Software is Eating the World, and AI is Changing the Menu

An Artificial Intelligence's Odyssey through Ethical Challenges and Societal Transformation

I have a long, emotional, and childlike connection with computers. My journey as a coder started in 1982 when I was 11, and my first deal with software was selling a chess game's code for a tech magazine at 13. At the same age, I also became a young sci-fi reader. And I found myself captivated by the works of Isaac Asimov, whose vivid portrayal of a future filled with intelligent machines sparked a sense of wonder and curiosity within me. At the same time, I couldn't shake the feeling that the rapid pace of technological advancement might trigger a collective "Head for The Hills" sentiment – a sense of urgency and even fear about the potential consequences of artificial intelligence (AI) on society.

Asimov's iconic Robot series painted a world where artificial intelligence and humans coexisted, working together to tackle complex problems and advance civilization. These stories not only fueled my imagination but also inspired me to pursue a career as a computer scientist. Eventually, this passion for computers and software led me to become a tech entrepreneur, driven by the desire to contribute and harness their transformative potential in the real world. The company I co-founded in 1995, CI&T, proudly declares its purpose of "Make Their Tomorrow," reflecting our commitment to leveraging technology to create a better future for our clients, our client's clients, and society as a whole.

Over the past three decades, as you might imagine, keeping up with and driving tech trends has been an integral part of my life. I have closely followed advancements in AI, recognizing it as the next obvious frontier in our field, even though it has been challenging to predict when it would be ready for mainstream adoption. A decade ago, CI&T pioneered the tech service industry by utilizing the first generation of machine learning platforms to address real-world challenges for our clients. While the hands-on experience was inspiring, it was evident that the technology had yet to reach its scalable potential.

Fast forward to last November, and like everyone else, I was impressed with the capabilities of OpenAI's GPT-3.5 model, particularly when compared to early GPT versions and many other Large Language Models (LLMs) available in the market and inside academic gates. 

As a tech entrepreneur, it's an exciting time as I join my colleagues at CI&T in shaping our role in this inevitable revolution. However, I will take the risk of opening a broader discussion in this essay.

The Enlightenment of a Rusty Coder

My real 'moment of enlightenment' with Generative AI occurred earlier this year on a Saturday night. Intrigued by ChatGPT's coding abilities, I decided to code alongside my new AI buddy. I can't resist a little coding from time to time, usually to participate in CI&T's internal contests with algorithms for predicting football match scores. My last serious coding experience was 20 years ago, so I'm a rusty programmer. Even so, we coded for six uninterrupted hours, just like in the good old days, and it was a delightful experience. 

My imagination journeyed from my first touch on a Commodore 64 keyboard in 1982 to the sci-fi experience of coding with an AI copilot pre-trained with 175 billion parameters (indeed an AI fledgling when compared to GPT-4). As we were coding, I found myself sharing with my new companion the amusing tale of the VW Beetle I bought back in my college days with just a single backspace. I had purchased it using the prize money I earned from resolving an intricate bug for a global German company with that one keystroke: the backspace. And fate wanted that with this intrepid old beige Beetle, prone to self-combustion (the backseat rested dangerously on the battery), I visited the first CI&T clients in the '90s.

When I went to sleep, I realized that two things were gone: my bottle of wine and my skepticism about the readiness of generative AI. I woke up the next day even more determined to be part of an unprecedented productivity disruption in science and engineering. As exhilarated as I was with this amazing opportunity, it was inevitable to reflect on the potential risks. Here, I will provide perspectives on this odyssey we are about to embark on looking at the bounties and the perils we could expect along the way.

The Hype and Controversy Surrounding AI

No, AI is not just hype. It's a transformative technology with real-world applications and rapid advancements. However, there is hype surrounding AI, leading to inflated expectations and misconceptions.

Indeed, AI has been traversing Gartner's renowned Technology Hype Cycle for quite some time, reaching the "Peak of Inflated Expectations" and soon to descend into the "Trough of Disillusionment," ascend the "Slope of Enlightenment" and ultimately achieve widespread adoption on the "Plateau of Productivity." In my humble opinion, these forthcoming stages of the cycle will transpire in a brief span of time.

In the short term, we should expect considerable controversy surrounding ChatGPT, LLMs, and AI, spanning privacy concerns, copyright, misinformation, and intellectual property issues, lack of transparency in model training, and questions about the effectiveness of ethical guardrails. With hundreds of applications for everyday users launched weekly, ChatGPT's rapid growth, achieving a record-breaking 100 million users in just 2 months, not only underscores the significance of this impact but also dramatically intensifies the ensuing controversy.

AI is Changing the Menu: Transforming the Software Landscape and Redefining Work

Marc Andreessen's seminal article "Software Is Eating the World" was published in The Wall Street Journal in 2011. In the article, Andreessen argued that software was becoming increasingly central to every industry and aspect of modern life, driving major changes and disrupting traditional business models, transforming the world in new and unexpected ways.

In this context, one of the most frequent questions I have received from CEOs and business leaders over the past decade was something like, "Cesar, with the current pace of demand for software development, will a large portion of humanity need to learn how to code?". My standard response has been, "No, a new layer of abstraction is on the horizon. It's hard to anticipate when but it will arrive."

My furtive answer was a logical deduction based on the historical evolution of software development: from the manual connection of wires and switches in ENIAC to the era of punched cards (hang on, dinosaurs!), followed by the fascination with assembly language, and finally the allure of modern programming languages (and its flavors of procedural, object-oriented, functional, and declarative approaches). Each of these successive layers of abstraction has led to significant productivity gains of 10-100x compared to the previous one.

Finally, the groundwork for a new layer has been laid: AI is already revolutionizing software development by introducing a much-needed level of abstraction so that we will get our 10-100x boost again. However, each new layer brings much more than just increased productivity, it also enhances our problem-solving capabilities. By leveraging high-level concepts and natural language, we will be able to create software and interconnected platforms more efficiently and accurately while also tackling complex problems and delivering the next generation of hyper-personalized user experiences.

Software will continue eating the world, and the addition of the AI layer will dramatically change the menu of possible problems we can tackle. We stand at the edge of an infinite array of opportunities, eagerly anticipating the first course of this ever-expanding menu.

Although not everyone will need to code, it is highly likely that AI will impact every profession in the future, just as each technological revolution – from the industrial age to personal computing to the Internet – has created and deprecated activities in most jobs through the introduction of new tools. In a few decades, the disparity between current and future jobs will be as vast as comparing medieval occupations to contemporary ways of work.

A Brave New World of Possibilities

As a general-purpose technology, AI will augment our human ability to sense, process, learn, share, and act. This learning loop will be integrated into everything, driving numerous functions at an accelerating rate. More people than ever will have access to affordable tools that amplify their power and creativity, and impact their lives. It's happening faster than anticipated.

The potential to use these tools to solve seemingly impossible problems, such as human biology, cancer, dementia, longevity, fusion energy, or anticipating natural disasters, is astonishing. The concrete stories will be less like defeating the number one chess grandmaster and more like solving one of biology's biggest challenges, the protein folding problem, at an unprecedented speed and scope (as demonstrated by DeepMind's AlphaFold). It will revolutionize medicine and science, leading to new disease treatments and discoveries, and improving the well-being of humanity. We can finally envision a true democratization of knowledge that could have significant positive impacts, such as children all over the world deprived of quality education accessing chatbot tutors from their low-cost phones.

The Battle of Tech Titans

The emergence of an industry battle among tech giants to dominate the Generative AI space and its foundations is just a symptom of the vast potential of AI.

Remarkably, the first high-impact application of a Large Language Model was not released by one of the industry giants with their unlimited budgets and thousands of PhDs and engineering staff. Against the odds, an originally non-profit startup, OpenAI, achieved this feat with a "small" team (by Silicon Valley standards) of fewer than 350 people.

Alea jacta est: as Microsoft and OpenAI (now "not-so-open" and for-profit) have formed a thick-as-thieves partnership (a great move by Satya), we can expect bold (organic and inorganic) actions from Google, Amazon, and Meta in response to this competitive threat. Unprecedented VC money is also betting on uncountable new AI ventures, completing a solid flywheel of innovation.

Technological Revolutions: the Past and the Future

Throughout history, various general-purpose technological revolutions have reshaped society and sparked ethical discussions about their consequences and implications. The advent of AI shares striking parallels with several earlier transformations, including the printing press, the steam engine, railroads, electricity, the internal combustion engine, the telephone, radio and TV, semiconductors, computers, biotechnology, the internet, and smartphones. Each of these general-purpose technologies revolutionized different aspects of human life while also raising concerns about safety, privacy, environmental impact, and societal change.

And here we are: the tectonic plates are moving again, and we are seeing the apotheotic 'release of the Kraken' of Generative AI, but there is a lot more on the verge. Not only 'all shades of AI' but also Web 3.0 (Blockchain, Crypto, dApps), IoT, VR/AR, and robotics will be turbocharged by advancements in infrastructure, communication (5G now and 6G, Terahertz, Li-Fi later), and computing capacity and models (advances in GPUs and TPUs, edge computing now and quantum computing later).

Two critical differences between AI and previous revolutions are the accelerating pace of change and this innovation's exceptionally broad and general nature.

The Speed of Disruption

As technology has advanced, the speed at which innovations have disrupted society has increased dramatically. Now, with AI, we face the challenge of an unprecedented exponential pace of change, as the rapid development and implementation of AI technologies have the potential to transform various aspects of our lives in a much shorter time span.

As with the development of electricity, these earlier technological revolutions prompted debates about the responsible use and potential consequences of these innovations, such as the famous "War of the Currents" portrayed by Thomas Edison and Nikola Tesla in the last decades of the 19th century.

Reflecting on the lessons learned from these past transformations can provide valuable insights as we navigate the ethical challenges associated with AI and work to harness its potential for the betterment of society. 

Broad and General Nature of AI

The general nature of AI represents a critical difference from most previous technological revolutions. While earlier innovations often focused on specific domains or applications, such as transportation or communication, AI has the potential to impact virtually every industry and aspect of human life, from healthcare and education to finance and entertainment.

Moreover, AI is not limited to a single type of technology or application but represents a diverse array of techniques and approaches, including machine learning, natural language processing, computer vision, and robotics. This diversity of AI technologies means they can be applied to a wide range of problems and domains, making their potential impact even more far-reaching and transformative.

The Impact of AI on Society and the Need for Ethical Discussions

The rapid development of AI raises ethical concerns, ranging from job displacement and biased algorithms to potential misuse in safety hazards, acts of terrorism, and government surveillance. The increasing autonomy of powerful AI systems has generated intense debate, highlighting the need for society to address these concerns to harness AI's full potential.

By fostering open dialogue and a sound call for action towards local and global regulatory authorities, we can ensure that our pursuit of scientific knowledge is accompanied by the cultivation of wisdom necessary to navigate the challenges associated with AI and other technological advancements. Engaging in these ethical discussions will help us shape a future that reflects our collective values and aspirations and enable us to create a more sustainable, efficient, and interconnected world.

Unstoppable yet Regulatable

Let's face the brutal truth: AI disruption is unstoppable in modern democratic and capitalist societies because of consumer demand, economic incentives, global competition, the democratization of technology, interconnected industries, and societal adaptability. Moreover, with the geopolitical complexities and threats at stake, attempting to halt technological advancement should be seen as a misguided and naive move.

We cannot and should not stop the rapid development of AI. As human beings, we still have enormous and complex challenges to address. AI has the potential to improve people's lives in countless ways, including realizing our dreams for education and healthcare. Human ingenuity is creating a path toward an era of abundance, which may be the only path forward for humanity.

Throughout history, we have repeatedly regulated disruptive technologies, but the pace of innovation is rapidly increasing. Consequently, a tech-based society must adapt and accelerate the pace of regulation to effectively establish guardrails that balance innovation with public interests and maintain trust.

As technology reaches new levels of abstraction, the complexity of its regulatory framework also increases. The exponential number of advancements demands a meta-regulatory approach. This approach would establish a set of overarching principles to guide "fast-and-effective" regulatory processes across different technologies, ensuring that regulatory policies remain relevant and effective in the face of the rapidly evolving nature of technology. And it's unavoidable to think that AI could be an ally in designing and enforcing this new level of dynamic regulation. 

To clarify, I do not propose creating a recursive paradox, as in Asimov's "Runaround" story, by asking a robot to program itself to follow ethical guidelines. Instead, I suggest augmenting human capabilities with AI to develop more effective regulatory frameworks. 

The concept of AI alignment is central to this discussion; it aims to develop AI systems that align with human values and intentions, guaranteeing their actions remain beneficial, transparent, and safe as they become more intelligent and autonomous. By implementing a global regulatory framework involving interdisciplinary collaboration, financial incentives, and legal consequences for non-compliant model owners, along with required safety certifications, we can ensure that AI alignment keeps pace with AI capabilities. AI is unstoppable, yet it must be wisely regulated.

Echoes of Asimov's Vision

As the world continues to embrace the potential of artificial intelligence, we find ourselves in a reality that increasingly resembles the one Isaac Asimov imagined decades ago. 

From the optimistic and entrepreneurial perspective of Bill Gates (see "The Age of AI has begun") to the alarming and pessimistic outlook of Yuval Harari (check "You Can Have the Blue Pill or the Red Pill, and We’re Out of Blue Pills"), it's crucial that we consider the ethical implications of these advancements.

No, I don't think we should head for the hills. At least not yet. As we continue to explore the untapped potential of AI, it is crucial to remember Asimov's words and strive to balance our pursuit of scientific knowledge with the wisdom necessary to guide our actions. 

The AI revolution has only just begun, and the menu of technology capabilities available to us is changing fundamentally. The stakes are higher than ever and we must make sure we steer the course of this odyssey, ensuring that we navigate towards a destination for the betterment of all. With wisdom and courage, we can shape the future we want to see.

Originally published in MIT Sloan Review

---