XMAS PRESENT #1 - GENAI BRINGING US TELEPATHIC UNDERSTANDING 2025

With the GenAI revolution, sentiment analysis has reached new heights, allowing us to "read between the lines" in digital communication. In this article, we will assess the opportunities and challenges this technology will bring 2025, exploring both its potential to transform human interaction and its ethical implications. Can GenAI-boosted sentiment analysis breach personal privacy or help predict needs and prevent burnout? Will it manipulate relationships or uncover hidden talents? We will explore the challenge posed by the 90% of unstructured data that, in a pre-GenAI world, remained largely unanalysed. We’ll dive into the billion-dollar sentiment industry that is tapping into this data and the strategic market potential for GenAI-boosted sentiment engines. Transitioning from the "why" to the "how," we will delve into four key technical projections and, in particular, examine the benefits and risks associated with using synthetic data in this pursuit. Finally, we’ll conclude with an outlook on the future telepathic capabilities of GenAI-boosted sentiment analysis and explore the potential dawn of a new era of human understanding.

A Historical Fascination with Telepathy

As we move from "the information age" to "the intelligent age," the AI revolution in general, and GenAI in particular, has elevated horizontal functionalities to new heights. From enhanced natural language understanding to advanced tools for text generation and analysis, few verticals have remained untouched. It's clear that the shift from pre-GenAI to post-GenAI is no longer a soft data evolution, but rather a dramatic data disruption.

One of the most dramatic advancements regarding these horizontal functionalities involves our millennia-old desire to read others' minds—not just "see" their actions or words, but to "read" their thoughts. That is, to read between the lines of what people "really" mean, what they think, how they feel, and even their intentions.

In this article, I will argue that with GenAI, that day is actually here. We have already seen indications of the possibilities, and we are on the brink of a development involving several full-fledged GenAI-based sentiment engines. These engines will, even without fancy additional hardware (e.g., tomorrow’s AR glasses/lenses), offer the "telepathic ability" required to fully understand who people "really" are, what they "really" mean, and even what they "really" want.

And this is regardless of whether the relationship is between a seller and a customer, a talent and a recruiter, a patient, and a doctor, two friends or colleagues, or two people dating or conversing.

A New Era of "Mind Reading"

Scary? Perhaps. Some of these applications are even being banned at the continent and country level, if it will be even possible to uphold, and to top that of we’ve recently seen some mindblowing examples of GenAI “agency”. On a company level, we've long discussed information on a "need-to-know basis," a view which with the rise of GenAI-powered sentiment engines is getting obsolete. Likewise, research has shown that the average person "lies" 2 times a day, a social lubricant we may now have to live without.

Whether we’re talking about countries, companies, or consumers, we are facing a drastic social change that some may find unsettling or even unpleasant. At the same time, many others speak of the value of "authenticity" and "transparency," seeing the process as liberating.

And here we’re not only talking about well-known commercial digital dinos like Meta ("Share more in order to make the world more open") or Google ("Make the world's information universally accessible and useful"). And not only about those less commercially inclined individuals who generally have "brutal honesty" as one of their core values. But even more like the GNU movement who since the end of the last century have advocated that "information wants to be free" (as opposed to being "expensive"), where we are talking about a truly broad movement.

"When information is generally useful, redistributing it makes humanity wealthier no matter who is distributing and no matter who is receiving" (Stallman, 1990).

And even politically, the value of free information has long been asserted, at least in the Western world, and also here from both political sides. Ronald Reagan, in his famous speech just months before the fall of Soviet and its Berlin Wall, envisioned how "Information is the oxygen of the modern age" (1989). Twenty years later, Barack Obama, in his speech in China, stated, "The more freely information flows, the stronger the society becomes."

Similar words from the most commercial digital dinos, to the most anarchistic digital movements, from the right at least as much as the left – and, of course, not least the global scientific community - all incredibly beautiful, and all quite easy to align with. Not least with the other side being populated by dictators and highly IP-aggressive corporations. Still the viewpoint is something that our moral philosophy guardian, Yuval Harari, lately warned us for as "The naive view of information" (in the most recent book from last month which should need-read by most regardless of stance on the subject).

Which standpoint will be the most viable one, the future will tell. What’s certain, however, is that even a gamechanging technology like AI and GenAI gathering and processing information has – as with any technology - no intrinsic value, neither good nor bad, but rather depends on who holds it, for what purposes and how it’s used. Just like a hammer can be used to kill someone or build a hospital, a powerful GenAI-boosted sentiment analysis, can - in the wrong hands - be used for devastating harm, while in the right hands, lead to liberating authenticity with direct communication and the reduction of both unintentional misunderstandings and intentional deceptions.

Below, we will briefly examine both the need for and the value of such a GenAI-boosted "telepathic" market. We will give a quick overview of where the opportunities will arise, followed by an outline of the technical alternatives, before concluding with a look into the future of this area.

The Challenge of Unstructured Data—Unlocking the 90% of Unanalyzed Data

One of the critical challenges businesses face in trying to understand customers, prospects, markets, talents, employees, and partners is unstructured data. It’s estimated that 90% of all data generated today is “unstructured” (i.e. non-numerical), meaning it cannot easily be handled, analysed and processed with ML and traditional tools of analysis.

The unstructured data includes both text-based and multimodal interactions such as customer reviews, emails, social media, chat messages, phone support, video meetings, and other internal and external sources. Previous ML-related sentiment analysis techniques were based on keywords and word frequencies, often missing the nuances of irony, indignation, warmth, or sarcasm. Words like "fantastic" or "oops" could mean something entirely different depending on the context.

The earlier models were simply not sophisticated enough to grasp the deeper meaning of words and how humans’ express emotions subtly in various contexts.

This is where the tremendous change comes with the current language models (LLMs). By training these models on trillions of words and sentences, we now have possibility to analyze emotional patterns and understand context in ways that never before been possible in history. This means that companies (and, when legally applicable, countries and consumers), with almost telepathic power, can analyze customers' emotional responses to products, services, relationships, and communication in real-time, allowing people to adjust their communication, interactions and offerings based on deeper insights than ever before.

The Economic Potential of Sentiment Analysis—A Billion-Dollar Industry

The dream of telepathic ability is not just an age-old dream of humanity, but also involves immense commercial potential. The market for big data analytics, covering a wide range of solutions for managing and processing copious amounts of data, was valued at $272 billion in 2022 and is expected to reach $655 billion by 2029, with a CAGR of 13%.

The biggest part of this growth is the rapidly expanding AI-market, valued at $136 billion in 2022 and expected to grow significantly through 2030. Within this market, the NLP segment is an area where practical usage is growing, from customer service automation to understanding consumer behaviour, expected to grow from $14 billion in 2021 to $61 billion by 2030. The global market for sentiment analysis is one of the major contributors to this growth, and even before the GenAI breakthrough, it was expected to grow from over $3 billion in 2021 to $10 billion in 2026, with a CAGR of 23%.

This highlights the increasing role sentiment analysis plays in corporate data management by turning unstructured data into strategically, tactically, and operationally valuable insights. It demonstrates how sentiment analysis bridges the gap between big data analytics and AI technologies like NLP, and with the GenAI boost it will become an even more indispensable tool for harnessing the value of today’s data flows.

As this technology evolves from weak signals to nearby “telepathy,” applications will expand exponentially. Understanding customers’ emotions and reactions in real-time gives businesses a clear competitive advantage. E-commerce companies uses sentiment analysis to adjust their product recommendations and improve the customer experience, based on customers’ emotional states and past behaviours. In healthcare, sentiment analysis is used to evaluate patients’ emotional responses to treatments, allowing doctors to adjust care plans and improve patient outcomes. In recruitment, companies can analyze the fit between a talent’s true personality and the corporate culture, as well as candidates' motivation and engagement for a role.

The rapid market growth for GenAI-based sentiment reflects a growing need to act on emotional signals and intentions in real-time. Eventually, this will even lead to “pre-instant-gratification”—understanding people so well that companies can offer them what they want before they even realize it themselves. This makes it economically likely that a dramatically larger number of companies will invest in GenAI-based sentiment engines in the coming years.

Strategic Market Potential for GenAI-Based Sentiment Engines

Turning to the strategic potential f?r the GenAI-based sentiment engines themselves, we must first understand how they will best position their solutions within the growing sentiment analysis market.

As seen above, one of the greatest strengths of GenAI-based sentiment engines will be their ability to analyze and understand vast amounts of unstructured data in real-time. This gives businesses a distinct advantage in understanding their customers' emotional needs and acting on these signals, or helping their customers to understand each other.

At the same time, one of the biggest weaknesses is the excessive cost of developing and maintaining these systems. Sentiment engines require continuous model training and updates, especially as language and communication channels change rapidly.

A potential threat for any innovator climbing into the market, is of course the growing competition from major players such as AWS and Google Cloud, which already offer powerful APIs and services for sentiment analysis. Smaller players may struggle to compete with these well-established platforms, making it challenging to break into the general market, while strategic windows exist within more specific horizontals and verticals.

Product Market Potential for GenAI-Based Sentiment Engines

Diving into a more technical product-market match, my take is that the opportunity will be lying in being the first to combine advanced vectorization, synthetic profile generation, and reinforcement learning (RL) to continuously improve models.

While many companies are using transformer models for NLP, integrating these technologies to enable predictions across multiple areas (e.g., recruitment, customer behaviour, and employee well-being) remains rare. The generation of synthetic profiles using generative AI is also a comparably new technique, not yet widely implemented across all industries.

This, combined with reinforcement learning for model improvement over time, gives these novel solutions the potential to become utterly unique.

GenAI-based sentiment solutions armed in this way will provide significant customer value by reducing decision-making time and costs in complex situations. Pricing for this perceived telepathic ability can be expected to be premium, with subscription models ranging from approximately $1,000 per month for basic features to over $10,000 per month for more “telepathic” enterprise-level features with full reinforcement learning and customized synthetic profiles.

When it comes to competition, major players such as OpenAI, Hugging Face, Google Cloud AI, and Amazon Comprehend already offer powerful NLP tools. However, none of them currently provide a complete solution covering the entire chain from synthetic profiles to reinforcement learning. IBM Watson offers advanced NLP but focuses mainly on text analysis and lacks a fully integrated solution with synthetic profiles and continuous improvements.

History also shows that when it comes to truly disruptive technologies, it is rarely the established giants who succeed in disrupting themselves. Instead, it is often startups quickly moving to scaleups that become the first to think outside the box, aggressively creating new markets. Like we’ve recently seen happen for both processors (NVIDIA) and AI (OpenAI) it is nu unlikely that it also in this case will not be the larger players who will have what it takes to truly innovate, but that history repeats itself and we’ll see another cool breed of ventures popping up.

In summary, the GenAI-based sentiment solutions on the horizon are likely to offer a historically unique capability if they succeed in combining transformer models with synthetic profiles, and reinforcement learning. Customers will benefit from a certain "telepathic" ability, reaping the benefits through improved decision-making quality, faster processes, and higher precision—justifying a premium price.

Today's competition may have certain parts of this solution that give them a right-to-play, but few current actors are likely to integrate all these steps to create a true right-to-win. Time will tell. ??

Four Technical Alternatives – Platforms vs. Plugins, LLM vs. ML

For companies aiming to build a sentiment engine capable of more telepathic precision, there exists a spectrum of technical choices. One simple model for understanding these choices can be to distinguish between whether the solution is platform- or plugin-based, and whether it is built on traditional ML or novel LLM. This evolution from simpler, low-cost ML libraries to sophisticated LLM platforms reflects the growing complexity and capability required to meet diverse business needs.

ML-based libraries and plugins, such as VADER and TextBlob, represent the most basic form of sentiment analysis tools. They are low-cost, often free or available at minimal subscription rates, and designed for ease of use, with minimal setup or technical expertise required. These tools are suitable for small businesses or projects with basic sentiment analysis needs, where quick implementation and limited customization suffice. However, while they offer simplicity and speed, these plugins are constrained by their lack of scalability and limited ability to capture nuanced emotional tones, making them unsuitable for complex or high-volume tasks.

Moving up the chain, ML-based platforms like Lexalytics and Gavagai (in memoorium) represent a more structured, scalable approach, catering to mid-sized businesses that handle structured data. These platforms provide moderate sentiment analysis capabilities, optimized for domain-specific applications such as finance or healthcare. While generally more affordable than LLM-based solutions, they require some level of feature engineering and technical integration. Their strength lies in their capacity to process larger datasets with domain-specific adjustments, yet they still lack the ability to understand the subtleties of human language fully, making them ideal for simpler contexts but less effective for dynamic, multi-layered interactions.

The shift to LLM-based solutions marks a rock hards leap in complexity and capability. LLM-based libraries and plugins, including Hugging Face and spaCy with integrated LLM models, provide a flexible, customizable option that allows developers and businesses with technical expertise to fine-tune sentiment engines for specialized needs. These solutions offer significantly higher sentiment quality if tailored properly, capturing complex emotional tones and context-dependent subtleties. However, this customization comes at the cost of significant computational resources and setup time. These tools appeal to technically adept teams who require the versatility of open-source frameworks but are willing to invest in their management and optimization.

At the forefront of this spectrum are LLM-based platforms, such as Hume AI and OpenAI GPT. These platforms offer highly scalable, cloud-hosted solutions with advanced natural language processing (NLP) capabilities that can handle vast datasets and process complex, nuanced sentiment data in near real-time. Their API integrations are designed for easy deployment, making them accessible to larger enterprises seeking comprehensive sentiment analysis across multiple domains. However, these platforms are resource-intensive, with high associated costs and technical demands. They deliver unparalleled accuracy and contextual understanding, making them the optimal choice for companies requiring robust, domain-agnostic sentiment engines capable of responding to a wide range of business needs.

In summary, as sentiment analysis technology progresses from traditional ML libraries to advanced LLM platforms, companies must weigh factors like cost, scalability, integration complexity, and sentiment quality. Each solution offers a unique balance, with ML libraries favoring simplicity and low-cost accessibility, ML platforms supporting structured but domain-specific applications, LLM plugins providing customizable flexibility, and LLM platforms delivering comprehensive, high-quality sentiment analysis at an enterprise scale.

But this is only the first high-level decision. While GenAI gives us an unparalleled ability to analyze unstructured data, we still face important technical and ethical challenges, not least regarding the use of synthetic data and feedback loops to improve models.

Synthetic Data – Necessary Catalyst or Potential Risk?

Synthetic data has become an indispensable catalyst for strengthening AI models in general and GenAI in particular, especially when sentiment engines need to handle large datasets or when access to real data is limited. By generating synthetic data, companies can expand their training databases and thereby improve AI model performance.

This is particularly useful in contexts where GDPR or other privacy regulations limit access to sensitive information such as personal data. Especially within sentiment analysis, where emotional nuances are critical, synthetic data could in theory enhance AI models' ability to process relationships between entities like customer and seller, employee, and manager, or even patient and doctor.

However, there are risks. If synthetic data is used without sufficient grounding in reality, models can become overfitted and begin identifying patterns that don't exist in the real world. This risk should be particularly prominent in sentiment analysis, where subtle emotional signals in real data can be difficult to simulate with synthetic datasets. By balancing the use of synthetic data with real insights, models can avoid overfitting and remain useful in practical applications.

An important application of synthetic data in sentiment engines is the generation of synthetic profiles that represent specific scenarios, such as a customer who is close to making a purchase, a talent who is a good cultural fit for a company, or an employee showing signs of burnout. These profiles are generated using GenAI models such as GANs and can be used to match real user data to these scenarios through vectorization methods like cosine similarity or Euclidean distance. In this way, synthetic profiles can add additional layers of insight and create better decision-making tools for businesses.

Using synthetic data for classification and profile generation may thus offer many advantages, such as avoiding biases present in empirical data and allowing the exploration of hypothetical scenarios. At the same time, it's important to recognize the limitations—synthetic data can never fully replace real data because it lacks the unpredictable complexity and dynamics that only empirical data can provide. Therefore, a preferable methodology for ensuring generalizability and relevance would be to use the processing of synthetic data as generating hypothesis, while securing empirical validation through feedback loops.

By continuously testing models with real data, companies can ensure that sentiment engines are not only tailored to current model’s use cases but also remain relevant over time, as language and social interactions evolve.

My conclusion would be that synthetic data will become a powerful tool in sentiment analysis, especially when real data is difficult to obtain or contains bias. By combining synthetic data with empirical validation in continuous feedback loops, future telepathic businesses can build robust sentiment engines that capture both generalizable patterns and complex emotional nuances, leading to deeper insights and better business decisions.

Towards a New Era of Human Understanding

We can conclude that with the GenAI revolution, sentiment analysis is reaching new heights, allowing us to "read between the lines" in digital communication. Still the question remains whether this breach personal privacy or help predict customer needs and prevent burnout? Will it lead to manipulation in recruitment or discover hidden talent? Will it make relationships more artificial or help us recognize romantic signals that humans miss. Overall - will it promote honesty and authenticity, or intrude too far into our personal and professional lives?

Based on the assessment, my take would be that while the future of sentiment analysis requires a balance between technological advances and ethical considerations, it is also full of possibilities. With the development of GenAI-boosted sentiment analysis, we will now actually have the tools to analyze human emotions and intentions in ways that were previously unimaginable. This opens the door to new telepathic ways of understanding and communicating, both within companies and between individuals.

However, with this power comes responsibility. Earlier tools like Chrystal knows and Whitebridge has been revealing enough, but the “mind reading computers” now coming is something totally different. The technology must thus be used in ways that protect individuals' privacy and autonomy. GenAI-boosted sentiment analysis, if misused, could become a tool for manipulation or invasion of personal privacy. Therefore, it is essential that sentiment engines are developed with clear ethical standards and guidelines regarding their use.

Finally, when discussing the future of GenAI-boosted sentiment analysis, we return to the question posed at the beginning of this article: Can AI give us true telepathy? In a sense, we can say that we are closer than ever to understanding each other on a deeper, more intuitive level, but the real "telepathy" lies not in reading people's minds. It lies in our ability to use technology to understand and communicate better—to create authentic and positive interactions in an increasingly complex world.

With the rapid development of technology and major advancements in GenAI-based sentiment analysis, we now have the potential to transform how businesses, society, our relations and even ourselves function. It is thus nothing but an existential question. A psychologist friend consulted during the research for this article even went as far as to call it “The next stage of human evolution.” Scary as hell, but at least as cool. The future must thus be handled with care so that we can wield these powerful tools in a way that benefits all parties involved. But the future also looks very, very cool ?? Alea iacta est.

Rufus Lidman, Fil. Lic.

Lidman is founder and board member of half a dozen ventures, the most recent being AIAR EdTech in Singapore, using cutting edge technology to reinvent learning for millions of needy in emerging markets, as well as chairing Digitalenta in Stockholm, fast becoming the leading talent acquisition venture within digital resources in the nordics. Prior to launching Lidman spent years as digital strategist in four continents for over 100 companies such as Samsung, IKEA, Mercedes, Electrolux, PwC etc. As an entrepreneur he has run half a dozen ventures with 2-3 ok exits, incl. two-fold award winner of growth (Gasell) and some of the world's largest apps in its areas with over 15 million downloads. He has been founder of IAB, digital advisor for WFA, is a top tech influencer with 50,000 followers, a recognized speaker with over 300 lectures, has published 4 books and the world's largest learning app in digital strategy loved by 200.000 people in 165 countries. Lidman received dual degrees in business administration and data statistics from Uppsala University, complemented with PhD studies and data science. For more inspiration feel free to visit www.datadisruption.ai