Artificial intelligence and the ChatGPT revolution

Artificial intelligence is certainly not a new concept. Most people have some idea of what the term means, mainly thanks to science fiction books and films. At the end of last year, when OpenAI made the ChatGPT bot available for free, the general public moved decisively from science fiction into the real world.

At Webikon we work on a wide range of projects and are always trying to move forward and improve ourselves and our work. Everyone in IT knows how important it is to keep learning and to follow the latest trends and technologies and how they are used. ChatGPT and the other language models (there are already several of them) are the latest such technology, and it is not just Bill Gates who thinks it is a revolution on the scale of the personal computer and the internet.

Keeping up with the latest technologies and their applications was already complicated and demanding, and artificial intelligence multiplies the problem. Yet AI itself will help us automate or significantly speed up most of the work of processing and generating text or instructions. That frees us from routine, time-consuming tasks, so we as people can focus on the more interesting and creative side of the work and have more room to keep learning.

What is ChatGPT

ChatGPT is one of the large language models (LLMs), trained on an enormous amount of text. It has been further trained for conversation and to perform certain tasks better. It was developed by OpenAI, founded in 2015 as a non-profit research organisation; since 2019 it has also had a for-profit arm through which it offers its models commercially.

ChatGPT is not just a bot you can chat with naturally about almost any topic. It is above all a tool for boosting productivity.

Getting to know ChatGPT

In December I created an account, logged into the chat interface and started a conversation. Naturally, the first thing I looked into was the nature of ChatGPT and what it actually knows about how it works. I liked straight away that it doesn't try to pretend to be human, or to have consciousness and think. It readily admits that it is “just” a language model. Based on data, algorithms and a great deal of computing power up front, it predicts which word follows the previous one. And it does this so well that in many cases you cannot tell it apart from a human and the way people express themselves.

Working and talking with ChatGPT

I soon realised that ChatGPT really is “just” a tool, an extension and, to some extent, a mirror of ourselves. So once I had satisfied my philosophical curiosity, I started thinking about what it could offer in practice. I began experimenting with different questions and examples to understand its concrete possibilities and limits: what the bot can respond to and what it can generate.

After a while I managed to put together a basic structure of what ChatGPT can actually generate and what people need or would like to see. With the help of ChatGPT, I wrote a book full of general but practical prompts (instructions for the bot). The book is written in English, because 50% of ChatGPT's training is in English and that is where it is strongest. The category names give a better idea of what ChatGPT can do at the most basic level.

ChatGPT as a digital content creator and assistant

While writing, I realised that although ChatGPT is a text generator, that doesn't mean it only writes text. It can generate code in practically any programming language and use various libraries and applications known up to 2021 (its knowledge doesn't go any further). As well as generating code, it can understand, explain or comment on code. It can then suggest improvements or fix its bugs straight away.

ChatGPT formats the generated code neatly in its interface, so we can simply copy it and use it. So I started looking into the different formats it can generate. I found that ChatGPT can produce formatted text in HTML, using a structure of headings, bulleted lists and tables. It can't draw, but it will generate SVG for you, which means icons and diagrams too.

A few months on, people are playing chess with it and generating ASCII art. You can even have the bot simulate a JavaScript console or a Linux terminal. You can then try out and practise commands, and the bot responds as if it were that interface. One fan even managed to create a virtual computer inside the ChatGPT interface, connect it to an alternative internet and talk to an alternative ChatGPT there.

Using ChatGPT

Given these developments, we decided to jump on the AI wave early and start using it internally. AI won't replace you entirely just yet, but a person who uses it certainly will. That is why practically any company where people work on computers and create digital content needs to start using ChatGPT and language models. For digital agencies specialising in app development or providing SaaS, this applies twice over.

The first phase is introducing our colleagues (developers, marketers and product managers) to what ChatGPT can do in its interface, so that everyone sees how they can use the bot properly in their own role to be more productive. We are preparing internal webinars so that, by using a tool that is already available, we immediately raise both productivity and the know-how of everyone in the team.

Since a GPT model can be trained on your own data, the next logical phase is building an internal bot for our purposes. It will know our projects and their structure, our clients and their needs, and all the documentation and code. It will combine the full “intelligence” of ChatGPT with all the information relevant to our work, and become an indispensable assistant and an unrivalled productivity tool.

More generally, you can build your own applications on the GPT API, and we are slowly reaching the point where the limit is our imagination, not the technology available. Beyond generating text, we know that AI already produces images, sound and voice, and video. Gradually all these bridges will connect and multimodal AIs will emerge, so that from a simple text prompt you will be able to create a script, music and a video with actors indistinguishable from real ones.

And finally, there is actually carrying out an action, in other words creating a real assistant that you give voice commands to and that performs all the digital actions for you. Once the key APIs and services are connected, the bot gains the ability to perform any action on a computer or mobile device. So, to finish the way I started (with a nod to science fiction and popular films), I think the film Her shows most accurately what this technology already makes possible.

How Webikon sees the rise of AI

ChatGPT is a big step for everyone, inside the IT world and outside it. At Webikon, we have jumped on this wave because we know that with ChatGPT we can work more efficiently and improve the quality of our results. We are not worried that machines will replace our work. Our know-how lies not only in technical skill but also in a creative approach to solving problems.

With ChatGPT's help, we can become even better coders, marketers, analysts, designers and project managers. Its capabilities help us build our own solutions, made to measure for our clients. We can leave the routine to it and spend more time on what makes each solution original.