Will AI steal my job? The burning question for creative professionals
The few things we know
Creative professionals have spent the last few years eyeing the towering and ever-growing elephant in the room: will AI steal my job?
The scariest part is that no one can say for sure. Dig into expert forecasts, and you’ll find everything, from zombie-apocalypse scenarios to utopian daydreams—nothing helps you pin down a definite “yes” or “no.”
Conflicting Crystal Balls
McKinsey (2017) envisioned tech-driven job losses by 2030 ranging from 10 million to a jaw-dropping 800 million, while simultaneously predicting 555 to 890 million new roles. If you think that’s a wide range, you’re not alone.
World Economic Forum (2025 Report) shoots for precision, claiming 22% of jobs will be disrupted by 2030, with 170 million new positions created and 92 million displaced, resulting in a net gain of 78 million jobs.
Goldman Sachs (March 2023) suggests 300 million jobs could be automated away, but we can expect new opportunities in the bargain.
Kai-Fu Lee has long warned that AI could replace half of all jobs by 2027, while billionaire Vinod Khosla predicts AI will handle 80% of work in 80% of jobs.
Geoffrey Hinton, one of AI’s foremost experts and a vocal critic of its potential dangers, recently awarded the Nobel Prize in Physics, suggests universal basic income can become a necessity so that wealth doesn’t end up clustered in a few lucky pockets.
On the other hand, many envision a “post-work society” where AI generates prosperity and frees humans to focus on their passions and their “self-actualization”. This vision has been shared, with different nuances, by figures like Ray Kurzweil, Sam Altman, and Bill Gates.
The takeaway? No one really knows what will happen in ten or even five years.
The impact on creative professions
While uncertainty reigns, there is a broad consensus that creative jobs are among those most affected by AI. Some professions will face more disruption than others. Take translators, for example: AI has been gaining ground in this field for years, starting with simple texts and increasingly tackling more complex ones. Will translators disappear? Probably not, as AI outputs will need coordination and refinement by humans. However, there will likely be less demand for translators of low-value-added texts. A similar fate awaits journalists who focus on clickbait content or low-value articles written solely for SEO purposes. For them, creativity and depth will become survival skills.
But it’s not all doom and gloom. Based on what I’ve seen firsthand—having worked in recent years on AI adoption with dozens of colleagues in publishing and other creative fields—there is significant potential for AI to enhance rather than replace creative professionals.
Consider designers who wield MidJourney or other text-to-image tools. The real magic happens when experts know exactly what to ask for. Take these prompts for a book cover:
• “Draw a pirate battle for kids.”
• “A dramatic scene of a pirate ship battle, illustrated in a dark and grotesque style blending Tim Burton with German Expressionism, featuring exaggerated caricatures, a dark-toned palette, and bright neon flashes.”
The second prompt is far more specific and nuanced, resulting in a more compelling output. Sure, ChatGPT can write detailed prompts for Midjourney, perhaps even better than most humans. But a trained designer knows exactly what to request to achieve a particular style and can interpret the results with an expert eye.
It’s the same with other fields: without expertise, even the most advanced AI can be a mediocre tool. If I asked AI to solve a complex mathematical theorem, I wouldn’t know how to evaluate its accuracy or apply it, but a mathematician could. Similarly, in creative and intellectual domains, competence remains irreplaceable—at least for now.
What lies ahead?
As history has shown, in industrial and technological revolutions, job destruction has always been counterbalanced by the creation of new professions, often in areas we could never have imagined. Will this happen again with AI? We can’t say for sure, but we can certainly hope so. One thing is clear: those who embrace and master these technologies will be better equipped to shape the future, rather than be shaped by it. And who knows—perhaps enormous opportunities await for those creatives who learn how to harness them.
What the next ten years will bring remains uncertain, but one thing is certain: it’s going to be a wild ride.
Benchmarking tools to evaluate AI even if you are not a developer
How do you choose your LLM to use? Access to benchmarking tools is crucial because they help evaluate and compare AI models across key dimensions like performance, efficiency, safety, and fairness. These tools provide standardized metrics, enabling informed decisions about which model fits your specific use case. Without these tools, it’s challenging to objectively measure trade-offs between accuracy, speed, scalability, and ethical considerations. It’s challenging even to choose the right tool to use.
1. LMSYS Chatbot Arena (LMArena)
Characteristics: LMArena is a platform for comparing large language models (LLMs) interactively through chat-based evaluations using Elo rating systems and crowdsourced feedback.
Pros:
Free to use without sign-up requirements.
Transparent data and open participation.
Cons:
Limited to conversational AI evaluation.
Subjective human feedback may introduce bias.
Best For: Comparing LLMs in conversational scenarios, even if you are not a developer but just curious
2. POE
Let’s say that you don’t want to have an environment that looks like too nerdy. Then you have to choose something else. Poe AI, developed by Quora, is an aggregator platform that allows users to interact with various LLMs, such as GPT-4, Claude, Llama, DeepSeek, and others. It also supports custom bot creation for specific tasks and provides a marketplace for sharing and monetizing these bots.
Pros
Easy to use
Convenience: Centralized access to multiple AI models eliminates the need to switch between platforms.
Flexibility: Users can route queries to the most suitable model for their task.
Customizability: The ability to create bots for niche tasks adds versatility.
Comparative Analysis: Side-by-side comparisons help users evaluate model strengths and weaknesses.
Cons
Variable bot quality: Marketplace bots may vary in performance since they are user-created.
Limited free access: Advanced models like GPT-4 or Claude 3 require a paid subscription.
Potential bugs: As the platform is still evolving, occasional technical issues may arise
3. Hugging Face
Characteristics: Hugging Face offers pre-trained models for NLP tasks with easy-to-use APIs and evaluation scripts.
Pros:
Simplified fine-tuning and evaluation process.
Large repository of pre-trained models.
Cons:
Some familiarity with Python is helpful but not mandatory.
Best For: Non-developers working on text-based AI tasks who want quick results.
4. Weights & Biases (W&B)
Characteristics: W&B is a cloud-based platform for tracking and visualizing machine learning experiments.
Pros:
Intuitive dashboards for real-time metrics tracking.
Minimal setup for basic usage.
Cons:
Advanced features may require technical knowledge.
Best For: Teams or individuals needing easy visualization of AI performance.
⚠️ Whichever benchmarking tool you choose, carefully read the terms and conditions of use