In the select world of elite computer programmers, could artificial intelligence reach a place where it will write a significant amount of their code? For one such developer, it already has.

One of the founders of ChatGPT’s OpenAI, Andrej Karpathy, left in 2017 to lead the computer vision team for Tesla Autopilot. Karpathy recently rejoined OpenAI and described his use of the Copilot AI-based coding tool in a December tweet.

“Copilot has dramatically accelerated my coding. It’s hard to imagine going back to ‘manual coding,’” Karpathy said. “Still learning to use it, but it already writes ~80% of my code, ~80% accuracy. I don’t even really code, I prompt & edit.”

This feature is an exploration of the themes emerging from theCUBE’s ongoing event coverage of KubeCon and Red Hat Summit.

Expanding use cases

Such a statement might have been unthinkable even three years ago, yet rapid advances in large language models such as ChatGPT have led to the emergence of powerful new AI-based tools for writing code. One of the most significant of these is Microsoft Corp.’s Copilot. Microsoft’s GitHub unit first debuted Copilot two years ago as a tool that used OpenAI’s GPT-3 algorithm to automatically generate software code. GitHub launched Copilot for Business in February and introduced Copilot X a month later.

The latest iteration of Copilot is powered partly by GPT-4 and includes an array of new programming features. Developers can now issue instructions by voice, scan for bugs and speed up the process for creating pull requests that explain how the code was created in the first place.

How is Copilot being used in the real world? One Microsoft programmer described using it to create a network security group for opening a port in Azure and generating inline documentation for his technical work. Another developer described how he used Copilot to write file parsers for extracting data and adding new UI components in an HTML page. Because Copilot is contextual, it can quickly sense what a developer is seeking to do.

One programmer needed to create mock sports data using elite college basketball teams as a model. When Copilot was given the task, it generated similar data for football, also using elite teams.

“It’s removing the very toil that sapped developers’ productivity and energy,” said Arun Batchu, vice president and analyst in the Software Engineering practice at Gartner Inc., in an interview for this story. “They are happier because they can pursue the creative work they signed up for.”

Gaming signs on

One of the areas where ChatGPT is beginning to have a transformational effect is in the gaming industry. Developers of video games have been major AI users for some time now, and the emergence of ChatGPT has created a flurry of new activity.

Developers have experimented with the use of ChatGPT to recreate existing games such as Snake and the 3D-based Doom. Game maker Inverse Inc. has been using generative AI to produce dialogue for a dating simulation called “Love in the Classroom.” Inverse programmers have noted that the use of ChatGPT has allowed them to input unique questions and receive unique answers from digitized characters.

Major game studios are getting involved as well. Ubisoft Inc. revealed its own generative AI tool in March that creates phrases or sounds made by non-player characters in response to an in-game event.

Ubisoft has been careful to position its new tool, called Ghostwriter, as a positive enhancement for taking away mundane tasks from developers. Yet in the aftermath of Ghostwriter’s unveiling, some game developers are worried that advances in generative AI could cost them their jobs.

Gartner’s Batchu sees basic cut-and-paste programming work being assumed by generative AI models at the very least.

“Those tasks are getting automated,” Batchu said. “The skill levels are shifting to those developers who understand how to tease out the best code from these AIs.”

Marketing potential

While generative AI may be poised to replace some developer tasks, it could also create new opportunities from the growing market that surrounds it. The rise of generative AI raises the potential for change in how business-to-business and business-to-consumer sales and marketing will be conducted.

Personalized marketing has long been the holy grail for advertisers, and generative AI now offers a potent tool for making this happen. A survey of commercial leaders by McKinsey & Co., found that more than half believed generative AI would have significant or very significant impact on lead identification, marketing optimization and personalized outreach.

In addition, the growth of AI-based marketplaces, such as GenesisAI, could create new opportunities for developers as interest in the technology continues to blossom. GenesisAI, which provides an open marketplace for AI applications, claims 2,000 registered users with over 45 AI models deployed.

Supercloud and security

In addition to a growing market around generative AI, there is an opportunity for ChatGPT and other tools to be employed in data-driven areas of platform engineering. A specific example of this can be seen in a rise of the supercloud, an abstraction layer that resides above and across hyperscale infrastructure.

As companies such as Snowflake, Databricks and MongoDB continue to build major businesses on top of cloud platforms, this emerging infrastructure will require new thinking about cybersecurity. Generative AI will play an important role in this, and researchers at Tenable Inc. have been testing the application of large language model tools to improve cloud security and protect against potential threats.

Tenable’s work includes the development of EscalateGPT, a Python tool designed to identify privilege escalation opportunities in AWS cloud environments. The tool can retrieve identity and access management policies and prompt for weaknesses that could be exploited for privilege escalation by a malicious actor.

Tenable’s security researchers have also developed a translation script for Ghidra, a reverse engineering tool released into the public domain by the National Security Agency in 2019, along with a code debugger that answers questions about runtime state or assembly programs and an intercept tool for analysis of HTTP requests and responses.

Creative uses of ChatGPT by security researchers at Tenable points toward the flip side of this equation. The AI tool, which has been downloaded by over 100 million users in record time, is also in the hands of cybercriminals. As noted by security researchers at the RSA Conference in San Francisco this spring, while there have been no signs that generative AI is being widely used in attack vectors, malicious actors have been leveraging AI tools for years.

Despite this ominous reality, there remains a note of optimism within the security community that generative AI will ultimately enhance network protection. Being able to ask questions of or prompt ChatGPT to deliver the right results could ultimately have enormous value. During the RSA gathering, Brian Spanswick, chief information security officer and head of IT at Cohesity Inc., spoke at an event hosted by Tenable and expressed a belief that generative AI will make a positive impact on the tech industry.

“It’s going to do so much more good than harm,” Spanswick said. “The trick is getting the question right.”

Image: freepik commons – kenshinstock