Chat Smarter! OpenAI Upgrades AI Voice Experience
On Tuesday, during an open stream event, Sam Altman, CEO of OpenAI, announced the much-anticipated release of the first major upgrade to ChatGPT’s image-generation capabilities in more than a year.
Now, the technology giant’s new model, GPT-4o, will allow ChatGPT to create and modify native images and photos. It has always fueled the AI chatbot platform until now without the ability to generate and edit images, only text.
According to Altman, GPT-4o native image generation is already live in ChatGPT and Sora, OpenAI’s AI video-generating product, for subscribers to the company’s Pro plan, which costs $200 a month.
OpenAI says the feature will soon be rolled out to Plus and free users of ChatGPT, as well as developers using the company’s API service.
“thinks” a little longer than the model it supplants, DALL-E 3, in order to make what OpenAI says are more accurate and detailed images: output from GPT-4o. It can also be used to edit existing images, including those with people, turning them or “inpainting” details such as with foreground and background objects.
Join our WhatsAPP community..
The feature was powered as OpenAI called the Wall Street Journal, saying it trained GPT-4o by “publicly available data,” in addition to proprietary data compiled using the house’s partnerships with companies like Shutterstock.
While in theory universal, generative AI firms mainly guard such proprietary training data because it is mostly seen to give them a competitive edge. But training data details also provide a fertile ground for IP-related lawsuits, something that disincentivizes companies from throwing up much.
“We’re respecting their rights,” said Brad Lightcap, OpenAI’s chief operating officer in a statement to the Journal. The artists in question are those whose rights are believed to be infringed by output that might have been produced via the use of AI. Policies are also in place that prevent these scenarios from happening.
OpenAI also has an opt-out form that content creators may fill out to officially seek the removal of their work from training datasets, and the company indicated that it respects requests to deny its web-scraping bots from collecting training data, including images, from websites.
Join our WhatsAPP community..
Coming hot on the heels of this release was Google experimenting on native image output for Gemini 2.0 Flash, one of its flagship models. This amazing feature made waves on social media but not necessarily for very good reasons.
There was little guardrails on the image component of Gemini 2.0 Flash, so it allowed people to remove watermarks and create images even of copyrighted characters.
This article was updated at 12 pm PT to include OpenAI’s statement to the Wall Street Journal about the training data for GPT-4o.
Join Our Social Media Channels:
WhatsApp: NaijaEyes
Facebook: NaijaEyes
Twitter: NaijaEyes
Instagram: NaijaEyes
TikTok: NaijaEyes
