Gen Discord and the Future of Work

Narayana Murthy, of Infosys, wants employees to work 70-hour work weeks. SN Subrahmanyan, of L&T,? upped that to 90-hour work weeks. Following these comments, there was an immediate media backlash with topics like #smart-work, #quality-time and the need for #work-life balance trending.

What was left unsaid in these media reactions, was that in any job function, you learn to work smart only after years of hard work. If you can spend quality time at work, couldn’t you do the same at home?

Everybody agrees that there must be work-life balance. Balance, for some, is shutting down their laptops at 5pm. For others, it is finding purpose and meaning in work. Work-life balance is not necessarily about the hours worked. It is about attitude; your attitude towards work and to life. Skills can be taught, but attitude is cultural and generational.

Today, Gen Z is entering the workforce. This is the first generation that has used technology and mobile apps from childhood, grew up on social media, and uses AI assists for most common tasks. Some consider Gen Z a privileged generation, digital natives, great at multi-tasking, and completely at ease with technology. Others consider them an employer’s nightmare, prone to behaviors like quiet-quitting and micro-retiring.

Can working longer hours compensate for the attitude deficits? Or, should companies reset expectations and adjust the nature of work to where they can succeed? Is the future of work agentic and automated, with custom AI agents and digital coworkers that perform most job functions? To answer these questions, we as parents, educators, managers, and employers, need to understand this generation better and how new technologies are disrupting learning and the future of work.

Screens and Sensory Deficits

Studies have shown that early exposure to screens can negatively impact child development. Gen-Zers have been exposed to screens since early childhood. Increase in screen times can adversely affect sensory processing, leading to chemical imbalances, sensory overloads, and behavioral problems.

Excessive screen time can result in neurological changes, including premature cortical thinning. This can impair critical thinking and reasoning skills. In the extreme, the immersive experience of the virtual world becomes more exciting than the real world. It can negatively impact fine motor and social skills.

Playing online games has taken the place of traditional childhood activities such as playing with toys, exploring the outdoors, and socializing with friends. Although these games are marketed as a group activity that can improve hand-eye coordination, problem-solving, and social skills, the primary engagement is not with other players, but with the game itself.

The main goal of these games is to maximize engagement time. All the data collected is to ensure that the game adapts to a player’s triggers, giving a personalized experience that keeps them addicted. It is just a matter of time, and before they realize, they are not playing the game, the game is playing them.

Social Media and Reality Distortion

Studies have shown that use of social media is affecting general wellbeing and our mental health. It has been shown to cause anxiety, fuels FOMO, and linked to depression, loneliness, self-harm, and even increased suicidal tendencies.

Social media distorts reality, creating an illusion that all our friends have the perfect lives, relationships, and experiences. Social media filters allow appearances to be changed, blemishes removed, and features altered. Social media algorithms create echo chambers where we see what we believe. These filters and algorithms further skew our perception and make us feel isolated and unhappy.

The dopamine rush of making new friends has been replaced by the simple click of a friend request. The true rush lies not in genuine connection, but in accumulating likes on social media. A lack of social media validation can lead to feelings of depression and isolation. The lines between real and virtual have become blurred for Gen Z, creating a distorted perception of reality that has become their new reality.

Agentic AI and Cognition Offloading

Studies have shown that frequent use of AI tools impacts critical thinking and reasoning through cognition offloading. Why learn grammar, why learn to write, why learn to draw, why learn a new language, when you can prompt the machine?

The calculator didn’t make everybody a math genius. It actually dumbed down the vast majority. Spellcheck and Grammarly didn’t make bad writers good. It again dumbed down the vast majority. Self-driving cars are only making bad drivers worse. When humans collaborate closely with machines to supplement their abilities, the human brain does what it does best – forget skills that it deems redundant or superfluous. In our accelerating pursuit of artificial intelligence, the only casualty is human intelligence.

As we offload reasoning functions to a machine, we slowly lose our ability for critical thinking. It creates knowledge gaps and erosion of skills. These tools empower us, but there is one thing that employers understand well, that a fool with a tool is still a fool.

Gen Discord

We have a generation that is Deficit in Senses, Cognition Offloaded, and Reality Distorted (DiSCORD). Gen Z, or Generation Discord, can be characterized by deficits in senses, cognition, and reality. This discord results in individuals who are entitled, unmotivated, easily offended, unreceptive to feedback, and lacking communication skills and a strong work ethic.

Before they could talk, we put an iPad in front of them. To give them an edge, we put them in schools that were leading adopters of technology and EdTech. Today, we are scrambling to introduce AI tools into schools, without preparing them with the ability to discern or challenge the outputs from these tools. Fueled by corporations preying on parental guilt, we have created a generation that is delusional, depressed, and dumber than their parents.

In 2016, Nobel Prize winner and AI pioneer Geoffrey Hinton predicted that radiologists would be out of a job in 5 years as machine learning would outperform them at their jobs. In 2024, Jensen Huang, CEO of Nvidia, predicted that programmers would be out of a job and kids need not learn to code, as AI would generate code using human language prompts.

We prepared Gen Discord for this future, where all work could be done by prompting a machine. We have created the workforce for that future. Unfortunately, that future has not arrived.

Jobs of the Future

We are told, the future is automated. A few years back, this future was to be built programmatically with Robotic Process Automation (RPA). RPA is software that automates repetitive tasks by mimicking human actions. Today, we have written off RPA and the future we are told is Agentic AI, where autonomous digital agents and digital coworkers collaborate and perform most common job functions. Is there a way that we can visualize this Agentic AI future to understand better the jobs of the future?

There are domains where AI models have excelled and outdone traditional programming. Stockfish is a chess engine that originally used brute-force computing and coded heuristics. In 2020, the evaluation function of Stockfish was replaced by a neural network evaluation function. The brute-force of traditional analysis using chess theory heuristics combined with the neural network’s positional knowledge of millions of previously played games, pushed it to an Elo rating of 4200, almost a 1500 point advantage over today’s grandmasters.

Stockfish can also synthetically generate millions of new games by having it play against another instance of itself, and learn and get better using a process called Reinforcement Learning. In 2023, Stockfish removed the coded heuristics and moved to a fully neural network based solution.

In chess, the no-code AI model has outdone the programmed model. Can we apply this no-code AI model technique to other real-world problems and create better solutions? Is the future a matter of renting a custom bot by the hour and having it do our work?

Chess is a well understood game with well-defined rules, and with players having complete information. It is also easy to create synthetic data by having the bots play against each other.

This is not true for most real world problems. In many cases, data is not available; where available, it may not be easily accessible. Even where it is available and accessible, data is rarely complete, consistent, or current.

AI solutions need large amounts of data. This data has to be manually generated, and tagged and manually annotated. These AI jobs of today are done by unseen low-paid workers, in remote parts of the world, repetitive soulless work, done without context, meaning, or purpose. Millions of workers, manually tagging and annotating images, videos, and text, which are then ingested into an AI model.

Despite the huge costs, every technology solution today is marketed as AI. Do these AI solutions offer a better return on investment compared to a programmed solution?

To Code or Not To Code?

The distinction between a programmed solution and a no-code AI solution is best illustrated through an example. Let's say that we are given the images shown below: a frowny icon, a human face, and a clock. Our task is to transform the frowns to a smile.

An image editor is a software program that provides high-level graphical functions (e.g. flip, rotate, drag, and distort) and allows for pixel-level editing. Flipping the frown to a smile on a symmetrical frowny icon is a relatively simple task for a novice designer. However, performing similar changes on an asymmetrical human face requires the expertise of a skilled designer. To alter a 'frowning' clock into a 'smiling' clock requires some imagination. However, even a child could determine that the solution involves moving the clock hands from the 8:20 position to the 10:10 position.

The results achieved using Adobe Photoshop and the Adobe Photoshop’s Smart Portrait Neural Filter extension are shown below.

The AI solution flags an error for the frowny icon and the clock. It successfully detected the human face and transformed the frown to a smile. However, it also distorts the nose, the glasses, and adds artifacts around the head.

To train the AI model to detect frowns and smiles, a dataset of over 100,000 pairs of images appropriately tagged must first be uploaded. This training process maps the images into a high-dimensional vector space called a latent space. In this latent space, similar images are grouped together, where the images tagged ‘smile’ form one cluster and those tagged ‘frown’ form another cluster.

Programmed solutions also have a data model, and the transformations are done using well understood math functions. A flip, rotate, or distort are coded math functions on the data. The AI solution, when doing the transform, is also executing a very high order math function. The difference is that the AI solution discovers this math function through a process of regression analysis, using a statistical technique called an Artificial Neural Network.

While the AI solution discovers the math function, it cannot express the math function notationally. What it can do is navigate a path through the latent space from the frown cluster to the smile cluster to change a frown into a smile. With the right training data and appropriate tagging, the same process that can change a frown into a smile, can also change your age, hair, or clothes.

Whether it is transforming or generating new media or code, what AI does is give a reasonable starting point. An expert designer can use the AI output image, select just the lips and apply it back to the original image. AI boosts productivity for a good designer and a good coder, but you still need a human in the loop. The programmed image editor is a general solution and still very much needed to finish off the task.?

Everything Everywhere All at Once

Manu Bhaker is a professional female Indian sport shooter. Manu also happens to be common as a boy’s name in India. Many auto-generated news reports from the Paris Olympics frequently used the pronoun ‘he’ for Manu Bhaker. In several AI models, Manu Bhaker was a boy.

Manu might have remained a boy in the virtual world, if not for a, more recent, controversy around her not being nominated for India's top sports award. As more human-written articles were posted about her, the models adjusted their weights and Manu Bhaker became a girl again.

The only thing worse than people talking about you, is people not talking about you. When synthetic content overwhelms human-written content, AI models become self-affirming. They start believing their own truths. The information entropy does not change, but the confidence in its answers multiplies. When new information is overwhelmingly machine-generated output, a model will bias itself to its own hallucinations.

Gabriel Garcia Marquez in his book One Hundred Years of Solitude wrote,

He had had to start thirty-two wars, and violate all his pacts with death and wallow like a pig in the dunghill of glory, to discover almost forty years late the privileges of simplicity

Most humans will pause on reading this sentence; it may trigger a chain of thought, and then, subconsciously, file the associated reflections, imaginations, and emotions in a remote part of the brain. A Large Language Model (LLM) does not have the equivalence of a human pause. For an LLM, it is just a sequence of tokens that get filed away in a numerical vector space, where similar tokens are clustered together. Unlike a human brain, the LLM does not reflect, imagine, or emote. It trains and it infers.

The Book of Samuel, Chapter 2 verse 8 reads,

He raises up the poor out of the dust, and lifteth up the beggar from the dunghill, to set them among princes, and to make them inherit the throne of glory

An LLM will cluster the tokens for both sentences together as the text ‘dunghill of glory’ are found in both, and is a rarely used figure of speech. Gabriel Garcia Marquez was writing about the hopelessness of wars. Hannah’s prayer in the Book of Samuel is about gratefulness. Humans understand the difference in contexts. LLMs are excellent at syntax, but semantics requires context, which, to infer, would require significantly more instances of usages of this figure of speech.

Artificial neural networks used the brain as a model, mimicking the structure of neurons to create interconnected nodes that process information in layers. Today, we are explaining the working of the brain using artificial neural networks as the model. We have come full circle. We are using a model that we do not understand to explain another model that we do not understand and drawing equivalence. If the brain is intelligent, then it must necessarily follow that the artificial neural network is displaying similar intelligence. Is intelligence just a high-order mathematical equation that we cannot express notationally?

In 1930, John Maynard Keynes wrote an essay Economic Possibilities for our Grandchildren about the level of economic life a hundred years into the future. He predicted that technological? advancements and the accumulation of capital would reduce the margin between production and consumption of material things. Economic prosperity would come when supply meets demand. This economic prosperity would, however, come at the cost of technological unemployment, where efficiencies in labor would outpace finding new uses for labor.

Economic prosperity and technological unemployment would, he predicted, lead to fifteen hour work-weeks. The challenge, in the 2030s, would be to fill our leisure time. He proposed that people should prepare for it by experimenting in the arts of life and in activities of purpose.

Keynes could not have anticipated the digital revolution, the abundance of digital content and our insatiable appetite to consume it. Mindlessly scrolling, rewarded with dopamine, fueling addiction and endless consumption. Our challenge today is managing time in the face of this overwhelming abundance of content. When everything everywhere can be accessed all at once, we are constantly busy and never bored.

Threading the Future Needle

The World Economic Forum’s Future of Jobs Report 2025 lists Analytical Thinking as the top core skill for employers. Additionally, Resilience, Communication and Collaboration, Creative Thinking, and Self-Awareness round out the top five core skills.

Over-reliance on AI tools diminishes analytical thinking and social skills. Excessive screen time and social media usage reduce resilience, self-awareness, and the ability to think creatively. How do we create this workforce that industry needs?

Parents have to decide if the technology edge that they want for their kids is worth the costs. Educators have to decide if we need more AI labs, or more tinkering labs and maker studios in our schools. Companies have to decide if they are getting their return on investment from their AI tools as compared to programmed solutions. Countries have to decide if they want to continue exporting data and importing intelligence.

An AI model that goes through multiple cycles of data ingestion, training, and inference will never get to artificial general intelligence (AGI). It cannot achieve singularity, unless we can teach it what it means to be bored, to pause, reflect, imagine, and emote. It is when you are bored that you notice an apple falling down or the water spilling out of a bathtub.

There is one lesson technologists never appear to learn, that a professional is much more than what they produce. A teacher does more than deliver lessons. A radiologist does more than scan images. A programmer does more than generate code. A designer does more than create images. The reports of the imminent demise of these professions has been greatly exaggerated.

Amidst all the AI evangelism and techno-prophet noise, there is one truth that cannot be denied, that an AI strategy cannot succeed without a data strategy. Even the best AI model is unreliable without good data. AI models need large amounts of data to validate and train the models. What transformed AI research was ImageNet, a curated and properly tagged dataset with millions of images, which was used to test and improve the accuracy of the models.

Countries that will lead the AI race are the ones that have outlined a national data strategy. When digital access is inclusive, when data models are open and accessible, when data barriers fall, it will create sustainable businesses, resilient communities, and responsive governments. The most transformative changes happen when we have the right digital public infrastructure.

Read More

On why imitation is not intelligence – https://hawkai.net/digital-transformation-guaranteed/

On Digital Public Infrastructure – https://hawkai.net/digital-public-infrastructure/

Gen-Zers apparently have low attention spans and rarely read long-form content. If you are a Gen-Zer and are reading this, kudos to you! You have a bright future. Like and comment below to celebrate your accomplishment.

Credits

• Chess bot images are from https://chess.com
• Portrait of the stern-looking man is from the painting American Gothic (by Grant Wood)
• Header image was generated in Adobe Firefly using the prompt – Agentic AI future with robots and people facing giant brain
• Google Gemini's assistance was utilized in the composition of this article.

Hawkai Data provides a Customer eXperience Platform (CXP) to quickly prototype, operationalize, and scale applications and services. Start your digital transformation today and create new business and customer experiences using Hawkai Data CXP.

If you have any questions, talk to us at [email protected], or follow us on LinkedIn at https://www.dhirubhai.net/company/hawkai-data/, or connect with us at https://hawkai.net.