Artificial intelligence came to public attention last year when US company OpenAI made the two AI tools ChatGPT and Dall-E freely available for everyone. The media is reporting on it extensively and everyone wants to try out the tools.
More than a million people registered for ChatGPT alone within a few days. Apart from Dall-E, it is primarily the two popular image generators Midjourney and Stable Diffusion that are being actively used at present.
Michael Maurer, what’s behind the AI hype?
Michael Maurer: The latest AI applications are very powerful and easy to use. It’s very understandable therefore that there’s also significant interest outside the tech bubble. I’ve been involved with artificial intelligence for a long time and have worked with earlier tools like GPT-3 and Stable Diffusion. These were also powerful, but even I was surprised by what ChatGPT can do.
Is ChatGPT a quantum leap for artificial intelligence?
No, but it’s a successful enhancement of GPT-3. The preoccupation with ChatGPT only came about because of the high level of public interest in the technology. In the meantime, Google has also launched a competitor to ChatGPT in the form of “Bard”. Other big tech players like Microsoft and Meta are likely to be working on similar solutions
“We will no doubt be amazed in a few years at how “stupid” Word, PowerPoint, etc. still were in 2023.”
Michael Maurer, Senior Consultant and Solution Architect at bbv.
What impact will such AI applications have on business and industry?
AI tools are already integrated into many systems today and will be found practically everywhere in the future. They are used wherever content, such as text, images and videos, is created. Microsoft has also announced that it intends to integrate AI solutions into its Bing search engine or Office programs at some point. “We will no doubt be amazed in a few years at how “stupid” Word and PowerPoint still were in 2023.” At the same time, however, the existing professional AI concepts will continue to be developed and will be used increasingly – for example large scale models, neural networks or generative deep learning models. Completely new applications are being found and business models are emerging.
In addition to the text-based Open AI tools, it is primarily AI applications for image generation that are creating all the hype. What changes will image tools like Dall-E, Midjourney or Stable Diffusion bring for companies?
AI image generators are likely to compete with stock image databases in the near future. Thanks to AI, symbolic images for web sites or blogs can be created at will and at the touch of a button. What I also find interesting, incidentally, is the tool Vall-E from Microsoft, which generates voices with neural codec language modelling and could be used wherever text is to be converted to speech.
Applications of ChatGPT
We asked ChatGPT what the business models are for using ChatGPT. This is what AI said:
“There are several business models that can be considered for using ChatGPT. Some examples include:
- Automated content creation: ChatGPT can be used to automatically create content like blog posts, product descriptions, news articles and more. This can be useful for businesses that need to produce a lot of content in order to maintain their online presence.
- Customer support: ChatGPT can be used to create automated chatbots that can answer customer enquiries and solve problems. This can reduce the cost of customer support and increase customer satisfaction.
- Speech generation in applications: ChatGPT can be integrated into applications like games, virtual assistants, voice control and more in order to enable natural speech interactions.
- Customised advertising: ChatGPT can be used to create personalised advertising campaigns by generating customised messages and offers on the basis of data such as user behaviour and interests.
- Language translation: ChatGPT can be used to generate automatic language translations that can translate in real time.
There are many other business models that can be created with ChatGPT and the possibilities will no doubt continue to evolve as more companies and developers start using the technology.”
What benefits do AI tools offer for software development?
They can be a valuable tool for programming. Usable code fragments can already be generated today with GitHub Copilot and used to generate software. While the AI application does not always deliver optimum results, as a developer I can adapt the fragments without needing to write the entire code myself. However, these tools will get better and better over time as they learn and recognise what is being used and not used. This is how they train and improve themselves in everyday use.
Can similar code fragments not also be assembled from relevant online sources?
Absolutely, but the results are generally created faster with AI and are more comprehensive. In my experience, such tools can support software developers by relieving them of routine tasks or solving individual problems. With AI pair programming, human and machine complement each other ideally, with the result that AI allows programmers to work more efficiently.
The tech sector is battling a skills shortage. Can AI help or is it a job killer?
There are certainly jobs that artificial intelligence will perform partly or fully in the future. But AI will also enrich our everyday working lives in the sense of an enabler. In other words, it helps us to work better and more efficiently. A good example of this is the chess discipline Advanced Chess. In this case, the combination of chess computer and human chess player produces the best results. This duo will not be beaten by either humans or computers.
What about the acceptance of AI in software development or in other industries?
As new technologies emerge, there are always people who fear them initially. Some also think that a human can basically do a better job than a machine. This is likely to remain an area of tension. Ideally, AI handles repetitive tasks so that we can concentrate on activities that require human interaction. This is already the case today with software, automated processes or robotics.
A further area of tension is the issue of copyright and usage rights. Who owns the texts and images generated by AI?
You generally have usage rights for images you create with Dall-E, for example. However, the problem is not the images generated, rather the data used to train the AI. These are mostly texts and images taken from the Internet whose authors have not consented to their use. This can also be problematic with AI-based software development.
Why is that?
The AI tool GitHub Copilot can be used, for example, to generate code fragments, which you can use in your own software. However, the Copilot models may have been trained with data for which public rights do not exist – especially if the products are used commercially afterwards. This will generate much discussion in the coming years and legal cases could arise for which current copyright laws do not yet have a solution.
Are there other risks involved in using AI tools?
Artificial intelligence can sometimes generate content that is racist, sexist or unethical. This is again down to the data that was used to train the AI. If this data is biased, then the models are also biased. For example, if HR data is dominated by white males, so too will the generated content. This means that non-white people could be discriminated against in AI-based HR recruitment.
What can be done about this?
Science is already actively researching this question today. One approach is to likewise identify such bias using AI. The bias is then evened out subsequently with artificial or synthetic data.
The longer AI is used, the more it will also be used in SMEs. How can companies successfully implement an AI project?
If a company has large volumes of good-quality data, then this is a good starting point. It is also possible to source data externally. The latest cloud technologies should be used preferably to prepare AI models. However, we are increasingly finding in SMEs, in particular, that sufficient volumes of data are not available. But such companies can also use AI solutions. For example, there are free and freely available AI models that can be used for speech or image recognition. In addition, tech giants like Microsoft or Google offer ready-made AI services and APIs, which can be integrated into your own software solutions.
What challenges face companies that opt to use artificial intelligence in their projects?
They have to find a use case or business model that offers tangible added value. It takes courage to use AI productively and to defend it against internal resistance. And there is still a lot of that! Lack of specialist staff is often a challenge as well.
Are specialists even needed in the first place to use AI?
Anyone can use AI tools like ChatGPT. This is precisely what creates new opportunities and business models. As with other emerging technologies, democratisation is taking place. In the beginning, it is only the specialists who have access to a new technology and are able to use it. Later, everyone can use it.
Will the hype around such simple AI tools subside again?
Quite the opposite. Integration of such applications into other systems, in particular, will further contribute to their spread. Corresponding APIs already exist for Dall-E and GPT-3 and it should soon be possible to integrate ChatGPT. There will be a further push towards AI if AI applications make it onto popular platforms and devices. In addition, while ChatGPT is surprisingly powerful, there are rumours that ChatGPT, which actually works with GPT Version 3.5, only has about 10 percent of the performance that Version GPT-4 will have one day. Version 4 might even be in a completely different league again and actually represent a quantum leap. I am also assuming that, apart from OpenAI, other tech giants will also try to get their solutions into the public domain. The real AI hype is yet to come.
Michael Maurer is a Senior Consultant and Solution Architect at bbv in the financial service provider & transport fields. He focuses on digital transformation, innovation management and the design of (cloud) software solutions. Thanks to his many years of experience in the SME area, he understands their challenges and can highlight targeted solutions and implementation options. He is convinced that digitalisation offers a major opportunity for all companies.