What is ChatGPT, the viral social media AI?

- ADVERTISEMENT -

A new artificial intelligence chatbot has taken the internet by storm. Humans are asking it questions, and it’s sending answers back that are eerily lifelike, chatty, sometimes humorous and at other times unsettling and problematic.

The technology, made by the same organization that launched the text-to-image generator Dall-E to the public, has already amassed over a million users since it was released last week, company officials said.

It is powered by a large language model, an AI system trained to predict the next word in a sentence by ingesting massive amounts of text from the internet and finding patterns through trial and error. ChatGPT was then refined using feedback from humans to hold a conversation – as well as a robot in 2022 could reasonably do so, that is.

OpenAI, an organization launched several years ago with funding from Elon Musk and others, warns that ChatGPT isn’t perfect and will sometimes give offensive or misleading answers. But that hasn’t stopped social media users from asking it creative questions and posting the results online.

So what exactly is ChatGPT and how does it work?

Q: What is ChatGPT?

A: ChatGPT is a project from the OpenAI organization and an attempt to make “talking” with an AI feel smooth and natural, almost like a conversation between humans would.

Users can ask the AI questions in an instant-message like feature, probing the knowledge collected for the model. ChatGPT will answer the questions in full sentences, trying to mimic the rhythm of a conversation. But be warned that its answers are not always correct, or appropriate.

ChatGPT doesn’t have a personality as such, instead it can combine knowledge collected from many sources, coupled with training from humans.

“As a machine learning model, I do not have feelings,” the model’s answer will tell you if you ask how it’s feeling. “I am designed to process and generate text based on the inputs that I receive.”

The model – which the company’s CEO said in a tweet costs an “eye-watering” amount to run – can get stuck on certain phrases, including reminding you that it is what it says it is.

“The model is often excessively verbose and overuses certain phrases, such as restating that it’s a language model trained by OpenAI,” the company writes online.

OpenAI’s philosophy is to release these models into the wild before all the guardrails are in place, in the hopes that feedback from users will help the company find and address harms based on interaction in the real world.

Q: What are examples of how people are using ChatGPT?

A: People are using the chatbot in seemingly endless ways.

One user asked the bot to write song lyrics about AI taking jobs, in the style of rapper Eminem. “AI taking over, making humans obsolete; I gotta get angry, gotta spit fire and heat,” was one line of the tool’s attempt at imitating the multi-Grammy award winning artist.

Another asked ChatGPT to write scenes that would fit into the NBC sitcom “The Office,” tailored to ongoing storylines such as a prank war between the characters Jim Halpert and Dwight Schrute.

Many have also asked it to write computer code in programming languages such as Python.

Others asked the chatbot to write college style essays.

When one Washington Post reporter asked the chatbot to write a headline for this article, it returned three suggestions.

“GPT-3 chatbot revolutionizes the way we communicate with AI,” the chatbot wrote. Another was “Meet the GPT-3 chatbot: the smartest AI on the planet,” while the final said “GPT-3 chatbot leaves humans in the dust with its advanced conversational abilities.”

Q: How does it work?

A: At its core, ChatGPT is a large language model, which uses algorithms to analyze a massive corpus of text, often scraped from the internet, to respond to user requests in language that can sound surprisingly human.

ChatGPT is refined using a tactic called reinforcement learning from human feedback, according to OpenAI. The company enlisted “human AI trainers” to converse with the model, playing both user and chatbot.

Trainers compared responses written by ChatGPT to humans and ranked their quality, to help reinforce humanlike conversation styles.

Since OpenAI first released its groundbreaking text generating model, called GPT-2, the company has continued to experiment and iterate on the system. ChatGPT is similar to InstructGPT, a version of the model that is better at responding to commands from the human typing in a prompt.

Q: Who created ChatGPT and OpenAI?

A: The language model was created by OpenAI, a San Francisco research lab that was founded in 2015 with a combined $1 billion pledge from chief executive Sam Altman, Musk, billionaire venture capitalist Peter Thiel and others.

OpenAI’s technology went viral earlier this year when it opened its Dall-E image generator for anyone’s use. The often realistic photos stunned many with the apparent advancement of AI technologies. OpenAI uses a model it developed known as GPT-3 and later iterations to power its projects, and it is expected to launch a successor to the model in the near future, though it’s unclear exactly when.

Musk parted ways with the organization in 2018. After ChatGPT started spreading widely on Twitter, Musk – who bought Twitter for $44 billion this fall – tweeted that he had learned OpenAI had access to Twitter’s database for training data.

“I put that on pause for now,” he tweeted on Sunday. “Need to understand more about governance structure & revenue plans going forward.”

Q: How accurate is it?

A: A disclaimer pops up when you start to use the technology, warning users that ChatGPT is not always accurate.

“While we have safeguards in place, the system may occasionally generate incorrect or misleading information and produce offensive or biased content,” it reads. “It is not intended to give advice.”

The model can write answers that can seem plausible, but contain errors under closer scrutiny. The frequency of falsehoods, gibberish, or minor errors have compounded doubts about how soon this type of AI could be relied on without human oversight.

Others note the chatbot can have racial and sexual biases, while some pointed out that it gets basic facts wrong.

Q: How can I use ChatGPT?

A: OpenAI is letting people try out the technology by creating an account with the organization and reading through a few simple disclaimers.

The technology notifies users that it may use generated results to better train the technology, and warns users not to enter sensitive information.

This article was written by reporters, not the ChatGPT tool.