AI improves programming efficiency, but generating too much code too quickly is not a good thing

News on June 1st, although generative artificial intelligence technology has improved efficiency and made software development easier, the head of the technical department is worried that generating a large amount of code too quickly will lead to subsequent problems, making their jobs more difficult.

The following is the translation:

Generative AI programming tools can greatly improve development efficiency, but some technology executives worry that generating too much code too quickly may have negative consequences.

IT heads at major companies including United Airlines, Johnson & Johnson, Visa, Cardinal Health and Goldman Sachs have said they are excited about the potential of generative artificial intelligence to automate some of the programming process and expect This can significantly improve work efficiency.

However, some IT executives worry that lowering the bar for code development could lead to increasing complexity, "technical debt" and confusion as they manage an ever-expanding stack of software products. "Technical debt" refers to developers making compromises in a short period of time in order to develop quickly, which ultimately creates additional burdens in the future.

Tracy Daniels, chief data officer at financial services company Truist, said that as deliveries accelerate, there is a potential risk of increasing "technical debt" and "orphan code" Worth paying attention to.

"People have been talking about 'technical debt' for a long time, and now we have a whole new credit card that can be used to accumulate 'technical debt' in ways that were not possible before," MIT Computing said Armando Solar-Lezama, a professor in the Science and Artificial Intelligence Laboratory. He added: "I think it's possible to accumulate a lot of bad machine-written code. Companies should rethink how they interact with new tools to avoid a similar situation happening again.

According to Saul Lezama , some code development processes are very tedious and time-consuming, and people have been trying to automate these development efforts for many years. The increasing size and accuracy of automatically generated artificial intelligence models has promoted the progress of automatic programming. In turn, this has also Promoted the popularity of chatbots such as ChatGPT.

Amanda Silver, Microsoft vice president and product leader of the development department, said that the shortage of excellent developers has also prompted the company to increase its investment in development tools. Investment.

Different companies are evaluating and deploying various tools, such as Microsoft-owned Github’s Copilot, and other tools launched by Amazon, IBM and startups Tabnine and Magic AI. These tools can often recommend generated code snippets and Conduct testing, or make technical recommendations in writing programs. But IT department leaders say there are risks.

EXL is a data analytics and digital operations solutions company. Vivienne, executive vice president and head of analytics Vivek Jetley said: "Even if it makes it easier for programmers, I think it makes the job of the chief information officer more complicated. "These tools do make programming easier so more and more employees can start writing code for new application scenarios. But as the amount of code explodes, CIOs This code needs to be controlled and managed, prioritizing which code to keep, which to throw away and how to run the system.

"There will definitely be more chaos," said Jaitley.

OutSystems CEO Paul Paulo Rosado (Paulo Rosado) said that "technical debt" and "orphan code" have long been problems that have plagued CIOs.

As the amount of code continues to increase, people will inevitably confuse some What the code does and how it was created. These problems are compounded when developers leave the company. Over time, the pile of code grows. Rosado is convinced that generative AI programming tools will exacerbate these problems.

These risks are real, according to Jason Birnbaum, chief information officer of United Airlines. Over time, the resilient design and security of cloud environments will become more critical , it will also become more challenging to release software that has not been properly vetted and tested.

Despite the risks, CIOs are moving forward. Birnbaum said United is Testing several generative AI

applications

, including the ability to automatically generate code. Healthcare company Cardinal Health also recently formed a cross-functional working group to evaluate related use cases and risks. Truist is working with vendors Explore new code generation and code annotation tools together. Goldman Sachs’ early pilot project efficiency improvements have reached double digits. (Chenchen)

The above is the detailed content of AI improves programming efficiency, but generating too much code too quickly is not a good thing. For more information, please follow other related articles on the PHP Chinese website!