Generative AI and No-Code — A Match Made in Heaven

Is Generative AI Changing the World?

ChatGPT, the popular chatbot from OpenAI, is one of the fastest growing consumer applications in history. According to a UBS study, it reached 100 million monthly users just two months after launch in November 2022.

ChatGPT: Fastest growing consumer application in history

Artificial intelligence (AI) has long been a futuristic concept that has been envisaged in popular culture and science fiction chronicles such as The Matrix and The Terminator. However, AI is no longer just mere fantasy, as advancements in computing technology have seen AI being incorporated into our daily lives — self-driving Tesla cars, real-time language translations using Google Translate, Amazon Alexa virtual assistants, and many more.

The most recent technology revolution has seen the rise of chatbots like ChatGPT, which can conduct human-like conversations and has the ability to generate all kinds of content such as essays, emails, code, and even poetry. As indicated by its name, ChatGPT is based on a type of AI model called Generative Pre-trained Transformers (GPT).

What are GPT and Large Language Models?

Generative Pre-trained Transformers (GPT) are AI large language models (LLM) created by OpenAI based on the transformer architecture. An LLM is a type of machine learning model that is used for natural language processing tasks. Trained using massive amounts of data and computing power, LLMs can remember the patterns and structures of language and generate content based on prompts.

Introduced by Google in 2017, the transformer architecture is widely regarded as the major breakthrough that now powers the capabilities of most modern LLM. Though GPT is arguably the most prominent model now, it is not the only transformer-based LLM model available, as there are many alternatives such as Google’s LaMDA and Facebook’s LLaMA.

Infographic: The Rise of LLM Chatbots — Source: Analytics India Magazine

The advancements of these models and explosive user adoption have led to many competing services being launched, where Microsoft’s Bing Chat, Google’s Bard, and many others are following in the footsteps of ChatGPT. Research and competition in this area are only just heating up, and future advancements will undoubtedly result in more amazing capabilities in the years to come.

How to Use Generative AI for Application Development

Generative AI has the potential to radically alter the way humans interact with technology. Bill Gates, the founder of Microsoft, says that the GPT AI model is the most important advance in technology since the introduction of the graphical user interface (GUI) in 1980. Just as how the GUI empowered a whole generation of users, an AI natural language interface can change the way people work, learn, communicate and more.

In the context of application development, generative AI is now capable of generating code based on natural language prompts. Many language models have been trained on billions of lines of code, and can generate code in a variety of programming languages. In March 2023, OpenAI conducted a developer livestream demo of a GPT-4 model generating a basic HTML website from a hand-drawn sketch. Tools such as GitHub Copilot and Amazon CodeWhisperer are already used by developers all over the world.

Nevertheless, some critics malign LLMs as merely auto-complete on steroids, which may not be too far from the truth because these models are actually predicting output from inputs. As AI technology advances, the day may come when AI singularity happens, a hypothetical future point in time whereby artificial intelligence exceeds human intelligence. However, technology has not reached that stage yet, and current AI models tend to err and make up facts in a phenomenon termed hallucination.

Although there are many use cases where generative AI can perform wonders, it is currently a huge risk for developers and enterprises to rely entirely on code generation especially for non-trivial enterprise applications. Having to understand, test and maintain large amounts of code generated by AI, which may contain bugs or even complete hallucinations, may become a developer’s worst nightmare.

A typical enterprise application would contain thousands if not millions of lines of code. Now imagine placing trust on an AI to produce that amount of code, and it would be clear that it is simply not a practical alternative in real world scenarios.

Generative AI hallucinating solutions — Source: Reddit

Developers can use generative AI as an aid to kickstart development, generate snippets and improve overall productivity, but for now AI code generation is not a magic bullet to develop, deploy or maintain complete enterprise applications.

Why Generative AI Complements No-Code / Low-Code

There is, however, an area of application development where generative AI can be leveraged and optimized — low-code and no-code platforms (LCNC) platforms.

LCNC platforms are application platforms that allow developers to rapidly build applications visually, with zero or little code. There are several reasons why generative AI complements LCNC platforms for enterprise applications:

1. LCNC platforms use a declarative approach. A declarative approach means that the focus is on “what” the developer wants to do, rather than “how” to do it. This approach jives nicely with how a user interacts with generative AI. So instead of prompting the AI to generate low-level code to achieve certain behaviors (the “how”), the developer prompts the AI on what he wants (the “whats”), and the AI can produce the end result.

2. LCNC platforms are visual in nature, with a drag-and-drop approach and a WYSIWYG (What You See Is What You Get) environment. So having AI generate apps that are visually verifiable and can be immediately executable makes it easy to validate the AI output, making it much more compatible with generative AI compared to code generation.

3. Governance is a critical aspect to consider when deploying generative AI technology. Unlike code generated directly from LLMs, LCNC platforms with governance features ensure that enterprises can apply their business frameworks, rules and governance to the output. In addition, by presenting it in a logical visual format, it is far easier to maintain and customize the application using an intuitive visual development approach.

Visual declarative development using Joget DX no-code/low-code platform

Leading market research company Forrester coined the term TuringBots for generative AI that assists software development tasks. In their research “Will AI Kill The Low-Code Market?” Forrester also postulated that TuringBots will dramatically increase low-code adoption.

So rather than replacing LCNC platforms, generative AI is an ideal complement to harness its capabilities and fulfill its incredible potential.

Generative AI Capabilities in Joget DX

Joget DX is an open source no-code/low-code platform designed as a convergence of both no-code and low-code. Prior to the introduction of generative AI, Joget already has integrated AI capabilities such as TensorFlow integration for use cases such as image classification, object detection, face recognition, sentiment analysis, and more.

With the rise of generative AI, Joget is introducing a new range of generative AI features and integrations that enrich the user experience of the platform regardless of persona — whether application end-users, no-code developers, and low-code developers. As an open source platform that empowers anyone to rapidly build enterprise applications with built-in governance, these new generative AI capabilities will further increase the accessibility of the Joget platform more than ever.

Based on Joget’s extensible plugin architecture, these AI capabilities are plugins that can be seamlessly installed. Significantly, these plugins are themselves designed to be pluggable, such that the AI models used can be switched. The new generative AI innovation in Joget DX now seamlessly integrates large language models (LLM) such as ChatGPT to merge the simplicity of natural language with Joget’s well-established visual development platform, resulting in a significant acceleration in software development.

The following are the new generative AI capabilities introduced for Joget DX:

No-Code Developer: Generative AI App Generator

Joget’s App Generator greatly accelerates the building of an app by allowing administrators to rapidly generate a complete base app directly from a form based on pluggable templates. With the new Generative AI App Generator, anyone can now use natural language prompts to simply describe the type of application they need, and the application will be instantly generated with comprehensive enterprise features including forms, lists, workflow processes, mobile-enabled user interfaces, and more. The complete base application can then be instantly customized using Joget’s intuitive visual development approach.

Low-Code Developer: Code Snippet Builder

Joget supports the full scale of the development spectrum, from no-code visual development to full-code functional coding by professional developers via plugins. There will be scenarios where applications that require snippets of low-code to achieve specific requirements, and this is where the Code Snippet Builder comes in. It can consolidate code snippets into a single repository for easier management, and with the integrated generative AI, developers can easily generate code directly from the code editor based on natural language prompts.

Application End-User: AI Writing Assistant

The AI Writing Assistant can be embedded into any application. Integrated to generative AI, it acts as an assistant to help end-users compose messages based on any specific writing style.

Application End-User: AI Assistant Live Chat

The AI Assistant Live Chat plugin can be embedded into any application UI. This practically allows developers to easily embed a chatbot into their applications for end-users to interact with using natural language.

Get Started with Generative AI on Joget

We are excited to share the new generative AI capabilities on the Joget platform with you! Get started and let us know what you think, along with what you would like to see in the future.

Download Joget Generative AI Plugins :

• Generative AI App Generator — Get the Generative AI App Generator plugin for Joget DX.
• Code Snippet Builder — Get the Code Snippet Builder plugin for Joget DX.
• AI Writing Assistant — Get the AI Writing Assistant plugin for Joget DX
• AI Live Chat Assistant — Get the AI Live Chat Assistant plugin for Joget DX.

Additional Resources:

• Get Started — Experience the Joget DX platform.
• Video Tutorials — Quick overview and build your first app.
• Knowledge Base — User and developer reference, samples and other documentation.
• Community Q&A — Ask questions, get answers, and help others.
• Language Translations — Translations for more than 20 languages.
• Joget Mindshare™ Series — Information sharing and educational content ranging from whitepapers, webinars, video tutorials, customer success stories and more.
• Joget PM-CD Evaluation Tool — Download Project Management-Citizen Development Cheat Sheet to evaluate the key considerations for bringing in Citizen Development tools for project success.
• Joget Academy — Self-paced online learning and certification.
• Joget Marketplace — Download ready made apps, plugins, templates and more.
• Joget Blog– Latest news, upcoming events, and educational content ranging from thought leadership articles, whitepapers, webinars, video tutorials, customer success stories and more.