Tech takes dark turn with ChatGPT

Asking a computer any question imaginable and getting an answer is now possible with certain black box artificial intelligence models.These models go through iterative processes of development, making the internal processes of how it works a mystery, yet a plausible output is always reached. 

AI is used in a variety of industries, from the financial sector to manufacturing environments, because of its ability to complete tasks faster and more efficiently than humans, generally making less errors. In line with this, rapidly advancing AI technology will change the way that information technology classes function for those who are aware of its effects. 

At the East Cooper Center for Advanced Studies, Neil Sullivan teaches cyber security and networking classes which are sensitive to technological changes. Students are able to utilize AI for probable answers to assignments, forcing teachers to change their grading practices or coursework. 

“It’s difficult to guard against [cheating through AI]. People aren’t understanding what they are losing. It’s a shortcut that’s degrading from your education or your ability to think,” Sullivan said. 

ChatGPT is a black box AI model created by OpenAI which features complex neural networks that use interconnected nodes to replicate biological networks and unsupervised learning, with version 3.5 being released on Nov. 30, 2022. Most models rely on deep learning and natural language processing. 

ChatGPT is becoming more popular because of its superior language processing ability, or way to distinguish what a human wants from their words. Users have been able to get better quality work out of it when using specific and clear prompts, such as meal plans and essays. 

“It can always be helpful, it’s a phenomenal tool, but any tool can be misused. You have to make sure you’re doing it the right way. It’s a great tool just to help your imagination…If you have somebody that reads what it comes up with and expands it and uses it as a starting point, it’s a wonderful way to work. It gives you an idea. If you’re using it that way, it’s incredible. If not, it could be brutally detrimental,” Sullivan said. 

One issue created from ChatGPT and similar models is the way that they obtain information. Black box AI models take in a range of inputs and parameters from the internet when it was created, but often this makes the developers unable to explain how a model works. It is difficult to verify if the information outputted from the AI model is factually correct or if it’s completely original work as ChatGPT compiles work from a variety of internet sources, such as Wikipedia. 

Senior Ben Patrick has heard about recent developments in the field of AI and machine learning, but isn’t concerned about its effect in the education system. Every student has a unique voice, allowing teachers to easily recognize authors when reading submitted assignments. 

“You could be not articulate at all when you have a conversation with [the teacher] but with the AI, you sound like a completely different person,” Patrick said. 

AI models, like ChatGPT, are trying to become close to artificial general intelligence, which would display basic human capabilities such as common sense, abstraction, and the usage of background knowledge. Oftentimes, it is clear when AI models are used to produce work because it is missing key human aspects, like one’s experiences and background.

Benjamin Rzyski teaches both AP Seminar and AP English Literature and Composition. These classes are known for the intensive essays and analysis meant to push students to go deeper into a variety of content and issues. AI models will be able to provide students with analysis for these topics instead of students practicing these skills themselves. 

“I think students are creative in a variety of ways, even by incorporating technology into not just schoolwork but in their daily lives. I don’t think it’s an issue of creativity, or any creativity that will be lost. I do think some critical thinking may be lost and the ability to focus on abstract concepts for a particular length of time may be lost,” Rzyski said. 

Technology will continue to be something that is utilized in education, with potential for it to be good and bad. It is difficult to predict where the future of AI will lead to because of the uncertainty of what advances in computing will bring.

“It’s not a pressing concern to me yet at the moment, but I know that it’s on the horizon and I think more changes because of technology are on the horizon. As a professional, I just need to prepare and to understand,” Rzyski said. 

Currently, there are many speculations on the future of AI, with many believing that there will be a large leap in its complexity and capabilities due to a large increase in the model’s number of parameters and inputs. 

Currently, ChatGPT-3.5 has 175 billion parameters, but CEO Sam Altman of OpenAI, the organization which created GPT models, declines commenting on the future abilities of next generation models. 

“You have to guard against technology. We’ve gone to such a rush to technology once COVID-19 hit and found out that it really didn’t work. It made it simple, but making it simple doesn’t make it better. It doesn’t make you educated,” Sullivan said.