ChatGPT – the Solution Architect of the Future?

A post by Pascal Espinouse made me ask the question whether an AI model like ChatGPT could someday replace us architects working in information technology. If you don’t have the time to read through my views here is my short answer: AI models, like ChatGPT have the potential to help and assist an architect. However, today AI models like ChatGPT are not advanced enough to replace an architect, as designing the “right” solution requires complex decision-making, planning, and problem-solving.

What is involved in architecting a solution?

As architects we design solutions by applying methods and techniques combined with technical expertise:?“The architect has a responsibility for ensuring the completeness (fitness-for-purpose) of the architecture, in terms of adequately addressing all the pertinent concerns of its stakeholders; and the integrity of the architecture, in terms of connecting all the various views to each other, satisfactorily reconciling the conflicting concerns of different stakeholders, and showing the trade-offs made in so doing.”[1]

The actual solution design process will follow a similar pattern to that of a creative technology design process[2]. An architect will combine and group the different components together according to stakeholder group and will, over several sessions, develop concept views related to key architectural components, establishing different options. Using an architecture framework like IAF (integrated architecture framework) will help with a pre-defined set of artefacts that are required to establish a logical and a physical design.

Deciding the “right” option will mean balancing the various criteria like functionality, value for money, compliance, quality, and sustainability. IT architecture design involves complex decision-making, planning, and problem-solving that require human expertise and experience.

That is where most of the architect’s work is focused on – using knowledge and experience to research a particular subject, to apply design thinking and to solve problems to establish a solution.

The capability of our current AI tools

Our current AI capabilities and models like ChatGPT are language model designed to process natural language inputs and provide responses based on patterns in its training data. ChatGPT is based on a deep learning architecture called a transformer, which was introduced in a research paper called "Attention Is All You Need" in 2017.

The transformer architecture is designed to process and understand sequences of words or tokens, such as sentences or paragraphs, by using attention mechanisms that allow the model to focus on specific parts of the input. ChatGPT can be invoked by a user entering a question in natural language or by an application using and API (application program interface). The model generates a response based on its training data and learned patterns. ChatGPT has been trained on a massive corpus of text data, including books, articles, and websites, and it is able to generate responses on a wide range of topics (about 570GB of data).

And this is where an AI model like ChatGPT can really shine: “to understand natural language input and can provide answers to complex questions or engage in a conversation on a given topic. Its strength lies in its[3] “ability to learn from large amounts of data and adapt to different contexts, making it a versatile tool for a wide range of applications.”[4]

The main limitations of an AI model like ChatGPT

Whilst there are many impactful use cases where an AI model like ChatGPT can make a real impact, there are several key limitations :

1. Lack of real-world experience: ChatGPT has been trained on large amounts of text data, but it does not have real-world experience. This means that it may struggle to provide accurate responses to questions that require practical knowledge or experience.
2. Biases in training data: ChatGPT's responses may be influenced by biases in the training data it has been exposed to. For example, if the training data contains a lot of language that is biased against certain groups, ChatGPT may inadvertently generate biased responses.
3. Inability to understand context: While ChatGPT can generate coherent responses, it may not always understand the context in which a question is being asked. This can lead to responses that are technically correct but do not fully address the intent of the question.
4. Being not UpToDate: Tools like ChatGPT must be “trained” it’s always slightly behind. The ChatGPT system I used for this article was using data up to September 2021, so with almost an 18-month gap.
5. Lack of emotional intelligence: ChatGPT may struggle to recognize and respond appropriately to emotional cues in human communication. This can make it difficult for ChatGPT to engage in empathetic or emotionally nuanced conversations, and most importantly
6. Inability to verify sources: ChatGPT generates responses based on patterns it has learned from the text it was trained on, but it cannot verify the accuracy or credibility of the sources it draws upon. This means that responses generated by ChatGPT may not always be reliable.

ChatGPT as a Solution Architect

Let’s get to the actual question of this short article – what activities, if any, could an AI model like ChatGPT take over. ?As outlined above as well as in my last AI article, all current AI models are so-called “limited memory” models that are considered as “narrow AI” oppose to “strong AI”. In summary, the main difference is that Narrow A is an application of AI to enable a system to replicate human intelligence for a dedicated purpose. Strong AI is referred to as artificial general intelligence (AGI) which is based on self-aware systems, which is still unachievable with today’s technology. This means that an AI model like ChatGPT cannot be used to when emotional intelligence, understanding context or where real-world experience is needed. Instead, AI models like ChatGPT could help a Solution Architect with:

1. Providing advice and guidance during research: ChatGPT has been trained on a vast amount of data, which enables it to provide advice and guidance on various topics related to IT solutions.
2. Requirements gathering: ChatGPT could be used to help gather requirements for the IT system by engaging in a conversation with stakeholders or end-users to better understand their needs and preferences. This could help ensure that the system being developed is aligned with the needs of the people who will use it.
3. User interface design: ChatGPT could be used to generate ideas for the user interface design of the IT system by providing examples of natural language interactions that users might have with the system. This could help to create a more intuitive and user-friendly interface that aligns with user expectations and behaviour.
4. Quality assurance: ChatGPT could be used to test the IT system by engaging in conversations with it and evaluating the quality of its responses. This could help identify any gaps or issues with the system's functionality and guide improvements.
5. Support and troubleshooting: ChatGPT could be integrated into the IT system as a support or troubleshooting tool, allowing users to engage in a conversation to ask questions, report issues, and receive assistance. This could help to improve user satisfaction and reduce the workload on support staff.
6. Supporting collaboration: ChatGPT can facilitate collaboration between architects by serving as a centralized source of information and context.
7. Enhancing creativity: AI models like ChatGPT can generate new ideas and concepts that architects may not have considered. By exploring new possibilities and pushing the boundaries of what is possible, AI can help architects create more innovative and effective IT solutions.

Summary

AI tools and models like ChatGPT, Bing AI, Bard, etc are already making a big impact. With the imminent release of GPT-4 it is likely that the accuracy and value will be further improved, which will help architects designing solutions across the information technology sector. However, whilst AI can assist an architect, it cannot replace an architect, as the role of an architect

Thanks for Reading & Thanks to ChatGPT for its support in developing this article

[1] https://www.capgemini.com/2015/05/the-it-architect-coming-of-age/

[2] University of Twente, A Design Process for Creative Technology, https://research.utwente.nl/en/publications/a-design-process-for-creative-technology

[3] All according to ChatGPT – I asked: "what are the main limitations of ChatGPT?"

[4] I asked ChatGPT “what is the main strength of ChatGPT?”