5 unfounded fears around AI-enabled code generation
Scott TumSuden
Vice President @ Cognizant | Chief Revenue Officer | Chief Product Officer | Chief Digital Officer | Growth Leader | Strategy & Transformation | P&L Management | Healthcare | Retail | Manufacturing | Energy | High-Tech
When it comes to GenAI, there’s the hype cycle and then there’s the fear spiral. As an IT leader, it’s important not to fall into the trap of either.
When it comes to managing the latter, we need to remember two things: 1. GenAI is an undeniable part of our future. 2. Not all use cases carry the same amount of risk.
Using GenAI to assist in coding and development is one of the best examples of a low-risk application: the use case is well-developed; the output is relatively straight-forward; and the organization has existing processes and protocols in place to evaluate those outputs with respect to performance, scalability and security.
It also happens to be a high-value use case. This technology can help automate routine tasks, like refactoring code and writing code documentation, generating efficiency gains of 20% to 50%.
And yet, many organizations remain hesitant to incorporate GenAI tools within their development processes. In this post we explore the underlying reasons for this reluctance and explain why, when it comes to coding, GenAI shouldn’t be a matter of concern but a source of competitive advantage.
?
Fear #1: The code won’t work.
At the bottom of the ChatGPT application there’s a small, but significant, note: ChatGPT can make mistakes. Consider checking important information.
When it comes to writing code, it’s this fear of mistakes that often holds some enterprise IT teams back. But how valid is this concern—especially when you consider that human developers can also make mistakes? Not very.
First of all, AI-enabled tools cannot yet autonomously produce code. Second, even if they could, that code should not be deployed without going through all the standard production steps, including review processes, testing, applying security controls, promotion processes, and approvals. Using a GenAI tool does not circumvent that process—it merely streamlines some of the research and legwork a developer does when coding.
So long as the organization has a robust process in place, the risk of a GenAI-enabled tool producing “bad” code that is inadvertently deployed isn’t any higher than publishing “bad” code written by a person.
?
Fear #2: GenAI will replace humans.
Whenever a new technology is introduced, people hyper fixate on its potential to replace jobs. That’s a natural fear, though perhaps not as common a scenario as we think.
The reality is that technology has traditionally augmented jobs rather than entirely displacing them. Excel has not eliminated the need for accountants any more than GenAI will make human developers obsolete.
At the same time, GenAI can improve the way people work, by making them more efficient and effective. When it comes to coding, it can automate repetitive and low-level tasks; it can streamline research processes; and it can help build skills naturally.
We should think of GenAI as an assistant—one that is trained extensively in every language and every platform. It’s a co-pilot that has encyclopedic knowledge that could answer questions and validate responses, saving developers at all levels countless hours of research and review.
At the end of the day, developers and IT teams should not fear GenAI any more than they fear GitHub. This tool isn’t giving an organization code, it’s giving their developers guidance and support.
?
领英推荐
Fear #3: We’ll give a leg up to our competitors.
One of the prevailing concerns at the enterprise level regarding public GenAI tools revolves around the inadvertent disclosure of sensitive information to competitors. But again, this fear is largely unfounded, especially when it comes to development.
Practically speaking, any reputable hyperscaler will work with clients to determine how to sandbox sensitive or proprietary data. Most of them offer a guarantee that this data won’t be re-ingested into the LLM to enhance the model’s intelligence. This significantly lowers the risk of critical code being leaked.
For the sake of argument, even if a snippet of code was fed into the LLM and discoverable by another user, IT leaders need to be realistic about how much value that really holds on its own. In most cases, a piece of code is not the source of a competitive advantage; the real value lies in how you stitch the code together with the broader system architecture and experience. In other words, the business process typically dictates how code needs to be written, not the other way around.
Further, a snippet of code is but a tiny fragment within a huge complex business process. It represents one specific action for one use case at a select point in time for that company. That information does not provide a window into the broader IT environment and business strategy, which means that even if someone may stumble upon it, it’s very unlikely that they can use it to infer any unique or differentiated aspects of the business model.
?
Fear #4: Any mistake could lead to serious legal ramifications, like a lawsuit.
When it comes to development, there isn’t much to suggest a high risk of legal ramifications for sourcing code that belongs to another party. Even if one company could prove that another organization was using a portion of their code and that it originated from an AI-enabled tool, there’s no equivalent of plagiarism laws in place around development.
Further, depending on what hyperscaler the company engaged, it’s possible that the business may be indemnified against such issues anyway. For example, Microsoft offers a Customer Copyright Commitment which pledges to “defend” Azure OpenAI Service users and take financial responsibility for any legal issues related to copyright infringement from Azure OpenAI Service outputs.
Finally, we need to keep in mind that many developers source code from online forums; they compare techniques with their network; they bring experience with them from one job to the next. Coding is as fluid as any other function and while AI tools may add a new wrinkle into how code makes its way around development world, it is certainly not the only way for snippets to be shared.
?
Fear #5: We don’t know what we don’t know.
In the world of evolving technology, there will always be the fear of the unknown. And when it comes to GenAI I’ll admit that there’s a lot that we don’t yet know. But, at the same time, is that a good reason to avoid doing something new? The technologist in me says no.
The consultant side of me also wants to remind people of the risk of inaction. If companies wait for all the laws and rules to be figured out before they event start to explore GenAI, will they ever be able to catch up? In the case of GenAI there’s a lot to be gained by moving fast and being first, but there’s even more to be lost by delaying and ignoring.
In a world of what ifs, we need to focus on what we do know: GenAI took the world by storm a little over a year and a half ago. The meteoric growth we’ve seen to date establishes that it will be part of our future. Software development offers a fantastic starting point for most enterprise GenAI journeys in that it is a mature, high-value use case that provides a relatively controlled entry point for companies to experiment with the technology, build skills, figure out their guardrails, develop a path for scaling initiatives and identify a cost model.
The GenAI landscape is sure to change—but that’s the nature of technology. IT leaders have a clear role to play in helping other members of the C-Suite and boards understand that, when done right, the rewards can far outweigh the risk.
?
Unlocking the value of GenAI in development
In a recent study by Cognizant, our research team forecasted that generative AI would drive up to $1 trillion in productivity gains by 2032. That’s an incredible source of value that companies should not ignore. ?
At the same time, IT leaders need to be thoughtful about how they use this technology. They must define the boundaries of appropriate use to simultaneously unlock the value of GenAI while also protecting the organization from risk, runaway costs, and traditional change management issues.
For more information on how companies can adapt their traditional technology roll-out processes to harness the power of generative AI in a safe, secure and effective way, please read my related article: A practical guide to introducing gen AI into the enterprise. I'd also love the opportunity to discuss this topic in more detail with you at any time. Do not hesitate to comment below or reach out to me or any of the 高知特 Cognizant team at any time to discuss your journey.
Senior Director - Process Improvement and RPA CoE at Cognizant
7 个月Well said Scott TumSuden. Thanks for the thoughtful insights.
CEO at Cognitive.Ai | Building Next-Generation AI Services | Available for Podcast Interviews | Partnering with Top-Tier Brands to Shape the Future
7 个月Exciting insights! Can't wait to dive into the benefits of GenAI. Scott TumSuden