By now, you've most likely heard of ChatGPT, but what exactly is it and how does it work?
GPT stands for “Generative Pre-trained Transformer,” which isn’t always clear to everyone, so Kent State Today asked ChatGPT to explain what that meant.
This is how it responded:
“It's a type of artificial intelligence language model developed by OpenAI. The name breaks down as follows:
- Generative: The model is capable of generating new content, such as text.
- Pre-trained: The model is initially trained on a large dataset before being fine-tuned for specific tasks.
- Transformer: Refers to the transformer architecture, a type of neural network architecture that has proven effective for a variety of natural language processing tasks."
ChatGPT, which is celebrating its one-year anniversary this month, has many uses: translating languages, making blog posts, writing code, brainstorming ideas, writing movie scripts, answering writer’s tech questions, and even telling knock-knock jokes. ChatGPT currently has more than 180 million users as of August 2023, according to Reuters.
So, how does this chatbot work, and what are the potential impacts of its use in academia? How could its increasing utlization affect future grads and their careers?
In layman's terms, ChatGPT is an artificial intelligence text chatbot made by the company OpenAI that was released as a free web-based tool on Nov. 30, 2022. It is trained on massive amounts of text and data from the internet such as books, articles, websites and social media to understand information, use grammar and apply reasoning.
Javed Khan, Ph.D., professor and chair of the Department of Computer Science, Ruoming Jin, Ph.D., professor of computer science and Mikhail Nesterenko, Ph.D., professor of computer science, hosted “The Power of ChatGPT” discussion in September to explore what ChatGPT is, does and will do.
In the presentation, Jin described ChatGPT as a large language model that has trillions of parameters to create an artificial neural network. Much like the neural pathways in a human brain, this network is what enables ChatGPT to showcase emergent properties, which are responses that it was not specifically programmed to do.
“It has started to show emergent properties, like sophisticated behavior or response,” Jin said, “like helping us program, which was not on purpose.”
During the event, Jin explained that what is put into ChatGPT is broken down into smaller units called tokens. The network processes these tokens to understand the context, which helps it learn patterns and associations.
ChatGPT not only considers the meaning of your input – for example the directive to “write me a winter poem” – but also the context of the entire conversation. Adding to your winter poem request that you would like it to include a stanza about driving conditions, ChatGPT will understand and rewrite the poem accordingly.
Because of how it scans the web, ChatGPT, like other AI models, can be prone to algorithmic bias, or negative outcomes related to the data set from which each AI model pulls, most of which are proprietary, according to the Associated Press style guide.
It's also not perfect. For instance, the images in the banner at the top of this story were created using the free OpenAI image creator on Bing with the prompt: "Create a banner image for a news story about ChatGPT and AI." As you can see, AI is spelled wrong in two places.
Still, ChatGPT can do so many things. But what does that mean for the future workforce? Will we all be replaced by AI? Khan does not think so.
“Those working in language fields will be affected, but they will most likely see it in the significant changes that happen to their tasks,” Khan said.
There may even be new jobs emerging from this technological advancement. An example would be what Khan described as a “knowledge encoder,” someone who encodes information such as the ways of working and other tasks into machine-processable information.
Khan compared how ChatGPT can affect the workforce to how the invention of the calculator did not replace the need for mathematicians, it only changed their workload.
Though AI is relatively new, Khan explained that humans naturally adapt to new inventions and that we can adapt to AI like how we adapted to the light bulb.
“The span of technological changes is becoming shorter and shorter,” Khan said, “The internet changes fast and AI will likely make changes even faster. We need to have extra antennas up to learn new things and how to do things differently.”