People looking to use OpenAI’s popular free version of ChatGPT chatbot soon will no longer have to sign into an account to access the service.
Officials with the company – that essentially kicked off the generative AI explosion when it released ChatGPT in late November 2022 – wrote in a blog post that they are rolling this out gradually “with the aim to make AI accessible to anyone curious about its capabilities.”
The move comes as the generative AI chatbot space becomes more crowded and more competitive with the likes of Google’s Gemini, Anthropic’s Claude, and Copilot from Microsoft, which is a close partner of OpenAI and has invested more than $10 billion in the smaller company. ChatGPT at one time was the record holder with the fastest-growing user base, which reportedly peaked at 1.8 billion web visits in May 2023 and has since seen its growth slow.
The company said that more than 100 people across 185 countries use the ChatGPT service every week.
There are Always Conditions
Using it without an account will come with some conditions. OpenAI said it may use the content generated by users to help train its large-language model (LLM), though users can go into the service’s settings and turn that off.
In addition, the company is including other content safeguards, including blocking prompts and generations in a broader range of categories, hoping to thwart the service being abused by people hopping on without an account.
“There are many benefits to creating an account including the ability to save and review your chat history, share chats and unlock additional features like voice conversations and custom instructions,” OpenAI wrote.
The company also is making ChatGPT more accessible less than two months after Elon Musk, an early investor in OpenAI, sued the company and CEO Sam Altman, claiming they violated its founding goal – that Musk agreed to – of being a nonprofit through its partnership with Microsoft. OpenAI executives pushed back against the lawsuit, saying Musk understood the need to become a for-profit company to raise the money needed to create artificial general intelligence (AGI), the point when AI systems can learn, comprehend and perform as well as humans.
Putting a Hold on Voice Engine
Dropping the account requirement for ChatGPT – the company still has paid versions of the service – is the latest item in what is a busy period for OpenAI. Last week, the company said its latest voice-cloning tool – called Voice Engine – can take text input and a single, 15-second slice of audio to create speech that closely mimics the original speaker, including emotive and realistic voices.
However, the company said in a blog post that it is delaying making Voice Engine widely available until it can set stronger rules and guidelines for its use. Officials noted that the quality of the output is so good that the synthetic voice service could be abused for deepfakes or voice-cloning scams.
“We recognize that generating speech that resembles people’s voices has serious risks, which are especially top of mind in an election year,” the company wrote. “We are engaging with U.S. and international partners from across government, media, entertainment, education, civil society and beyond to ensure we are incorporating their feedback as we build.”
Pulling the Reins on Voice Cloning
Voice cloning already reared its ugly head at least once during the election season, when robocalls sent to Democratic voters in New Hampshire urged them – in a voice that sounded like President Biden’s – not to vote in the primary. The Federal Communications Commission within weeks made it illegal to use AI-generated voices in robocalls.
“We believe that any broad deployment of synthetic voice technology should be accompanied by voice authentication experiences that verify that the original speaker is knowingly adding their voice to the service and a no-go voice list that detects and prevents the creation of voices that are too similar to prominent figures,” OpenAI wrote.
The company also said it is promoting a number of steps, including phasing out voice-based authentication as a security measure for accessing such sensitive data as bank accounts, establishing policies to protect the use of people’s voices in AI, helping the public understand deceptive AI content and other capabilities and limitations, and the creation of techniques for tracking the origin of visual and audio content to make clear whether someone is dealing with AI or a real person.
OpenAI, Microsoft and an AI Supercomputer
In addition, OpenAI and Microsoft reportedly are partnering to create an AI supercomputer called “Stargate.” According to a report by The Information that cited three unnamed sources, the plan is to build a data center and supercomputer for as much as $100 billion, with the plan to build it in five phases and complete it in 2030. The companies are reportedly in the middle of the third phase.
Demand for compute resources generated by the rapidly expanding AI market is forcing hyperscalers to develop plans for massive data center expansion in the coming years. Given the costs associated with handling AI workloads, much of the work – about 70%, according to cybersecurity vendor wiz – is being done in the cloud, necessitating the need for both the systems that can run them and data centers that can house it all.
Microsoft and OpenAI’s plans echo the work Amazon Web Services (AWS) and Nvidia are doing to build an AI supercomputer codenamed “Ceibo.” Nvidia, the world’s top maker of GPUs – which are critical in AI workloads – likely will factor in with Stargate as well, putting an even tighter squeeze on the global market for the Nvidia chips.
