Understanding Generative AI Technologies
As we rapidly approach a world where generative AI tools like ChatGPT, Dall-E, and others play a significant role in marketers’ day-to-day work, it’s essential for organizations to make a smooth transition. To achieve this, it’s important to have a solid understanding of how these technologies work, their capabilities, and their limitations. This knowledge can help businesses deploy generative AI effectively and avoid costly errors or embarrassing customer experiences.
AI-Generated Code: The Mechanism
Generative AI large language models (LLMs) write code similarly to how they write text – by calculating probabilities. However, while tools like ChatGPT and Bard are trained on literature, articles, and social media content, LLMs for coding, such as GitHub Copilot, are trained on billions of lines of publicly available source code. This enables them to recognize patterns in coding and string code elements together by identifying the next-most-likely code element to include.
Six Modes of Using Generative AI for Code
Generative AI for coding has a wide array of use cases, with six common code-writing applications that marketers can leverage:
- Writing full code of projects: Best suited for small, simple projects like basic website landing pages.
- Writing pieces of code: For larger projects, breaking them up into functional blocks is more effective.
- Debugging code: Generative AI can help fix code that isn’t working as intended or even perform live debugging, saving valuable time.
- Refactoring code: AI can compress and optimize existing code, making it more efficient and lightweight.
- Asking for coding advice: Broad LLMs, like ChatGPT, can provide helpful advice on coding or debugging.
- Documenting code: AI can document code written by anyone, facilitating future code changes.
Being specific in coding requests can significantly improve the accuracy of generated results. Just as expertise in graphic design helps users get more out of image-generating tools, being an experienced coder helps users get more out of generative AI coding tools.
Companies Adopting Generative AI for Coding
According to a Bain & Company survey of nearly 600 companies across 11 industries, 46% are already using generative AI for code completion, generation, and copiloting. However, some brands are taking ill-advised risks.
Confidential and Non-Confidential Code
Using proprietary code in generative AI tools can be dangerous because it is unclear what happens to that information. Notable cases, such as Samsung employees pasting proprietary code into ChatGPT, highlight the potential risks. Cyberhaven research reveals that over 4% of employees have put sensitive corporate data into LLMs.
It’s also worth noting that LLMs trained on publicly available code are less likely to be helpful in generating proprietary code, as they lack exposure to an organization’s internal libraries and resources.
However, for code that isn’t confidential or sensitive, generative AI can be beneficial. Examples include website code and email code, which are already visible to the public.
Website and Email Coding
Currently, generative AI is better suited for writing website code than email code for several reasons:
– There are more large, public repositories of web code than email code available for AI training.
– HTML coding for websites has well-established standards set by the World Wide Web Consortium, while email coding lacks official standards.
– Web coding is discussed more extensively on the public web than email coding.
These factors make it challenging for generative AI to write complete email code, as demonstrated by The Email Factory. However, it can still be useful for writing pieces of code, debugging, and providing coding advice when given specific prompts.
Future Developments in Generative AI Code-Writing Tools
In the coming months and years, we can expect generative AI code-writing tools to improve significantly as they are trained on larger amounts of content and their interfaces allow for more brand-tailored responses. This progress should enable skilled coders to become even more productive, further enhancing the organization’s ability to leverage generative AI capabilities in their marketing efforts.