Hardly a day passes when we do not find ourselves gasping for sanity in a hailstorm of audio leaks. But what exactly are ‘audio leaks’? In the hidden corners of our digitised existence, whispers transform into weapons, as audio leaks take on various forms. They range from simple wiretapping and skilled merging of out-of-context audio clips to deepfakes – the most spine-chilling of them all. Deepfake technology, given enough data, uses AI to clone anyone’s voice, image or video. The technology has been used to deceive people through scams, create conflict and erode trust in the digital world.

Yet, it is also important to recognise that AI technologies also possess tremendous potential for good and can positively impact our individual and collective lives. In the realm of music and art production, AI has emerged as a powerful tool for generating novel compositions and enhancing creativity. AI-powered systems can analyse vast musical libraries and assist musicians in composing unique melodies and harmonies, offering fresh perspectives and new ideas. Similarly, AI algorithms can help artists to generate visual artworks, helping them to explore new styles, experiment with different techniques – and unlock their artistic potential.

AI technology has also shown remarkable promise in creative writing. Large Language Models (LLMs), such as ChatGPT, can assist writers by providing inspiration, suggesting ideas and offering alternative ways of expressing their thoughts. When used responsibly and ethically, AI-powered tools substantially enhance human productivity and creativity.

Indeed, the history of technological advancements is marked by transformative inventions that have significantly enhanced human productivity and reshaped societies. The invention of the steam engine marked the First Industrial Revolution. The Second Industrial Revolution witnessed significant strides in harnessing electricity, the introduction of assembly line production and the advent of the telegraph. The Third Industrial Revolution was defined by the rise of the internet. With each successive breakthrough, human productivity surged, fundamentally transforming not only how we live and work, but also reshaping our thought processes. In this context, the emergence of LLMs like ChatGPT represents a groundbreaking advancement, with the potential to become intellectual helpers or thought companions for present and future generations.

Even before the First Industrial Revolution, a pivotal invention altered the course of knowledge dissemination. It was the invention of the printing press in the 15th century which made the mass production of books a reality and democratised access to knowledge on an unprecedented scale. The printing press ignited the flames of the Renaissance, fuelling scientific and cultural revolutions and nurturing a collective thirst for enlightenment that has endured across centuries. Later, the invention of the steam engine mechanised production processes and drove the growth of factories. It unleashed the potential of steam power and enabled the efficient transportation of goods, revolutionising trade and transportation. Electricity provided a new source of energy that could be used to power machines, leading to further advances in manufacturing, transportation and communication. The telegraph, introduced in the early 19th century, allowed rapid long-distance communication through electrical signals over wires. It revolutionised global communication, enabling near-instantaneous transmission of information across distances. The internet connected the world like never before. It transformed communication, commerce and collaboration by enabling the instant exchange of information, fostering global connectivity and giving rise to new forms of social interaction and economic opportunities.

In this lineage of transformative technologies, LLMs (born out of the Fourth Industrial Revolution) represent a new era of intellectual assistance. LLMs are AI-powered systems that can understand and generate human language, bridging the gap between humans and machines. ChatGPT, developed by OpenAI, is a remarkable example of this technology. It has been trained on vast amounts of text data and in the process acquired an impressive ability to understand context, generate coherent responses and even provide nuanced insights.

In this lineage of transformative technologies, LLMs (born out of the Fourth Industrial Revolution) represent a new era of intellectual assistance. LLMs are AI-powered systems that can understand and generate human language, bridging the gap between humans and machines. ChatGPT, developed by OpenAI, is a remarkable example of this technology. It has been trained on vast amounts of text data and in the process acquired an impressive ability to understand context, generate coherent responses and even provide nuanced insights.

As humans, our language and conversations serve as sources of experiences, knowledge and collective evolution. A major reason why LLMs, exemplified by the development of ChatGPT style AI, represent a groundbreaking breakthrough in AI capabilities is because they excel at learning from conversations and natural language experiences, enabling them to engage in meaningful interactions with users.

LLMs use vast amounts of text data from diverse sources, analysing patterns, structures and semantic relationships within language. Through a process known as pre-training, they develop a comprehensive understanding of grammar, context and even cultural nuances. This pre-training is then followed by fine-tuning, where the models are trained on specific tasks or domains to optimise their performance. The ability of LLMs to generate coherent and contextually relevant responses stems from their proficiency in predicting the most probable next word or phrase based on the given context, allowing them to answer a simple question, despite all the text they generate: “How likely is a human to utter this sentence in the given context?”

This process involves complex computations performed by machine learning models with millions or even billions of parameters – knobs that can be precisely adjusted so that the outcomes match our expectations. LLMs can be fine-tuned to specialise in various domains, from customer support to creative content generation and education. With their capacity to process and generate natural language, LLMs have not only transformed traditional human-computer interaction, they have also paved the way for breakthroughs in art – with DALL-E exemplifying their creative potential. The potential of these technologies empowers us to unlock new frontiers of understanding, creativity, and efficiency, positioning LLMs as a driving force in the ongoing Fourth Industrial Revolution.

ChatGPT, Whisper, and DALL-E represent a trio of technologies that have captured the public imagination and heralded a new era of possibilities. These advancements have demonstrated awe-inspiring potential and revolutionised human-computer interaction, natural language processing and image generation. From conversational companions to audio enhancement and visual creativity, these AI innovations have opened up a world of positive use cases.

ChatGPT has invaluable applications in customer service, virtual assistants and educational platforms. Its capacity to comprehend complex queries and provide coherent responses has elevated the quality of online interactions, enhancing user experiences across various industries. Students can seek personalised assistance, professionals can access on-demand information and individuals can engage in meaningful conversations, making knowledge more accessible than ever before. ChatGPT has the ability to write poetry and prose, translate text across languages, including Urdu, and even write computer programmes. As an example, the following poetry was produced by ChatGPT when asked to write two rhyming couplets praising the people of Pakistan:

In Pakistan’s land, a spirit so grand, Resilient people, united they stand.

From ancient heritage to futures untold, Pakistan’s people shine with hearts of gold.

Whisper, an AI-powered technology focused on audio enhancement, has brought significant advancements in audio enhancement, improving voice quality and facilitating better understanding in real-time communications. The technology has been invaluable for individuals with hearing impairments, facilitating better understanding and communication. Whisper’s noise reduction capabilities have transformed the quality of audio recordings, ensuring crisper sound in podcasts, videos, and teleconferences. Whisper also has the ability to convert human speech to text (like an assistant taking dictation).

DALL-E, an AI model trained to generate images from textual descriptions, has captivated artists and designers by bridging the gap between words and visuals. The technology can generate unique and vivid images that match specific descriptions, bringing to life fantastical scenes and concepts, once limited to the realm of imagination. From dreamlike landscapes to whimsical creatures, DALL-E has become a source of inspiration for visual artists, enabling them to explore uncharted territories and push the boundaries of creativity.

Recently, more products along the lines of conversational agents, audio and visual art generation have hit the market; they are easier to use and more accessible on desktop computers. These innovations are proving their worth across diverse industries. As we continue to witness advancements in AI, the positive use cases of these technologies serve as a testament to the potential and impact that AI can have on our lives.

However, the pursuit of augmenting human productivity can often come with a price. Hallucinations, inaccuracies and biases in AI technologies can lead to financial, emotional and societal harm. A few real-world incidents serve as cautionary tales, reminding us of the need for responsible AI development.

In 2023, Google’s parent company Alphabet experienced a loss of $100 billion in market value, following a high-profile debacle involving their new chatbot. The chatbot, touted as a revolutionary AI-driven customer service assistant, shared inaccurate information in a promotional video, causing confusion among users. This incident exposed the potential risks of over-reliance on AI systems without adequate checks and balances. The fallout of this event underscored the importance of ensuring that AI technologies are trustworthy, reliable and free from harmful inaccuracies. Another example highlighting the dangers of unchecked AI bias was when Amazon was forced to scrap a sexist AI tool. In an effort to automate the recruitment process, the company developed an AI system that screened job applicants based on historical data. However, the system exhibited gender bias, favouring male candidates and penalising female applicants. This incident raised serious ethical concerns and shed light on the inherent biases that can be embedded within AI algorithms. It served as a wake-up call, compelling organisations to acknowledge the potential pitfalls of AI and work towards fair and unbiased systems.

These examples illustrate specific instances of AI failures – and they also represent broader issues. LLMs like ChatGPT, which leverage large-scale training on diverse text sources have the potential to generate fake news, amplify societal biases, reinforce marginalisation and normalise inequity. As powerful as these models may be, they must be wielded responsibly. The responsibility lies not only with developers and organisations, but also with policymakers, researchers and society as a whole. Democratising AI is a crucial aspect of responsible AI development. It involves ensuring that access to AI technologies, knowledge and resources is available to a diverse set of individuals and communities. By democratising AI, we can mitigate the risk of concentrated power, biased decision-making and exclusionary practices. Transparency, accountability and inclusivity will be pivotal in shaping a future where AI benefits all of humanity.

Incidents of hallucinations, inaccuracies and biases remind us of the need for robust safeguards, ethical considerations and ongoing scrutiny. As we navigate the transformative effects of AI, it is imperative to prioritise responsible AI practices and strive for a future where AI is a force for good, enhancing our lives while ensuring fairness, equity and respect for all.

AI holds immense potential for developing countries, presenting opportunities to drive progress, empower local communities and foster economic development. Beyond their mainstream applications, these technologies can be harnessed to address unique challenges faced by developing countries.

A crucial aspect of leveraging AI in developing countries lies in the localisation of these technologies for use with under-resourced languages. Language models like ChatGPT have the ability to understand and generate text in multiple languages, making them invaluable in regions where resources for translation and natural language processing are limited. By adapting and training AI systems in local languages, developing countries can overcome language barriers, foster inclusivity and provide access to digital services to a broader population. Such localisation efforts pave the way for innovation, e-commerce and enhanced communication in local languages, empowering communities and driving economic growth.

MIT economist Daron Acemoglu’s research on the impact of AI technologies on the future of work and economic development provides valuable insights for developing countries. Contrary to concerns about extensive job displacement, Acemoglu suggests that AI can complement human skills, enhancing productivity and enabling individuals to focus on tasks of higher value. This paradigm shift creates new opportunities for skill development, prompting countries to invest in education and training programmes that equip their workforce with the necessary skills to thrive in an AI-driven era. By embracing AI as a tool for enhancing productivity and leveraging their unique human capital, developing countries can position themselves for economic growth and increased competitiveness in the global market.

For countries like Pakistan, several steps can be taken to improve AI competitiveness and accelerate economic development. First and foremost, investing in education and research to develop a strong pool of AI talent is crucial. By fostering partnerships between academia, industry and government, Pakistan should invest in AI centres of excellence, promote knowledge sharing and drive innovation. Additionally, creating a supportive regulatory environment that encourages AI adoption, protects data privacy and ensures ethical AI practices is essential.

Collaboration and international partnerships are vital in bridging the AI divide. Developing countries can benefit from knowledge exchange, technology transfer and the capacity-building initiatives offered by organisations, universities, and industry leaders. By embracing these collaborations, countries like Pakistan can accelerate their AI readiness and tap into global resources and expertise.

In 2018, the Government of Pakistan took a step forward by establishing the National Center of Artificial Intelligence and the National Center in Big Data and Cloud Computing. Although these centres have made notable contributions, it has become evident, as their terms come to an end, that bolder and sustained interventions are imperative to foster Pakistan’s AI talent, bridge the growing skills gap between developed and developing countries and enhance competitiveness on the global landscape. By investing in long-term and strategic initiatives, Pakistan can position itself as a global powerhouse in AI, unleashing a wave of innovation, economic growth and prosperity for the nation.

Dr Agha Ali Raza and Dr Ihsan Ayyub Qazi are Faculty in the Department of Computer Science, Syed Babar Ali School of Science & Engineering, LUMS.
agha.ali.raza@lums.edu.pk
ihsan.qazi@lums.edu.pk