Can Generative AI & No-Code Coexist
7th November 2024
In this article we look at the question, can generative AI & no-code coexist?
As the founder & CEO of a no-code startup, investors have been asking me the question, “can generative AI & no-code coexist” since the release of the generative AI model ChatGPT on November 30th, 2022.
Because of the frequency of this question, I thought it was worthwhile to write a quick note on why we at Dittofi think that generative AI models such as ChatGPT & Google Bard will coexist with both code & no-code platforms in relatively predicable ways for the foreseeable future.
Generative AI & no-code will coexist
Generative AI models such as ChatGPT are what are known as Large Language Models (or LLMs). They work by learning the statistical structure of language by optimizing their ability to predict missing words in sentences. The video by Steve Seitz published on YouTube explains how LLMs actually work.
Although LLMs are very sophisticated, as you have seen, they are ultimately search engines. Like all search engines, LLMs can only return data that is included in its database. Therefore, its answers are limited to what it has seen before. For anything truly original, the LLM cannot generate a solution &, in this respect there is always the possibility to develop something original either with code or with no-code.
Whilst traditional no-code tools are also limited in their abilities to produce totally custom apps, there is a new class of hybrid no-code tools such as Dittofi. Hybrid no-code tools seek to provide a vastly more powerful visual development experience that is 100% analogous to coding. Because of this, hybrid no-code tools already integrate very well LLM models such as ChatGPT. Furthermore, hybrid no-code tools can be used to solve some very complex coding challenges with the use of visual programming that traditional no-code tools cannot.
How generative AI & no-code will coexist
In the short to medium term we believe that generative AI systems will be used to enhance the experience of developing an app for both traditional software developers & for no-code developers.
Today, generative AI is already enabling both coders & no-coders to build increasingly complex & better quality apps faster than ever before. The way that this looks in the world of code is with generative AI plugins, such as Githubs co-pilot that acts as an AI assistant to help coders write code faster.
The way coders & no-coders use generative AI today is exactly the same. The only difference is the method by which they test & deploy their generated AI. To demostrate this, below are two examples of how the coders & the no-coders workflows look in practice.
Example 1: The Coders Generative AI x Code Workflow
In the first example you can see a developer writing a series of prompts to ChatGPT. These prompts are designed to get help from ChatGPT to build parts of his app. The coder then takes the generated code & uses his integrated development environment (or IDE) to integrate this generated code into their app. In this example the generated code is integrated entirely with code.
Example 2: The No-Coders Generative AI x No-Code Workflow
In the second example, there is a no-code developer writing a series of ChatGPT prompts to generate SQL code for their app. In this example, the no-code developer is able to generate a database schema for their app. The no-code developer then takes this SQL code & uses their no-code development environment to integrate the code into their app. The no-code tool used in this case is the Dittofi hybrid no-code platform.
Notice that in both of the above cases, humans are still required to check, test, modify, debug & deploy the code produced by generative AI. These tasks represent another suite of activities that LLMs cannot by default do & could be a place where no-code platforms sharpen a competitive edge in the future for example, in the development of better quality parsers.
In conclusion
We think that generative AI systems & no-code platforms will continue to coexist. This is because (1) LLMs are limited by their data sets in what they can develop & (2) because LLMs are unable to check, test, modify, debug & deploy the code that they generate. In regards to point two, this could imply that no-code platforms will eventually become really good at parsing responses from LLMs.
We have already seen LLM tools such as ChatGPT significantly improve the app development experience for coders & no-coders & this is a trend we think will continue into the future. We also see that LLMs are already starting to integrate into our technology stacks & everyday user workflows. For example, OpenAI has released their ChatGPT marketplace.
At Dittofi we plan to stick close to generative AI solutions, as we aim to leverage these technologies to enable no-coders to build increasingly complex & better quality solutions, faster than ever before.