This article explores the concepts and current thinking related to AI and the author has a chat with AI to find out what AI ‘thinks’ about tax stamps.
AI is everywhere these days and, make no mistake, it is, and will continue to be, the next big thing in the line of technologies that have disrupted almost everything. It’s in the workplace, the classroom and now, even art, music, literature, poetry, and conversations that were heretofore the sole domain of humankind.
The mere notion of AI invokes different thoughts and feelings amongst many carbon-based life forms, ‘AKA humans’. For some, it’s a true game changer in the technology world that will bring widespread benefit to the way we work, live and play. For others it’s the beginning of the end – an AI future where the machines turn on their creators.
This has been the subject of countless sci-fi books and films for generations. You know the ones, with dark skies, bombed out cities and toxic storms on the horizon where shiny silver robots tread through the dunes of ash and human remains.
While this latter scenario may seem a bit pessimistic, it may not be so. Indeed, arguably some of the world’s smartest and tech savvy people, including the late Steven Hawking, Elon Musk and Bill Gates have warned that untethered AI has the potential to literally wipe out the human race.
But, for now, most of the tech gurus are abuzz about a new world where AI will literally translate into exponential benefits across virtually all sectors, bringing with it an estimated $15.7 trillion in growth to the global economy as it reshapes entire industries.
Given the nature of the tax stamp and traceability industry and the multitude of diverse technologies offered by it,
AI is already having an impact, and will undoubtedly play an instrumental role in the solutions of the future.
The term AI was first used in 1956 by John McCarthy of Stanford University but its origins may go back as far as the ancient world.
In Greek mythology, Talos was a giant robotic creature constructed of bronze who acted as guardian for the island of Crete. AI is intimated in Mary Shelly’s Frankenstein and featured in the article ‘Darwin Among the Machines’ written by Samuel Butler, where he raised the possibility that machines were a kind of ‘mechanical life’ undergoing constant evolution and that eventually machines might supplant humans as the dominant species.
In more recent times, the foundation of machine learning was blazed by British computer scientist Alan Turning in World War II during his development of the Enigma machine used to crack Nazi codes. The world was shocked when chess master Garry Kasparov was bested by a computer in 1997 and when IBM’s Watson beat human players on the US game show Jeopardy.
In simple terms, the AI of today essentially describes computers that can learn – sort of – as opposed to simply doing what their coding instructs them to do. Computers still cannot entirely think as humans do but they can be taught to analyse information and data in many different ways, identify patterns and make decisions, including writing their own programming based on what they have learnt.
Given the vast sea of data available to AI, it can quickly see what humans cannot and go deep into vast data sets to improve on its results. But it has its limitations as it cannot make judgements based on prior experience, feelings, or instincts.
There is no shortage of critics when it comes to AI and particularly when it comes to the arts. An article in The New Yorker magazine likened the chatbot ChatGPT to ‘a blurry jpeg of all the text on the Web. It retains much of the information on the Web, in the same way that a jpeg retains much of the information of a higher resolution image, but, if you’re looking for an exact sequence of bits, you won’t find it; all you will ever get is an approximation’. It is also prone to glaring inaccuracies and misrepresentations that are referred to as ‘hallucinations’, or nonsensical answers to factual questions.
The most successful versions of AIs in recent years have used a system known as a neural network, which is modelled at a very simple level on how humans think a brain works. There is no easy way to categorise AI and the field is growing so rapidly that new approaches are popping up almost every day. The four main categories, however, include:
Reinforcement learning (RL) – this is perhaps the most basic form of machine learning that relies on an age- old technique used by humans: trial and error. The AI learns from feedback provided based on its own actions and experiences.
Large-language models (LLM) – these are trained by pouring into them billions of words of everyday text, gathered from sources ranging from books, websites, blogs, tweets and virtually all knowledge that exists digitally. These models draw on all this material to predict words and sentences in certain sequences and with correct grammar.
Generative adversarial network (GANS) – this is a way of pairing two neural networks together to make something new. They are an exciting recent innovation in machine learning. GANs are generative models: they create new data instances that resemble training the AI.
For example, GANs can create images that look like photographs of human faces, even though the faces don’t belong to any real person. GANS are used in creative work in music, visual art or filmmaking.
Symbolic AI (SA) – this involves the clear embedding of human knowledge and behaviour rules into computer programmes. This approach looks to the past for inspiration. For example, if you ask the question ‘what is an orange?’ SAs would reply that an orange is a round, orange, citrus fruit. The symbols used by AI explain the orange in terms of colour, shape and kind which are all symbols in human thinking.
A chatbot draws on the AI using large- language models. A chatbot is trained on a vast amount of information that is culled from the internet and can respond to text prompts in a variety of ways, including: conversational-style responses, conducting research, creating computer code, preparing travel itineraries and even writing essays, stories and poems.
The most famous example is ChatGPT. It has been developed by OpenAI, a San Francisco-based company developed and funded primarily by Microsoft. Launched as a simple website in November 2022, it became an overnight sensation reaching more than 100 million users within its first two months.
The chatbot gives plausible-sounding – if sometimes inaccurate – answers to questions. It can also write just about anything and summarise lengthy documents and, to the alarm of teachers, draft papers and essays.
The conversation depicted below took place over a number of weeks (since AI is constantly adding more content to its vast library) using an application that pulls from the most popular and, generally agreed, best AI chatbot, ChatGPT.
Since ChatGPT uses a large-language model learning paradigm, this conversation is limited by information that already exists somewhere in the digital world. It would not have access to presentations, programmes or other knowledge that has not been digitised or made public, so the discussion must be taken with a grain of salt.
In the revenue/customs/border and supply chain security world, AI is already being used by some of the world’s leading agencies for risk management to target risky consignments, traders and taxpayers. This has created a true step change in how agencies deal with both strategic and tactical risks and allows dynamic, real time risk assessment that is yielding significant results. The reliance on AI in this domain will only increase as AI functionality and usage expands.
The conversation depicted here shows a very rudimentary understanding of the tax stamp and traceability industry. When asked to prepare a technical specification for a tax stamp and track and trace system the chatbot offered some generic text similar to that which is displayed in the questions on this page. But it also noted some of the key best practices, albeit at a very high level, some of which looked familiar, if not identical, to this author in particular.
AI may be the new bane for many, and one has to have some empathy and perhaps even concern for those effected. The potential for AI to be used for malicious purposes or evolve beyond a point where we carbon-life forms can control it, should be a concern.
But the good news is that AI still has a way to go before it displaces tax stamps entirely. And indeed, in its current form, it is not capable of writing detailed specifications for tax stamps or render a tender document.
This provides comfort to consultants like this one. At least for now.
AI generally has a favourable view of tax stamps and understanding of their core functions and even best practices with respect to technical design and implementation. However, it is a bit inconsistent on the subject depending on how questions are asked eg. Q1 and Q2 versus Q3.
