Juan Andres Guerrero-Saade specialises in disassembling malicious software to determine how it attacks computers.
It’s a relatively unknown cybersecurity field, which is why he hosted a weeklong seminar at Johns Hopkins University last month to teach students the difficult practise of reverse engineering malware.
Several of the students had little to no coding experience, but he was confident that a new tool would make it easier: he told them to sign up for ChatGPT.
“Programming languages are languages,” Guerrero-Saade, a Johns Hopkins adjunct lecturer, said of the ChatGPT software. “As a result, it’s become an incredible tool for prototyping things and getting very quick, boilerplate code.”
ChatGPT went public in November and quickly drew millions of users who marvelled at its uncanny ability to mimic nearly any writing style, from Seinfeld scripts and limericks to religious texts and Shakespearean sonnets.
While much has been made of its potential to disrupt writing jobs, some computer scientists are now wondering if its most immediate impact will be on people whose jobs were once thought to be “futureproof.” There are already videos on YouTube and TikTok of people showing how they’ve discovered ways to have ChatGPT perform tasks that previously required a significant amount of coding ability, such as building entire websites or scraping information from the internet.
“English is the hottest new programming language,” Adrej Karpathy, a former senior director of artificial intelligence at Tesla and a founding member of OpenAI, tweeted.
ChatGPT’s ability to mimic a specific author or style stems from developers training it on freely available and public information available on the internet, which includes vast repositories of published computer code and discussions on how to troubleshoot it. This provides a solid foundation for ChatGPT and GitHub Copilot, a similar programme designed specifically for coding, according to Grady Booch, IBM’s chief scientist for software engineering.
“They’ve got an open book — they’ve got the internet,” Booch explained. “They’ve most likely discovered answers to questions that have already been answered. As a result, it becomes easier and faster.”
This will not put professional programmers out of work in the near future, but it will speed them up, according to Booch. Even before ChatGPT, coders who ran into a problem would frequently use Google to find a solution.
“It doesn’t change the way I do business. But it does speed things up for me,” he explained. “It’s not a game changer. It’s a natural progression.”
David Yue and two other engineers won a competition in San Francisco last week against around 300 other programmers to create the most interesting AI software programme. The chatbot was used in his team’s project, “GPT is all you need for backend,” to automatically build some of the necessary but not particularly unique parts of how apps work.
Yue stated that while software engineers have been developing such tools for years, the speed with which they have recently taken off has surprised him.
“I don’t think there was any doubt about it. “However, the rate at which it happened is quite surprising,” he said.
ChatGPT and related technologies are not without flaws. They can introduce coding errors, and some have questioned the security of the code they generate. However, as long as they have human minders with some programming knowledge, this may not be a major issue. Siddharth Garg, a computer engineering professor at New York University, said he and his colleagues recently completed a first-of-its-kind study in which he assigned coding assignments to groups of students but only allowed some of them to use ChatGPT or Copilot for assistance.
“We didn’t see a significant difference in the incidence of security bugs between human-generated code and code generated by Copilot or ChatGPT,” Garg said.
“There are security bugs, but humans also produce security bugs. We didn’t notice a significant difference, at least.”
What does this mean for the many people who learned to code in the hope of finding a lucrative job? Not everyone is gloomy about the future.
“Generative AI can automatically generate code, making it easier to create software and amplifying the power of a software engineer,” wrote Hadi Partovi, CEO of the nonprofit tech education organisation Code.org, in a lengthy Twitter thread about the topic. “As a result, the creation of (and demand for) software will accelerate, and more people will become software engineers,” he concluded.