OpenAI Will Give U.S. Safety Institute Early Access To New Models - 1

Photo by Levart_Photographer on Unsplash

OpenAI Will Give U.S. Safety Institute Early Access To New Models

  • Written by Andrea Miliani Former Tech News Expert
  • Fact-Checked by

Sam Altman, OpenAI’s CEO, announced today on X that the company will provide early access to the U.S. Artificial Intelligence Safety Institute to the upcoming foundational model.

“As we said last July, we’re committed to allocating at least 20% of the computing resources to safety efforts across the entire company,” wrote Altman, “Our team has been working with the US AI Safety Institute on an agreement where we would provide early access to our next foundation model so that we can work together to push forward the science of AI evaluations.”

Altman said he was excited about this decision and added that they want current and former employees to express their concerns about the AI technologies the company is developing. He emphasized that it was important for the company and its safety plan.

According to TechCrunch , the U.S. Artificial Intelligence Safety Institute—part of the National Institute of Standards and Technology ( NIST )—addresses and assesses risks in AI platforms and OpenAI has already worked on a similar agreement with the United Kingdom just a few weeks ago.

Altman’s announcement comes after people and organizations have criticized the company’s safety measures. In June, multiple current and former employees from OpenAI signed an open letter to express concerns and warn about AI spreading misinformation and becoming a risk to society. Various workers have also quit the company due to safety concerns, including key OpenAI researcher Jan Leike , after OpenAI dissolved the Superalignment team, in charge of addressing AI risks.

Despite the security concerns, OpenAI keeps moving forward. The company just launched the voice feature for ChatGPT Plus , and a new AI-powered search engine called SearchGPT is available for selected users.

The new efforts to add security measures seem to prove that Altman and his team have not forgotten about AI risks and concerns.

OpenAI Rolls Out Voice Feature For ChatGPT Plus - 2

Photo by Miguelangel Perez on Unsplash

OpenAI Rolls Out Voice Feature For ChatGPT Plus

  • Written by Andrea Miliani Former Tech News Expert
  • Fact-Checked by

“We’re starting to roll out advanced Voice Mode to a small group of ChatGPT Plus users,” wrote OpenAI on X, “Advanced Voice Mode offers more natural, real-time conversations, allows you to interrupt anytime, and senses and responds to your emotions.”

We’re starting to roll out advanced Voice Mode to a small group of ChatGPT Plus users. Advanced Voice Mode offers more natural, real-time conversations, allows you to interrupt anytime, and senses and responds to your emotions. pic.twitter.com/64O94EhhXK — OpenAI (@OpenAI) July 30, 2024

The company added that users who get to test the alpha version will get an email with instructions to learn how to use Voice Mode.

According to Forbes , OpenAI delayed its rollout—initially expected to be deployed in June—due to safety concerns and adjustments as the company claimed to need more time to reach “launch standard.” OpenAI set back after the actress Scarlett Johansson requested more information about the AI assistant’s voice called “Sky” as it was “eerily similar” to hers.

The voice Sky—used during the demonstration at the live event—is not available in the new release, and OpenAI hasn’t disclosed more information about the source of the voice.

According to MIT Technology Review , the new voice model powered by GPT-4o will be capable of more natural conversations compared to Siri and Alexa, and capable of recognizing emotions, responding to interruptions, and combining and integrating text and visual capabilities.

OpenAI expects to get more feedback during this period before releasing the voice-enabled chatbot to a wider audience later this year, in the fall. The company said it has tested the model with over 100 external testing team members in 45 languages across 29 countries.

One of the safety measures considered before launching the voice feature was audio deepfakes. The new voice model will not be allowed to create voice deepfakes; it will only use four voices chosen in collaboration with actors.

OpenAI expects to release more features soon—like video and screen sharing—for its AI assistant model. Just a few days ago, the tech company released a new AI-powered search engine called SearchGPT expected to compete against Google.