ChatGPT received a major update this week, as OpenAI announced GPT-4, the next-gen AI that powers the chatbot. The upgrade is available only to ChatGPT users who purchase the $20/month Plus subscription. But that’s a price some people might be willing to pay, considering GPT-4’s multimodal capabilities.
What exactly does that mean? Well, thanks to GPT-4, ChatGPT can receive different types of input, not just text. You can show the chatbot a picture and the AI will see it. If that’s still not clear enough, there’s a mind-blowing example of what ChatGPT can do. The bot now gets memes, as it can understand the comedy in a photo just like a human.
That’s not to say OpenAI delivered the ultimate AI bot or that the chatbot will stop offering incorrect answers to your questions. It’s just that GPT-4 brings brand new powers to ChatGPT, which you’ll be able to put to good use in different ways.
Twitter user Riley Goodside posted two examples of GPT-4 multimodal support. One shows ChatGPT correctly interpreting a funny meme. The other shows the bot solving a problem in an image. Both are equally impressive, although the meme is the best way to exemplify the novel powers of ChatGPT. After all, memes play a big role in day-to-day communication and social media.
As you can see below, the user asked ChatGPT what was funny about an image featuring three panels to see whether ChatGPT got the joke. The images showed an iPhone connected to a Lightning charger that looks like an old VGA cable, which you’d use to hook up a monitor to a desktop computer.
The user wanted ChatGPT to describe the three images panel by panel while explaining the humor.
ChatGPT put its new GPT-4 powers to good use to understand the meme and explain it accordingly. Here’s the bot’s conclusion:
The humor in this image comes from the absurdity of plugging a large, outdated VGA connector into a small, modern smartphone charging port.
ChatGPT executed the command just as the user had requested. As you can see in the image below, the AI uses the GPT-4 multimodal input to see all three photos and then describes each panel in detail.
Now that ChatGPT can easily understand memes, it’s only a matter of time before it can produce memes of its own.
The second example in the tweet above is also impressive. The user gives ChatGPT instructions in English, asking the AI to solve a problem written in French in an image the user provides. GPT-4 multimodal features shine again as the bot quickly offers the answer in English, as instructed.
Unlike the meme, which everyone gets, the problem-solving example is harder to verify. One could always screenshot the reply and ask ChatGPT to explain everything with memes. Still, even if you don’t understand the solution, it’s clear that the AI bot understood the task and solved the French problem in English.