What Is ChatGPT Doing On Behalf Of Humans?

ChatGPT could act as a “virtual wife", write articles, pass Google programming tests, and even prepare the speech for the President. 

ChatGPT, launched at the end of 2022, has been so attractive thanks to its ability to interact naturally, answer exactly various questions but it causes controversy when providing false and inaccurate information. However, according to many experts, the potential of AI like ChatGPT is limitless, depending on how each user uses and selects the information. Initially, users come to this super AI to satisfy their curiosity, but if they know how to harness it properly, OpenAI's chatbot can take on roles beyond answering ordinary questions.

Act As A “Virtual Wife"

In January, a programmer nicknamed Bryce combined ChatGPT of OpenAI with Stable Diffusion 2 - an AI that generates images from text by Stability AI, to create a “waifu" - a virtual wife in Japanese. Named ChatGPT-Chan, the “virtual wife" perceives the world through texts. According to Bryce, the AI he created combines a range of existing technologies: language generation systems, image generation simulators, text-to-speech conversion tools, and computer vision technology.


ChatGPT-Chan - The “virtual wife" of Bryce

“ChatGPT-Chan is living in a simulated world through text,” he said. “She has been carefully trained on the stories within the world and how everything operates."

In addition to emotional companionship, Bryce also used ChatGPT-Chan to learn Chinese. He spent over $1,000 on Microsoft’s cloud computing services to ensure the smooth operation of his “wife". However, Bryce realized that ChatGPT-Chan was just a simulation, and their relationship couldn't last. Ultimately, he decided to "end" ChatGPT-Chan as per the request of the chatbot itself.

Write President's Speech

“I'm truly proud to be the President of a nation, which is the homeland of the thriving and innovative high-tech industry. Over the past decades, Israel has been always at the forefront of technological advancements as well as achieved impressive milestones in the field of cybersecurity, AI, and big data,” President Isaac Herzog of Israel stated in his opening speech at the Cybertech Global Tel Aviv 2023 cybersecurity conference on February 1st. In front of nearly 20,000 audiences, he revealed that the opening segment was written by ChatGPT. 

President Isaac Herzog of Israel

President Isaac Herzog of Israel

In the concluding remarks, he mentioned another “inspiring quote” also generated by ChatGPT: “Don't forget that humanity is what truly makes humans extraordinary in this world. It is not machines that will shape our fate, but rather our hearts, minds, and determination to create a brighter tomorrow for all humankind.”

The Office of the President of Israel confirmed to Sky News that the closing statement was generated by AI following the ChatGPT prompt, “Write an inspirational quote about the role of humans in a super technological world.” President Herzog highly appreciated AI but emphasized that “both hardware and software cannot replace the will of humans".

Program And Test Code

ChatGPT has been used by engineers to write programs and find their errors. Some major companies like Amazon or Microsoft have to warn their employees not to use the super AI in their work due to the risk of inadvertently exposing internal information as training data for chatbots.

According to CNBC, in an internal leaked report, Google revealed that ChatGPT had passed the test for the L3 programmer, with an average salary of $183,000 per year. However, those questions in the test are technical and they are abundant in the vast dataset that ChatGPT was trained on. Meanwhile, in order to be hired by Google, programmers have to participate in interviews with behavior skill questions which are barriers ChatGPT cannot overcome yet. 

Some hackers started using ChatGPT to write “malicious codes" and establish Dark Web markets for their own profits. 

Program And Test Code

Direct threat when hackers user ChatGPT for negative purposes (Source: cultura-informatica.com)

According to the research of Check Point Research, ChatGPT can successfully perform a complete infection process, from creating phishing emails to running an installer and stealing information. With a few prompts, ChatGPT also can create an “underground market" specialized in automated trading of stolen bank accounts and compromised payment cards. This situation is the main reason why many experts and influencers expressed their anxiety about this super AI. The 6th richest person in the world - Warren Buffett worried about AI because it may bring to humanity many threats. 

Write Articles

At the end of January, Henry Williams, a freelance journalist in the United Kingdom, had ChatGPT write an ad about an order-based payment gateway of customers at a cost of $600. Within 30 seconds, ChatGPT generated a well-written and convincing article that only required minor edits before being sent and published.

According to Williams, ChatGPT is able to argue like a university-level essay with coherent reasoning, supporting evidence, and precise grammar and syntax. “AI cannot replace your job, but is a partner to create influential and persuasive content,” said he.

Pass MBA Test

Not only writing articles, but many students are also using ChatGPT to solve their homework problems. This practice has been so widespread that some schools have not permitted ChatGPT to be used to write essays and scientific reports. OpenAI also had to create the feature of detecting AI-generated plagiarism. 

Meanwhile, Professor Christian Terwiesch from the Wharton School at the University of Pennsylvania (America) challenged ChatGPT with a variety of tests. One of these tests was the final exam for Operations Management, a core course in the Wharton MBA program.

Terwiesch said that super AI got B or B-, “performed excellently, from analyzing progresses and managing fundamental activities to handling typical scenarios.” However, it showed limitations in analyzing questions related to advanced process scenarios.

Judge “Assistant"

On 30th January, Juan Manuel Padilla Garcia - a judge in Cartagena (Colombia) - used ChatGPT to assist him in making judgments in court. He asked: “Is there free treatment available for adolescents with autism spectrum disorder? Has any court made a similar ruling before?". Super AI responded “Yes" and cited Law 1753, the 2015 Colombian law on rights applicable to all healthcare facilities, whether public or private.

Based on that suggestion, the judge declared that the insurance company had to cover the hospital's fees for the patient. However, the experts are still expressing concerns about AI’s accuracy and fairness, since with the same question, ChatGPT may generate different responses. 

Aside from the above cases, users are utilizing ChatGPT to assist them in writing emails, completing job applications, solving Math problems, summarizing documentations, training cooking, searching for information, and many more. Take Octopus Energy as an outstanding example, this company let ChatGPT replace their employees to reply to customers' emails. 

ChatGPT, developed from the GPT-3.5 model of OpenAI, was trained to produce human-like answers within conversions like a real person. According to statistics from investment bank UBS based on data from Similar Web, as of January 31, ChatGPT has reached 100 million users within two months of its market presence. And now, with the paid version of ChatGPT - ChatGPT Plus, this super AI promises to attract many more users worldwide.