News

OpenAI Targets Families With New Voice, Image Features



ChatGPT continues to evolve past simple text prompts. Last week, OpenAI announced plans to integrate its latest image generator, and today it added voice and image capabilities to its app.

The chatbot can now accept images in prompts, as well as speak on its own with hyper-realistic AI-generated voices. The goal is to offer a “new, more intuitive type of interface,” OpenAI says— and to entice subscriptions for its $20-per-month ChatGPT Plus service.

Curiously, most examples in the announcement are about how children and families can use these features. Perhaps OpenAI has identified a new target audience—or an antidote to the apocalyptic predictions currently associated with its product.

The promotional video for the new AI voice generation capability begins with someone asking to hear a bedtime story about “the super duper sunflower hedgehog named Larry.” ChatGPT replies: “Larry was a unique hedgehog unlike any other. He had bright sunflower petals instead of spines.” The user then asks, “What was Larry’s house like?” It’s easy to imagine kids enjoying asking it questions to expand on the story, giving parents a break.

(At its recent fall event, Amazon previewed a similar feature: Explore with Alexa, an exclusive addition to Amazon Kids+ that will let kids ask Alexa questions about animals and nature.)

“The new voice technology—capable of crafting realistic synthetic voices from just a few seconds of real speech—opens doors to many creative and accessibility-focused applications,” OpenAI says. The company also points out you can use it to “settle a dinner table debate,” or have conversations on-the-go.

To transcribe the voice prompts, OpenAI will use Whisper, its open-source speech recognition system model. It warns, however, that “the model is proficient at transcribing English text but performs poorly with some other languages, especially those with non-roman script, [so] we advise our non-English users against using ChatGPT for this purpose.”

Regarding the new ability to accept images, OpenAI recommends using it to “help your child with a math problem by taking a photo, circling the problem set, and having it share hints with both of you.” Users can “circle the problem set” using a new drawing tool in the mobile app.

In fairness, OpenAI provides non-family-related examples about how image prompts can be useful: Taking a picture of the contents of the fridge and asking for recipe ideas, snapping a photo of a landmark to get information about it, and analyzing a complex graph at work. It’s a whole new dimension for prompt engineers to explore.

As with all OpenAI communications of late, the announcement ends by acknowledging how bad actors can exploit the new technology. It says voice generation could be used to impersonate public figures, or even disparage any member of the public.

The company says it has “taken technical measures to significantly limit ChatGPT’s ability to analyze and make direct statements about people since ChatGPT is not always accurate and these systems should respect individuals’ privacy.”





Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button