OpenAI’s conversational ChatGPT AI is over confident but otherwise brilliant

Imagine if Artificial Intelligence (AI) could write you a college essay, or spit out a movie review in the style of Shakespeare. Well, recently we’ve seen a ran ge of AI’s, from the likes of Google and OpenAI, do just that and a lot more including auto-generating blogs, imagery, videos, and even re-creating science experiments, doing homework, and magicking up university lecturers from nothing more than text inputs. And, elsewhere other AI’s are generating code, games, and even new drugs and rocket engines using just the most basic of inputs.

OpenAI last week opened up access to ChatGPT, an AI powered chatbot that interacts with users in an eerily convincing and conversational way. Its ability to provide lengthy, thoughtful and thorough responses to questions and prompts – even if inaccurate – has stunned users, including academics and some in the tech industry.

The Future of Artificial Intelligence, by keynote speaker Matthew Griffin

The tool quickly went viral. On Monday, OpenAI’s co-founder Sam Altman, a prominent Silicon Valley investor, said on Twitter that ChatGPT crossed one million users. It also captured the attention of some prominent tech leaders, such as Box CEO Aaron Levie.

“There’s a certain feeling that happens when a new technology adjusts your thinking about computing. Google did it. Firefox did it. AWS did it. iPhone did it. OpenAI is doing it with ChatGPT,” Levie said on Twitter.

But as with other AI-powered tools, it also poses possible concerns, including for how it could disrupt creative industries, perpetuate AI biases and spread misinformation.

ChatGPT is a large language model trained on a massive trove of information online to create its responses. It comes from the same company behind DALL-E, which generates a seemingly limitless range of images in response to prompts from users. It’s also the next iteration of text-generator GPT-3.

The Future of synthetic Content, by keynote speaker Matthew Griffin

After signing up for ChatGPT, users can ask the AI system to field a range of questions, such as “Who was the president of the United States in 1955,” or summarize difficult concepts into something a second grader could understand. It’ll even tackle open-ended questions, such as “What’s the meaning of life?” or “What should I wear if it’s 40 degrees out today?”

“It depends on what activities you plan to do. If you plan to be outside, you should wear a light jacket or sweater, long pants, and closed-toe shoes,” ChatGPT responded. “If you plan to be inside, you can wear a t-shirt and jeans or other comfortable clothing.”

But some users are getting very creative.

One person asked the chatbot to rewrite the 90s hit song, “Baby Got Back,” in the Style of “The Canterbury Tales;” another wrote a letter to remove a bad account from a credit report - rather than using a credit repair lawyer. Other colorful examples including asking for fairy-tale inspired home décor tips and giving it an AP English exam question which it responded with a 5 paragraph essay about Wuthering Heights.

In a blog post last week, OpenAI said the “format makes it possible for the tool to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.”

While ChatGPT successfully fielded a variety of questions submitted by users, some responses have been noticeably off. In fact, Stack Overflow – a Q&A platform for coders and programmers – temporarily banned users from sharing information from ChatGPT, noting that it’s “substantially harmful to the site and to users who are asking or looking for correct answers.”

Beyond the issue of spreading incorrect information, the tool could also threaten some written professions, be used to explain problematic concepts, and as with all AI tools, perpetuate biases based on the pool of data on which it’s trained. Typing a prompt involving a CEO, for example, could prompt a response assuming that the individual is white and male, for example.

“While we’ve made efforts to make the model refuse inappropriate requests, it will sometimes respond to harmful instructions or exhibit biased behavior,” Open AI said on its website. “We’re using the Moderation API to warn or block certain types of unsafe content, but we expect it to have some false negatives and positives for now. We’re eager to collect user feedback to aid our ongoing work to improve this system.”

Still, Lian Jye Su, a research director at market research firm ABI Research, warns the chatbot is operating “without a contextual understanding of the language.”

“It is very easy for the model to give plausible-sounding but incorrect or nonsensical answers,” she said. “It guessed when it was supposed to clarify and sometimes responded to harmful instructions or exhibited biased behavior. It also lacks regional and country-specific understanding.”

At the same time, however, it does provide a glimpse into how companies may be able to capitalize on developing more robust virtual assistance, as well as patient and customer care solutions.

While the DALL-E tool is free, it does put a limit on the number of prompts a user can do before having to pay. When Elon Musk, a co-founder of OpenAI, recently asked Altman on Twitter about the average cost per ChatGPT chat, Altman said: “We will have to monetize it somehow at some point; the compute costs are eye-watering.”