After spending a year immersed in testing Generative AI (artificial intelligence) models, gathering data, and deploying solutions, I’ve encountered numerous assumptions about this specific technology of AI. Some believe it will entirely replace humans, while others consider it flawed or intimidating. Over the past year, we subjected our models to extensive validation. We used diverse datasets to examine biases, tested how well the AI responded to unusual situations and established human checkpoints for critical decisions.
Here are the key lessons I learned about generative models, hopefully clarifying much of the hype.
1. AI Won’t Simply Replace Humans
One of the loudest concerns about AI is that it will entirely replace humans in the workplace. While it’s true that AI can automate repetitive tasks, fill gaps in manual processes, and even operate with minimal oversight, this does not automatically signify the end for human employees. In my experience, AI is not capable of completing a large project without human assistance, but it is excellent at focusing on a specific project/tasks with human supervision. AI cannot be relied upon to make critical decisions, and that won’t happen anytime soon (I’ll explain why later).
When AI is used for low-risk tasks—like consultations, identifying potential fraud, or recommending a playlist—mistakes are more forgivable, and AI can perform much better than humans. However, in high-stakes fields such as healthcare, finance, and aviation, a single error could have serious consequences. In these scenarios, AI’s outputs must be continually monitored and validated, and generative models that are causing a lot of buzz today are not suitable for that.
Pros:
- Frees humans to focus on critical thinking and creative projects.
- Can reduce boredom by removing repetitive tasks with atouch of creativity.
- Helps companies be more efficient.
Cons:
- Some jobs may change significantly or fade out if they rely mostly on routine tasks.
- Workers need new skills to manage or collaborate with AI systems.
2. AI Doesn’t Know Everything
There’s a common misconception that Generative AI, by definition, knows everything in every domain it’s applied to. Part of this belief came from AI’s impressive ability to process vast amounts of data and detect patterns beyond normal human capacity. While AI can certainly be powerful, it isn’t an all-seeing oracle.
The reality is that AI systems operate only within the bounds of their training data and design, simulating human behaviour based on statistics and mathematics. They do not understand the meaning of words, data, topics, behaviour, or emotions and lack any inherent awareness of polite or impolite behaviour; all AI are merely models. Whenever an AI encounters inputs it has not been trained on, it makes errors. Therefore, an AI’s “knowledge” is limited by the data and parameters provided by humans. Additionally, when an excessive amount of data is given, it may begin to confuse, hallucinate, and skew results.
Generative AI always answers questions somehow, but that doesn’t mean it provides the best answer. It is important to set expectations and understand what right or wrong means for us. Due to the vast amount of data, there are many answers for every single question, and AI is not in a position to find the best and most accurate answer today.
Pros:
- Excellent performance in focused tasks (e.g., customer support, analyzing data).
- Can help humans spot patterns and trends faster.
Cons:
- Limited to what it has been trained on—no “common sense.”
- Requires regular updates and retraining to stay accurate.
3. Will AI Take My Job?
The fear of job loss due to automation has dominated discussions about AI for years. In certain roles, especially those involving highly repetitive tasks (like data entry or assembly-line routines), the rise of AI may reduce manual labour. By combining robots, these capabilities will extend beyond the digital realm and become part of our everyday reality. We should consider that the job market is constantly evolving, and this represents another revolution, similar to the industrial or IT revolution, which enhanced human job positions and minimized dull, repetitive tasks, consequently creating many new opportunities. Today is no different, and new opportunities are emerging while some traditional positions are retiring.
Pros:
- Opening fresh opportunities in many fields and new skills.
- Reduces human effort for repetitive or hazardous tasks.
Cons:
- Could displace jobs focused on routine or manual processes.
- Employees need to learn new skills to stay relevant in evolving industries.
4. Can (or Can’t) I Trust AI?
Trust in AI isn’t about blind belief—it’s something that must be earned through rigorous testing, continuous evaluation, and careful deployment. Over the past year, we subjected our models to extensive validation, leveraging diverse datasets to identify biases, testing AI’s responses to edge cases, and integrating human oversight for critical decisions.
Through this process, I came to a crucial realization: the very question of “trusting AI” is flawed. Generative AI is not a conventional tool that delivers deterministic outputs like traditional software. Instead, it operates on probabilities and pattern recognition, which means we must adjust our expectations accordingly. Trust in AI is context-dependent—it requires understanding how it works, where it excels, and where its limitations lie.
Generative AI is built for creativity, idea generation, and problem-solving rather than rigid precision. By design, GenAI thrives on non-deterministic behaviour, exploring multiple possibilities rather than providing a single, fixed answer. In this landscape, trust isn’t about expecting absolute accuracy, it’s about ensuring alignment with intended goals, ethical safeguards, and responsible usage in a world where AI adapts, evolves, and generates new possibilities.
Pros:
- Adaptability & Innovation, providing creative solutions and diverse perspectives rather than rigid, predefined answers.
- Context-Dependent Trust: Encourages a nuanced understanding of AI, where trust is based on testing, validation, and oversight rather than blind faith.
- Bias Detection & Evaluation: Recognizes the importance of bias testing and human oversight, making AI safer and more reliable.
Cons:
- Non-Deterministic Behavior Can Be Unpredictable: AI doesn’t always produce the same output for the same input, which can be problematic for tasks requiring absolute consistency.
- Complexity in Understanding AI Decisions: Since AI works on probabilities and patterns, explaining why it made a particular decision can be difficult, making transparency challenging.
- Potential for Misinformation: If not properly constrained, generative AI can produce misleading or incorrect information, requiring extra safeguards.
Footnote:
Thank you for your time in reading this article. It is based entirely on my personal observations and experiences with leading Generative AI initiatives. I will elaborate on more topics in upcoming articles and provide my suggestions.