Gen AI: Beyond and Above the Hype
Enrico Santus

Gen AI: Beyond and Above the Hype

The opinions expressed in this article are solely my own and do not reflect the views or stance of my employer.

In a recent report by Goldman Sachs bearing the provocative title "GEN AI: TOO MUCH SPEND, TOO LITTLE BENEFIT? ," the curtain is pulled back on Generative AI, revealing a landscape that diverges starkly from the typically exuberant narratives commonly being spun in the media.

Amid predictions that investments in AI could soar to over a trillion dollars in the near future, this report draws together the opinions of industry experts and academics to ponder the true worth of this technological leap.

Professor Daron Acemoglu of the Massachusetts Institute of Technology (MIT) offers a particularly pessimistic perspective, estimating that it will be economically feasible to automate only a scant 4.6% of work tasks in the decade ahead using generative AI. His analysis, which is premised on the assumption that only a fraction of all tasks are amenable to AI and that even fewer could be automated cost-effectively, forecasts a modest uplift in productivity and a cumulative GDP growth of merely 0.9% over 10 years.

Though Acemoglu's viewpoint is certainly overly conservative — dismissing even the task of summarization as a viable Gen AI application — it opens the way to more reasonable reservations voiced by analysts at Goldman Sachs. These concerns touch on the barriers to AI’s value, including (1) the steep costs of adoption due to the required human-computation ecosystems; (2) the risk of efficiency gains being eroded by competitive market dynamics; and (3) the logistical and technological hurdles in hardware production and power infrastructure.

There’s a palpable risk that a significant portion of the investments in AI might not meet their intended goals, largely due to companies succumbing to a Fear Of Missing Out (FOMO), propelling them into the AI fray without the necessary leadership, human talent, computing power, and data resources.

Yet, there lies a silver lining for those entities that approach this paradigm shift with the requisite strategic foresight. By meticulously crafting the appropriate ecosystems, they could well be the architects of Industry 4.0, heralding an era characterized by adaptive processes seamlessly orchestrated through human-AI collaboration.

The Magnificent Seven (i.e., Amazon, Apple, Google, Facebook, Microsoft, Alibaba, and Tencent) stand out as prime examples of successful AI adoption. These tech giants have demonstrated remarkable strides in leveraging AI for various purposes, such as enhancing customer experience through personalized recommendations and improved service, optimizing operational efficiency in areas like inventory management and supply chain, monetizing data for targeted advertising and financial analytics, enhancing security by preventing and responding to breaches and fraud, and fostering innovation through research collaborations with startups, academia, and industry partners.

As someone with over a decade of experience studying and working in the field of AI and computational linguistics, let me try to provide some insights on the current and future state of AI.

Artificial Intelligence through Time

Having dedicated more than 12 years to the field of artificial intelligence, I have witnessed a significant shift take place. From my studies in computational linguistics at the University of Pisa (where my goal was to model the structure of language) to my work developing grammar checkers for Microsoft Office (in seven languages, including English, Portuguese, German and Italian), and then transitioning to pursue a Ph.D. and several Postdocs in ontology generation, healthcare, pharma, finance, and fake news detection across two continents (Hong Kong Polytechnic University, Singapore University Technology and Design, and Massachusetts Institute of Technology), I have witnessed nothing short of a remarkable evolution of this discipline. And, in the last 5 years, I have worked with talented colleagues to push this evolution into products, first as Director of AI in the pharmaceutical sector (Bayer) and now as Head of Human Computation in the financial industry (Bloomberg).

Throughout my career journey, artificial intelligence has transitioned from a niche interest confined to the laboratories of engineers into an ever-present force that shapes our daily existence. Today, AI is not just a topic of academic fascination but a lived reality, from the voice assistants that greet us each morning to the algorithms that curate our digital experiences and the autonomous vehicles that herald a new era of transportation. Yet, as expansive as this journey has been, we stand on the brink of even greater advancements, that — in my opinion — were already embedded in the origin of the term "Artificial Intelligence."

When John McCarthy coined the term during a workshop at Dartmouth College in 1955, not all of his contemporaries were enamored with the term. Some proposed alternatives like "automata studies," "complex information processing," and "advanced automatic programming." While descriptive, these alternative names lacked the visionary appeal of "Artificial Intelligence," which suggested the almost mythical potential for humans to create a form of non-biological intelligence. This naming moment played a crucial role in shaping the public's imagination and the field's trajectory. I believe it is fair to say that the future of AI is not just written in code, but also in the words we use to speak about it.

Is AI replacing our jobs?

As we navigate through these technological advancements, the specter of AI looms large, casting long shadows over the future of human labor. With the advent of deep learning and large language models, a chorus of voices has risen, oscillating between awe and apprehension. The media’s richly embroidered narrative paints a picture of a world teetering on the brink of full automation, where machines might soon render the human workforce obsolete.

Yet, amidst this clamor lies a stark discrepancy between perception and reality. The conversation, often charged with speculative fervor, begs for a grounding in hard facts. Recent data from the U.S. Census Bureau provides a sobering counterpoint to the notion of an AI-dominated landscape. It turns out that AI's integration into the American business ecosystem is far from ubiquitous. In fact, less than 10% of companies have woven AI into the fabric of their operations. This adoption skews heavily towards the corporate world, with a significant majority of AI-utilizing firms boasting a workforce of more than a thousand employees.

A closer examination reveals that the vanguards of automation inhabit sectors already steeped in technological innovation. Information services lead the charge, with around 20% of businesses in this realm employing AI in some capacity. They are closely followed by the domains of professional, scientific and technical services, as well as educational services, each flirting with the 15% mark. The finance and insurance sector also sees a modest engagement, with just over 10% of its entities harnessing AI.

AI Adoption in the Industry

Moreover, historical employment data paints a compelling picture about technology’s influence on the job market. Recent research from MIT economist David Autor and his colleagues reveals a staggering statistic: 60% of today's workforce is engaged in occupations that didn't even exist back in 1940. And here's the kicker - Autor's findings suggest that more than 85% of employment growth in the last 80 years can be attributed to the emergence of new occupations driven by technological advancements.

This landscape prompts a recalibration of our expectations. The notion of AI as being poised to supplant human endeavor appears, at least for now, to be more myth than reality.

Humans and AI Collaboration

As we navigate the 21st Century, the critical challenge isn't whether AI will replace the human workforce, but how it will integrate and enhance it. Karl Popper's distinction between Clock and Cloud problems provides a crucial framework for understanding this integration. While some challenges are straightforward and mechanical (Clock problems), and can therefore be mostly automated, others are inherently complex and unpredictable (Cloud problems). In the era of Generative AI, addressing these Cloud problems requires moving beyond traditional engineering approaches to embrace a more human-centric, creative, and empathetic perspective.

This new era demands a partnership between AI and human intelligence focused on enriching the human experience by turning the feedback-recommendation loop into simple, engaging, and insightful communication. It requires rethinking the workplace ecosystem — designing AI interfaces that feel natural to users, investing in upskilling employees, and creating an environment where human creativity and machine efficiency work together seamlessly.

Where AI will help the most

So, where can AI help the most? The answer includes tasks ranging from data mining to decision-making, and from monitoring to modeling. Imagine AI systems that streamline information gathering, making knowledge more accessible and actionable. Envision algorithms that optimize workflows, prioritize tasks, and automate administrative tasks. Picture AI as the ultimate assistant, monitoring processes and environments, ensuring compliance, and enabling humans to focus on the strategic and the creative – both uniquely human tasks.

What is AI bringing to the World?

Amid the sensational headlines in the media and the cautious views of certain academics, our mission is to ground ourselves in reality. In the unfolding narrative of Industry 4.0, AI emerges not merely as a tool for incremental improvement but as a harbinger of profound transformation. This revolution is marked by more than just enhanced efficiency and cost reduction; it heralds a deeper, more nuanced understanding of the world around us.

The proliferation of sensors everywhere — from wearables to IoT devices — further amplifies these capabilities, enhancing the volume of observable events and their interactions. This sets the stage for groundbreaking discoveries, thereby unlocking the potential for early detection of environmental shifts, anticipation of market trends, and more.

In its advanced capacity, AI will serve as a powerful microscope, revealing micro-patterns in vast seas of data, long before they surface to human consciousness. Imagine the difference between a flower captured in the lower resolutions of yesteryear versus one in stunning, high-definition clarity aided by AI. This analogy extends beyond mere visual beauty. It also serves as a metaphor for our enhanced ability to perceive and understand the intricate details of our environment and the processes within it.

Same image at low(er) and high(er) resolution

Human-AI collaboration will redefine adaptiveness in processes. This synergy between human ingenuity and machine intelligence creates a dynamic ecosystem, inherently flexible and swiftly responsive to disruptions, market shifts, and evolving client needs. Operations will no longer be rigid structures but fluid entities, morphing in real-time to address immediate challenges and opportunities.

From this fertile ground of adaptiveness and responsiveness springs the bloom of customization. Industries powered by AI will craft experiences, products, and even medical treatments tailored to the individual. It will radically transform how we interact with technology and one another.

Conclusions

Like other major technologies, artificial intelligence is rapidly transforming our lives. While there is a significant amount of investment fueled by hype, only a portion of it will yield value in the short term. The rest will likely result in failures and learnings that will gradually contribute to the so-called "plateau of productivity," which in the long run will benefit everyone.?

Successful innovation requires a clear vision and strategic planning, which then builds atop trust. Impulsively diving in without a strong foundation is a common mistake that leads to the wastage of resources. On the other hand, delaying AI adoption due to fear could hinder opportunities for growth and insight, giving competitors an advantage.

There is no one-size-fits-all approach to success in this dynamic landscape, but we must heed the words of Steve Jobs, who once said: “Innovation distinguishes between a leader and a follower.” There is no room for following in this journey.

In Dante Alighieri's “Divine Comedy,” Ulysses urges his companions to join him on an expedition into the unknown, portraying it as a quest for knowledge: “Consider well the seed that gave you birth: you were not made to live your lives as brutes, but to be followers of worth and knowledge.”

So, stay focused on your purpose, establish the right ecosystem, and move forward with bold innovation!

Disclaimer: Oh, and just so we are crystal clear — this piece was crafted in part by interacting with multiple cutting-edge large language models.

Victor Kovalets

PhD Researcher in Psychology | UCL | LSE Alumni Association | Southampton University | Edtech Founder | Nonprofit

1 周

Thanks for sharing, Enrico!

Sebastian Andersson

Lead AI Solutions Architect at SiloGen

1 个月

I still remember some of your earlier work ?? To me, the discussion around Gen AI is the same as was/is happening in human/machine translation. Study what is still an ongoing process there and you will see what will happen in some/many other fields (in my humble opinion)

Fabio Ardossi

AI Strategy and Innovation | CEO and Board Advisor | DataIQ100 (2024, 2023, 2022) | Help Organizations to Adopt and Scale Next Gen AI Solutions

2 个月

Very well written article! ?? agree. GenAI has been a good hype because easy to consume by diverse roles in organisations, it will still help different business ares but I do expect new techs coming up next year and more relevant to improve orgs efficiency.

Rory Raftery

Head of PV Benefit Risk Management Data Analysis at Bayer

2 个月

Very nice piece Enrico Santus ! The fact is that humans cannot deal with the data explosion of recent times and this will only increase with time, so we need AI to sift this data for patterns, relationships and to summarise it for us. The main challenge I see is that the field is moving so fast with new models and techniques almost every day. We need tried and tested products which are kept up to date with latest best practice, plus easy to deploy and adaptable components such as fine tuned agents which can be directed to only answer questions on specified data sets.

回复

要查看或添加评论,请登录

社区洞察

其他会员也浏览了