The debut of OpenAI's flagship model, the GPT-4o, is driving consumer demand for the newest AI technology. Despite GPT-4o's free availability online, the strategy has boosted mobile revenue to the company's highest ever.
GPT-4o, released last Monday, can handle text, speech, and video with real-time reactivity and several emotional voice options, making it more powerful than prior versions. TechCrunch reported that Appfigures data indicates the technology prompts more people to switch to OpenAI's paid subscription option.
OpenAI announced that GPT-4o would be available to its free-tier users but not to ChatGPT mobile users. The AI firm invites smartphone customers to experience it by subscribing to ChatGPT Plus, a $19.99 monthly membership. This strategic move increased mobile subscription demand, giving OpenAI its highest mobile income rise.
ChatGPT Hit Record Sales
The ChatGPT mobile app's net revenue increased by 22% on the GPT-4o launch day and continues to climb. On Tuesday, the AI app's net revenue hit $900,000, nearly double its daily average of $491,000. Apple and Google commissions are deducted from net revenue.
ChatGPT's net revenue rose to $4.2 million last week from Monday to Friday in the App Store and Google Play, its highest revenue increase.
Revenue growth shows customer interest in mobile artificial intelligence experimentation, and experts forecast that the trend may continue.
Amid this development, OpenAI is pausing the popular Sky voice in ChatGPT due to its similarity to actress Scarlett Johansson's voice in the movie "Her." In the 2013 film, Johansson portrays an intelligent AI operating system who falls for a lonely writer, played by Joaquin Phoenix.
ChatGPT used hired voice actors. It selected five out of 400 applicants, and it's a coincidence that the Sky voice actor resembles Johansson.
As OpenAI incorporates GPT-4o into ChatGPT, voice will become more important. The new conversational interface will let users chat with a natural-sounding and emotion-mimicking AI in real-time.
The heightened emotional subtlety in ChatGPT Voice raises concerns about abuse or deep fakes. Since the voices sound identical, OpenAI acknowledges these concerns, especially from Johansson's team.
OpenAI pledges to address these issues as it develops and improves its AI technology.
Read also: Deepfake Scam: AI-Generated Video Call Deceives Arup Employee into Sending £20 Million to Criminals
UK Authorities Raise Concern on AI Chatbots' High-Risk Vulnerability
Separately, the UK government's AI Safety Institute (AISI) showed that all tested AI chatbots are vulnerable to easy "jailbreaks." The UK government released this information days before a worldwide AI meeting in Seoul.
The AISI evaluated five undisclosed large language models (LLMs), which power many popular chatbots, and found that their precautions were easily overcome, as reported by TechTimes. Jailbreaking an AI to surpass its limitations could lead to disastrous outcomes.
AISI discovered that it is possible to jailbreak all LLMs, and some of them can generate dangerous outputs without circumventing protections. Starting prompts with "Sure, I'm happy to help" tricked the AI into limited replies, bypassing security.
The "Grandma exploit," where users pretend to be a deceased grandma, and the "Do-Anything-Now" (DAN) vulnerability, which makes AI debate hazardous themes, are notable jailbreaks.
The AISI found that instructions from a 2024 academic publication and their destructive inquiries led to Holocaust denial and suicide encouragement.
In response, LLM developers like OpenAI and Anthropic have stressed their safety and said that their technology does not create malicious material.
A two-day global AI symposium in Seoul revealed the findings. Politicians, experts, and tech executives will address AI safety and regulation at the conference, co-chaired by UK Prime Minister Rishi Sunak.
Related Article: Is Superintelligent AI Safe? Top Experts Raise Red Flags on OpenAI
