ChatGPT talks to consumers using voice while OpenAI faces European regulators
ChatGPT is still filling headlines by finding new uses, including voice chat. Its maker, OpenAI, however, faces more government scrutiny that could lead to bans, fines and requests and even deletion of data and models, particularly as biometric data and age verification concerns rise to the fore.
Sensory wants ChatGPT to talk to consumers using voice
AI voice company Sensory says its SensoryCloud server can translate speech to text and back, responding to input with a synthetic voice. Armed with that code, Sensory developers are working to integrate voice-enabled consumer electronics with text-based ChatGPT.
The company has created a demo of its voice chat engine, powered by its VoiceChat software. In it, a traveler books a hotel room through a ChatGPT concierge model.
Sensory reportedly avoids unpredictable and incorrect responses, known among developers as AI hallucinations, with careful response selection, the company says.
The voice assistant uses Go, gRPC, Nvidia’s Triton and Amazon’s Global Accelerator.
“Integrating this powerful new technology with our robust voice AI stack is a game-changer for the market,” Todd Mozer, CEO of Sensory, says in a press release. It “allows our customers to create a new generation of infinitely capable voice assistants tailored to a variety of customized domains.”
(Similar code has also been announced by General Motors for an AI voice assistant using ChatGPT.)
Generative AI’s data protection dilemma
That might be the easier task for OpenAI, especially compared to assuaging the fears of some government officials in the European Union.
Regulators and privacy advocatess want the company to be transparent in key aspects of building algorithms, products and services. OpenAI, for example, is being asked to prove it is collecting training data – including personal demographic information — for algorithms legally in the EU, either through consent or “legitimate interest.”
The European Data Protection Board two weeks ago created a task force to coordinate investigations and regulatory enforcement of ChatGPT.
Doubts already have prompted the government of Italy to ban ChatGPT’s use.
Officials have given OpenAI until April 30 to show it is complying with consent and legitimate interest regulations. The company, for example, must display a notice describing OpenAI’s data policies and mandate Italian residents to declare they are at least 18 years old. Italy also wants OpenAI to allow both users and non-users to rectify or erase personal data generated incorrectly by the service.
Some observers say it might be impossible for OpenAI to comply the growing number and market sophistication of rules in developed economies.
A report in MIT Technology Review says AI models collect content from the internet, making it difficult for developers to comply with regulatory regimes. OpenAI could be forced to delete models and data or face fines and more bans.
Not all European countries, however, have taken a tough approach to OpenAI: Ireland’s Data Protection Commissioner Helen Dixon said at a Bloomberg event last week that generative AIs, such as ChatGPT, need to be regulated, but governing bodies should not rush into prohibitions that “really aren’t going to stand up” in court.