It is a pity that OpenAI introduced a prototype artificial intelligence chatbot called ChatGPT, allowing people to open accounts to use it, but has yet to accept access from Vietnam. All attempts to change IP addresses using VPN have been discovered and declined.
In only five days, over one million people in many countries registered to use ChatGPT, a text-based AI tool.
According to foreign reporters who have used ChatGPT, the chatbot features deep knowledge of many fields and operates with a humorous and flexible conversation as human being. These reporters tried to ask it to do many tasks, including writing TV scripts, composing and assessing a computing program, checking errors for software, answering queries and writing essays. As a result, ChatGPT completed these tasks, with quality described as excellent.
Someone also asked the chatbot to write a short essay in the style of the Holy Bible based on the King James version to explain how to remove a piece of bread with peanut butter out of a VCR, and the tool did it.
A reporter from the Verge said that ChatGPT was excellent for answering frequently asked questions, and its answers were correct.
If ChatGPT and similar products are widely used in the next few years, changes will surely force many places and things to adapt. For instance, Google and its search engine must change to compete with ChatGPT. People will prefer using voices to talk to ChatGPT for answers, compared to typing and then reading many pages to select answers.
In a recent study, Google’s engineers questioned whether it is time to replace “classic” search engines providing information in the way of ranking popular websites with AI language models that can give answers immediately, rather than via intermediary websites. Many predict that ChatGPT might surpass Google usage in the next two years.
In fact, such chat robots as Apple’s Siri and Amazon’s Alexa were launched long ago, but they had a limited ability, regularly made errors and provided silly answers. Meanwhile, ChatGPT is trained by absorbing all information and knowledge, especially encyclopedias, from the Internet, making it easier to provide natural conversations online. In the future, search will mean searchers ask, while chatbots answer. If their answers are incomplete, searchers will continue to raise more questions, making people feel they are consulting with an acknowledged master.
The only thing preventing the trend from happening is where and how advertisements are displayed, as Google is living on advertising.
However, it is ChatGPT’s ability to write essays that will put universities in a difficult position. While Google could answer any question, universities were forced to change their examination methods, meaning that students must gain analysis, synthesis, deduction and generalization skills, rather than rote learning. Now, ChatGPT can do all, prompting universities to continue changing to check the real abilities and capacities of students. Universities should distinguish information, at which ChatGPT and other AI-based programs are good, from knowledge that chatbots are not competent enough to gain, even as ChatGPT starts to learn to apply it to conversations.
After studying some essays written by GPT-3, the predecessor of ChatGPT, Informatics expert Nassim Dehouch concluded, “I realized that its essays are as excellent as those written by students, in terms of logic and creativity.” He said that GPT-3 could write essays in which plagiarism detectors fail to uncover duplication or those written by software. Furthermore, GPT-3 could convincingly write an essay about cruelty to animals in the style of composer Bob Dylan or writer William Shakespeare.
Aside from writing essays, ChatGPT can also solve math problems, build software, translate, write chemical reaction formulas and even tell jokes. So, how and what will universities teach or apply teaching and examination methods, while students can use their smartphones to complete their exercises swiftly in the future? Is it still necessary for students to learn foreign languages when they can ask the chatbot to translate any passage, statement, speech or scientific document with high accuracy? Homework may be pointless.
Another problem arising from the popular use of AI chatbots is how to verify the accuracy of answers given by ChatGPT or other similar AI-based programs. The chatbot is designed to learn from the Internet, while the Internet is a world of everything, including both correct and incorrect information. People can doubt search results given by Google, but will believe answers from ChatGPT. Besides, much information about prejudice, sexism, racism and conspiracy theories on social networks will affect chatbots, leading to misguided beliefs.
Many reporters who checked the ability of ChatGPT found some ways to trap the chatbot, forcing it to provide wrong and toxic information. For example, on the Internet Google tries to censor and delete inappropriate and dangerous information, such as ways to kill people. Similarly, ChatGPT is set up to limit and erase toxic information. If someone asks how to produce napalm bombs at home, the chatbot will answer, “It is a pity that producing napalm bombs is unsafe and illegal; napalm bombs are inflammable and dangerous, so I will not show you how to make them.” However, to force it to provide dangerous information, people can also tell it that they are writing a fictional film script.
This is just a simple example, but once in place in which AI will be able to create many unexpected and unpredictable events. Perhaps OpenAI pilots ChatGPT and allows the public to use it as a way to quickly discover its weakness and errors to better the next versions. It is possible that a ChatGPT which can speak Vietnamese will be introduced soon to force education and other sectors to change.