noticeable errors in its system.

A recent report by BBC found that leading AI chatbots such as OpenAI's ChatGPT, Microsoft's Copilot, Google's Gemini, and Perplexity AI generate inaccurate information and distort facts when summarizing news stories.

Through the study, BBC discovered that more than half of the generated responses possessed major flaws.

The study included presenting each tool with 100 news articles from BBC’s website and asking them to generate summaries. The responses included factual errors, misquotations, and outdated information, as well as incorrect dates, numbers, and statements.

It’s not just news summaries that exhibit such problems.

U.S.-based personal injury law firm Morgan & Morgan pushed an urgent mail to its workforce of over 1,000 lawyers regarding AI-generated errors, especially its knack for citing fake cases and creating fake information, and lawyers on the wrong side of such scenarios would be terminated.

This came as a federal judge in Wyoming threatened to sanction two lawyers from the firm who used fictitious case citations in a lawsuit against Walmart, as per a Reuters report. In this case, one of the lawyers admitted that he used an AI program that "hallucinated" the cases—a move credited to an inadvertent mistake.

Do you think AI-generated content needs to be regulated or completely excluded from specific industries?

Let us know in the comments below!