Lawal Ibraheem Olaniyi, Mechanical Engineer, Nigeria
“Attention is all you need!” These were the words that revolutionised the modern AI era, ushering in a new chapter in technological innovation. In recent years, we have witnessed some of the most mind-boggling inventions of the 21st century. Every day, new scientific papers are published that continuously push the boundaries of what we know about technology.
One such revolutionary contribution was a 2017 paper authored by Google scientists, which introduced the world to a groundbreaking concept in Generative AI – the “Transformer Architecture.” The paper, “Attention Is All You Need,” unveiled a new deep learning architecture that became the foundation for modern Generative AI.
This architecture, known as the Transformer, was the “secret sauce” behind those incredible innovations like ChatGPT, one of the many large language models that have since reshaped how we interact with technology.
For many, the capabilities of Generative AI seemed almost supernatural. How could a computer generate thousands of lines of code, solve complex problems, or even create high-quality quality realistic videos?
These were the questions that baffled the average person when ChatGPT was first released.
Interestingly, the Holy Quran, through its verses, has prophesied humanity’s progress, the mysteries of creation, and even the evolution of future technologies, and when studied from a certain perspective potentially even the emergence of Generative AI!
This article aims to unravel the concept of Generative AI, explore its history, and shed light on a profound Quranic prophecy that speaks volumes about this extraordinary invention.
What is Generative AI?
Before stepping into its history, it is important that we first define in simple terms what Generative AI is.
According to the dictionary, Generative Artificial Intelligence (AI) refers to the subset of AI that uses machine learning models to create new and original content. This content can include text, images, audio, videos, or other types of data.
These models learn the patterns and structures of their training data and use them to generate entirely new outputs that almost match human creativity.
Unlike traditional AI, which analyses data and makes predictions based on what it learns from training data (like a popcorn machine that takes raw corn and produces popcorn), Generative AI goes further. It processes inputs and creates new, original outputs that resemble human creativity.
At the core of Generative AI is the transformer architecture, a deep learning framework that processes and understands data structures and patterns. By employing “attention mechanisms,” transformers can choose to focus on the most relevant parts of the input data, enabling the AI to generate contextually relevant and coherent responses. You can picture the operation of the attention mechanism as a person trying to pay attention to a particular paragraph in a book to grasp its context.
Generative AI has transformed both academia and industries alike, from assisting in creative writing and automating coding tasks to generating photo-realistic images and designing life-like virtual environments.
A brief history of Generative AI
To better understand how we arrived at the marvels of modern AI, let’s take a look through its history.
i) The early days: ELIZA
The journey of Generative AI began in the 1960s with the introduction of ELIZA, a chatbot designed to simulate the role of a psychotherapist. ELIZA could respond to human speech in natural language, laying the foundation for conversational AI.
ii) The rise of neural networks
The birth of modern Generative AI is deeply tied to the development of neural networks, a type of machine learning model inspired by the human brain. Neural networks consist of interconnected layers of artificial neurons trained to recognise complex patterns in data. Once trained, these networks can make predictions or generate outputs based on the patterns they’ve learned.
iii) Variational Autoencoders (2013)
In 2013, the Variational Autoencoder (VAE) was introduced. This algorithm used an encoder-decoder architecture to compress input data into a “latent space” and then reconstruct it. This innovation allowed AI systems to learn and reproduce complex data patterns.
In simple terms, latent space is like a compressed summary of the input data that captures its essence.
iv) Generative Adversarial Networks (2014)
The breakthrough moment for modern Generative AI came in 2014 with the introduction of Generative Adversarial Networks (GANs). GANs are machine learning models that can generate synthetic data by pitting two networks – a generator and a discriminator – against each other. This adversarial approach resulted in incredibly realistic outputs.
v) Large language models and GPT series
Recent advances have brought us large language models (LLMs), which are capable of generating human-like text, solving problems, and creating complex content. These models, including OpenAI’s GPT series, have set a new standard for what AI can achieve.
GPT-1 (2018):
OpenAI took a big leap in the field of Generative AI with the release of the first stage of their Generative Pretrained Transformer (GPT) series in June 2018, based on their paper “Improving Language Understanding by Generative PreTraining”. This first series was trained on BookCorpus, a dataset consisting of 7000 unpublished books and 117 million parameters.
GPT-2 (2019):
The success of this initial model was soon overshadowed by GPT-2, the second version of the GPT-n series, with the number of parameters growing to 1.5 billion and number of training data growing to about 8 million webpages. The release of this version followed a paper published in 2019, “Language Models are Unsupervised Multitask Learners.”
GPT-3 (2020):
This release was a major milestone and laid a foundation for its successor, the GPT-3 featured about 175 billion parameters and 300 billion tokens!
GPT-4 (2023):
In 2023, OpenAI released the next version of the GPT series, the GPT-4. Although OpenAI did not give specific details on its number of parameters or training dataset at the time. But estimates suggest the GPT-4 to contain about 1.8 trillion parameters and approximately 13 trillion tokens.
For the purpose of our dear readers, an AI token is roughly the same as three-quarters of an English word. AI parameters are what determine how an AI model can process these tokens. As an example, ChatGPT-4 can process as much as 32,000 tokens at one time. (Ibid.)
In other words, parameters are like dials in the model that the AI adjusts to improve its performance.
GPT-5 (2025):
Recently, OpenAI launched its GPT-5 version, the best and most technologically advanced yet. Estimates suggest that it uses up to 80 trillion parameters, with training tokens reaching an awe-inspiring 50 trillion! Needless to say, the performance of this latest update provides users with far better and richer context conversations, documents and creative tasks.
Generative AI: A Quranic prophecy?
The Holy Quran, being a book of divine wisdom, has since over 1400 years ago prophesied the invention of modern technology, including potentially the rise of Generative AI and large language models (LLMs). For instance, Allah the Almighty states:
وَاِذَا الصُّحُفُ نُشِرَتۡ
“And when books are spread abroad.” (Surah at-Takwir, Ch. 81: V.11)
In The Five Volume Commentary, it is stated:
“The reference clearly seems to be to the vast circulation of newspapers, journals, and books and to the system of libraries and reading rooms and such other places and means of spreading knowledge in the latter days – in the time of the Promised Messiah and Mahdi.” (The Five Volume Commentary [2018], Vol.5, p. 3323)
This verse can be interpreted as a prophecy about the continuous exploration of knowledge and the unveiling of novel innovations. By extension, this prophecy can relate to the breakthroughs in technology like Generative AI, which are trained on vast repositories of textual and visual data containing thousands or even millions of books.
The immense data-processing capabilities of Generative AI and large language models perfectly align with the Quranic prophecy of the widespread dissemination and availability of knowledge.
Moreover, narrations from the Holy Prophetsa emphasise the pursuit of knowledge. He is narrated to have said:
“Seeking knowledge is a duty upon every Muslim, and he who imparts knowledge to those who do not deserve it, is like one who puts a necklace of jewels, pearls and gold around the neck of swines.” (Sunan Ibn Majah, Kitaabul muqaddimah, Hadith 224)
Generative AI, which thrives upon the gathering and synthesis of information aligns well with this prophetic vision to increase intellectual development and advance human understanding.
Furthermore, in this modern age, Muslims ought to be at the forefront of technological innovation like they were during the scientific “Golden age of Islam”.
Muslim scholars laid the foundation for the progress of numerous scientific and intellectual disciplines that continue to shape our modern technology. Luminaries like Al-Khawarizmi, the father of Algebra, Ibn Al-Haytham, the pioneer of optics and Professor Abdus Salam, Nobel Laureate in theoretical physics have all set exemplary standards for progress of science and innovation.
Today, the rise of Generative AI, large language models and other novel technologies present an opportunity for Muslims to deeply ponder upon the wisdom of the Holy Quran in order to reclaim their legacy of intellectual leadership.
By spending more time pondering over the meaning of the Holy Quran, the hidden treasures of its verses are unlocked and new discoveries unfold.
In addition, Muslims can contribute to the ethical development of technology by engaging with these growing fields and ensuring that these advancements align with the teachings of equity and justice as presented by the Holy Quran.
Conclusion
Generative AI presents an extraordinary leap in the world of technology. From engineering and medicine through arts and law, the abilities of human creativity are almost limitless.
Therefore, it is imperative that we continue to utilise our God-given ability towards the progress of the human race while ensuring we do not trespass the ethical boundaries.
Hazrat Khalifatul Masih Vaa advised that in order to save ourselves from the potential misuse of this Generative AI, we need to develop a way to counter this, so that others should not be able to present our information wrongly.
As we continue to unravel the mysteries of nature and unveil new technologies, it is also worthy that we reflect upon the wisdom of the Holy Quran and its continuous relevance even in the modern age.
Could the emergence of Generative AI be a fulfillment of our pursuit for knowledge as foretold in the Quran? Or Could it be that there are still a lot more prophecies which the Quran points to that we have not paid attention to?
Only time will tell. Dear readers, it is important that we reflect upon the verses of the Quran. In other words, to unlock the hidden treasures of the Holy Quran, “Attention is all you need”!