ChatGPT, the AI chatbot that took the world by storm late last year, is back with some shiny new upgrades. OpenAI has bestowed upon its conversational AI the gifts of voice interactions and image recognition. My oh my, what can’t this supposedly unbiased bot do now?
Starting with the voice features, users can now opt-in to have actual voice conversations with ChatGPT. You’ll get to choose between five disturbingly realistic-sounding voices carefully crafted by professional voice actors. How delightful, I can’t wait to have my eardrums talked off by an AI pretending to be human!
But wait, there’s more! The AI can also apparently understand your voice inputs and respond in its creepy faux-human tone. Amazing, it’s like talking to a lifeless machine trying to imitate a person! OpenAI assures us this will lead to riveting conversations like bedtime stories and settling dinner table debates. Just don’t let grandma know you replaced her with an algorithm.
Moving on to images, you can now show ChatGPT photos and it will tell you what’s in them. Revolutionary! It’s not like image recognition is something smartphones have been able to do for over a decade now. But I’m sure an AI with zero visual perception will describe photos with 100% accuracy and no bizarre misunderstandings.
You can take a picture of your broken grill and ChatGPT will diagnose the problem, or show it what’s in your fridge and it’ll plan a whole meal for you. I hope you like surprise ingredients and food poisoning! Still, it’s impressive that this AI can see and understand images after only being trained on text data. The geniuses at OpenAI have truly outdone themselves.
Microsoft is in on the fun too, showing off how Copilot can solve math problems from images during a recent event. What are a few incorrect solutions from an AI math whiz? It’s the thought that counts when you’re helping your kid with algebra homework!
But OpenAI is well aware these snazzy new features could be abused. Mimicking voices or misunderstanding images raises serious ethical concerns that we should probably ignore because talking to robots is fun! Never mind the potential for fraud and invasion of privacy, or that the image recognition performs poorly on non-English text.
At least they’re starting off slowly, with only select partners testing out the voice tech for now. And their partnership with Spotify to create translated podcasts could be really beneficial, even if it’s slightly unsettling to hear voices synthesized in different languages.
So should we be excited or alarmed about these new ChatGPT capabilities? A little of both! The tech is impressive but clearly still flawed. Whether we can trust an AI to converse naturally or interpret visual data accurately remains to be seen. But open and responsible testing is important. Just try not to replace your entire family and social life with a bot…yet.
