The AI chatbot ChatGPT has received a significant update from OpenAI that adds speech and visual capabilities. It implies that consumers would be able to speak to the chatbot and send photos to it in response to questions.
The AI lab recently disclosed the change. Over the following two weeks, they will first be made available to ChatGPT Enterprise and Plus subscribers of the company, with access to other user groups to follow.
Millions of individuals signed up to use ChatGPT when it first started last year, igniting an artificial intelligence boom. OpenAI is attempting to capitalize on its success and just released its corporate tier for business users, despite the fact that its popularity has now declined. However, it competes in a field that is becoming more and more saturated, with alternatives offered by businesses like Google’s Bard and Anthropic’s Claude.
New ChatGPT features imitate Alexa and Siri
With the voice feature, users will be able to ask ChatGPT questions via a microphone in a manner similar to how you might ask Apple’s Siri or Amazon’s Alexa. Users must enable the feature in order for ChatGPT to react using a voice produced by AI.
“The new speech capability is generated by a new text-to-speech model. Which is able to generating human-like voice from just text and a couple of seconds of sample speech,” according to OpenAI’s blog post on the update. “We collaborated with skilled voice actors to create each of the voices. In order to translate your spoken words into text, we additionally make use of Whisper, our freely available speech recognition system.”
According to the business, their new voice technology “opens opportunities to many innovative and accessibility-focused applications”. The statement does, however, caution that “these capabilities also bring new risks. For instance the possibility for fraudulent actors to pretend to be public figures or commit fraud.”
It claims to be utilizing the technique just for this one particular use case. However, today it was revealed that Spotify, a music streaming service. It is also using it to translate podcasts into more languages by employing AI-generated voices that mimic the presenters’ voices.
ChatGPT is introduced into the picture by OpenAI
In other places, visitors can include one or more photographs in their search. This will allow you to, in the words of OpenAI, “troubleshoot why your grill won’t start. Or explore the inside of your fridge to prepare a meal, or analyse a complex graph for job-related data.”
Furthermore, the GPT-3.5 and GPT-4 multimodal big language models from the company are what enable the system to comprehend images. The business acknowledges that this can put users’ safety in danger.
Moreover, in the blog article, it is said that using vision-based models “presents new difficulties. Ranging from delusions about people to relying on the model’s ability to interpret images in high-stakes domains.” “Before broader deployment, we assessed the model with a variety of alpha testers. And red teamers for risk in areas like radicalism and scientific aptitude.”
Visit CxO Global FORUM or CxO News Live for all the latest updates.
