ChatGPT is evolving faster than we can assess how it may affect us personally

Lu Feiran
The latest version of the human-like language app can interpret photos on your smartphone and write prose more creatively.
Lu Feiran

Barely a month after people began worrying that ChatGPT might replace them in their jobs, the artificial language model has been updated to deliver more information, including the ability to interpret images and sounds.

Released last week and available in limited form via ChatGPT Plus, GPT-4 is more "reliable, creative, and able to handle much more nuanced instructions than GPT-3.5," according to its developer, US-based OpenAI.

The new version can recognize and analyze images.

For an example, if one shows GPT-4 the photo of a smartphone using a monitor cable to charge, the app is able to recognize that it's actually a joke. Again, if it is shown a picture of a bunch of balloons tied to the ground and asked what might happen if the strings are cut, the app will reply that the balloons will fly away.

ChatGPT is evolving faster than we can assess how it may affect us personally

Screenshot of a promotional video for OpenAI's GPT-4 model.

However, this image identification function is currently available only on applications for programmers, but not everyday users.

Still, people are already experimenting with GPT-4 and say it outperforms previous versions.

Liangziwei, a technology blogger, said he found that GPT-4 has a greater ability to solve advanced mathematic problems and is able to create small video games with more concise and organized codes.

"Take the classic single-player sliding tile puzzle game 2048, for example," he said. "We directed GPT-3.5 to create such a game within 80 lines of codes and it failed. But GPT-4 did it and also managed to solve a calculus problem in a simpler way."

GPT-4 seems to be a better "author" than its predecessor as well.

Liangziwei asked both versions to write a statement on behalf of the troubled soccer club Guangzhou F.C.

GPT-3.5 gave a quite straightforward and "dry" version, while GPT-4 added many details about the club's glorious past and the problems it is now facing, using emotional words. In short, it sounded more like a human statement.

Wang Lei, secretary general of Shanghai AI Industry Association, told Shanghai Daily that the concept of GPT-4 must have been in preparation for a long time and its release is aimed mainly at commercializing the technology.

"The key factor in the operation of GPT-4 is low-cost public participation," he said. "When people can spontaneously provide their own social interaction and identity-characteristic information, the commercialization of such AI models will become possible. It may occur in fields such as social networking, online searches, answer-and-reply, art and music creation, and image identification."

ChatGPT is evolving faster than we can assess how it may affect us personally

Office 365 users will be able to use GPT-4 in various Office applications.

In fact, Microsoft's Nuance Communications, a cloud and software company, announced this week that GPT-4 has been introduced into its Azure OpenAI Service.

With an extra payment, Office 365 users will be able to use GPT-4 in various Office applications, such as Word, Excel and PowerPoint.

Peter Durlach, chief strategy officer of Nuance, said the data inputted into GPT-4 "dwarfs" ChatGPT.

According to Zhang Qi, a computer science professor at Fudan University, this introduction might be a gamechanger in the workplace.

"Take writing an article, for example," he said. "Our current method is to search for information online, write an outline from the collected information and then do a story draft. In the future, these processes are likely to be unnecessary, but we do need to learn how to interact with a machine to write -- considerations like what keywords we should input so that it can generate a better article."

Zhang believes that GPT-4 will soon spread to other industries, such as education, finance, legal services and even autonomous driving.

"The multi-modal feature of such AI means that it can simulate and respond more to the on-site environment so that its application can be very quickly developed," he said.

Now here is the question: While international technology companies are head over heels in love with GPT-4, what's going on in China?

ChatGPT is evolving faster than we can assess how it may affect us personally

Chinese technology giant Baidu expects its AI language model Ernie Bot to be a "gamechanger."

Wang Lei said China, despite its capabilities in the development of artificial intelligence, is not yet widely promoting the technology nor applying it to domestic platforms.

Last week, Chinese Internet giant Baidu became the first mainland company to unveil an multi-modal AI-powered chatbot, called Ernie Bot.

More than 30,000 companies have applied to join that Baidu platform, which is being hailed as a "gamechanger."

Meanwhile, another Chinese-language generative AI tool, issued by veteran investor Kaifu Lee, is expected to make its debut in the near future.

"To some extent, such products in China are somehow 'driven to take action,'" Wang said. "The product launch of a certain AI model was displayed through pre-recorded screenshots rather than a real-time demonstration. We hope that such companies will shoulder social responsibility now that they enjoy somewhat of a domestic monopoly. And we expect that they will develop products in the general public interest."

Special Reports