I love AI. And that is exactly why I allow myself to be honest about what it actually is, and what it is not. Because hype and fear are two sides of the same coin, and neither of them leads us anywhere sensible.
A personal, critical, and technical look at artificial intelligence.
What is AI, really?
Artificial intelligence is not one thing. It is an umbrella term for systems that can perform tasks that normally require human intelligence: recognising patterns, predicting the next step, generating text, images, or decisions. Behind most modern AI systems lie neural networks: mathematical structures loosely inspired by the structure of the brain, trained on enormous amounts of data.
The AI we talk about today, ChatGPT, Claude, Gemini, and the like, are so-called large language models (LLMs). They are not intelligent in a human sense. They statistically predict the most likely next word, based on all the text they have been fed. That sounds trivial. The result is not.
An LLM does not know what it is saying. But it has become so good at saying it that it almost does not matter, until it does.
Where is AI used today?
AI is already everywhere. Most people just do not notice it. The spam filter in your inbox is AI. Netflix and Spotify recommendations are AI. X-rays analysed for cancer: AI. Navigation systems that predict traffic: AI. Facial recognition on your phone: AI.
And then there is the big wave of generative AI tools, the ones we talk to, draw with, and write with. They came in through the front door and they are not leaving.
AI is used today in:
- Healthcare: diagnosis, drug development, patient pathways
- Finance: fraud detection, algorithmic trading, credit assessment
- Law and justice: case processing, contract analysis, risk assessment
- Military: autonomous drones, intelligence analysis, cyber attacks
- Creativity: images, music, text, film, game design
- Education: personalised teaching, automatic feedback
- Climate: weather modelling, energy optimisation, CO₂ calculation
What AI is good at: the right starting point
AI is extraordinarily good at processing large amounts of information quickly. At finding patterns in data no human would have the patience to go through. At generating a first draft, a starting point, an idea. At translating. At summarising. At helping those who do not know what they do not know.
The right starting point for AI is as an assistant, not as an authority. As an amplifier of human capacity, not a replacement for human judgment. A skilled doctor who uses AI to make sure she does not overlook something is a strong image of technology's potential. A caseworker who blindly trusts an algorithm's decision is a horror story.
AI is a tool. An unusually powerful tool. But tools are harmless. It is the intentions behind them that decide whether they build or destroy.
The critical and the dangerous side of AI
Here I need to turn down the enthusiasm. Because there are real problems, and they are not small.
First: hallucination. AI systems make things up. They generate answers that sound authoritative but are directly wrong. Legal references that do not exist. Medical advice that is dangerous. Historical facts that are invented. And they do it with the same confident tone as when they are right.
Second: bias. These systems are trained on human data, and humans discriminate. So does the data. Facial recognition systems that struggle more with darker faces. Credit scoring algorithms that systematically pull in the wrong direction for certain groups. Recruitment tools that score applicants by how closely they resemble past hires, and thereby reinforce the same patterns again and again.
AI is not neutral. It reflects precisely the prejudices, data, and design decisions built into it. And because the system appears objective, it is harder to challenge than a human saying the same thing.
Third: concentration of power. Five to ten technology companies, primarily American, control the infrastructure behind global AI development today. That is a historic concentration of technological power, and its consequences are not yet fully visible.
Fourth: the job market. AI will not replace all jobs, but it will dramatically reshape them. And historically, it is always those with the fewest resources who are hit hardest in technological transitions.
Why do people love it?
Because it works. For the first time in the history of technology, there is a system you can talk to in your own language, and it responds sensibly. It is not necessarily intelligent, but it is accessible. It breaks down a barrier between human and computer that we have struggled with for decades.
For the programmer: AI writes the code faster than she can think it. For the student: AI explains what the teacher did not have time for. For the entrepreneur without money for consultants: AI is the advisor, lawyer, and copywriter in one. It is the democratisation of access to knowledge, and it is real.
And then there is something much more primitive: it is fascinating. It is almost magical. And humans love magic.
Why are people critical and afraid?
Fear of the unknown is human. But fear of AI is not irrational. It is just often aimed in the wrong direction.
Many fear the superintelligent robot from science fiction. That is not the immediate problem. The immediate problem is more prosaic: that we roll out the technology before we understand it. That we let profit drive the pace. That regulation is behind. That nobody can explain why AI made a particular decision, not even those who built it.
Disinformation is a real fear. Deepfakes are already a problem. Automated manipulation of elections, of discourse, of reality. These are not future scenarios. They are happening now. And they are happening faster than our institutions can keep up.
Grounded concerns, not sci-fi:
- Mass production of disinformation and deepfakes
- Lack of transparency in AI decisions (black box)
- Surveillance and large-scale breaches of privacy
- Automation that increases inequality rather than reducing it
- Loss of human control over critical infrastructure
Security and AI
AI security is a field in its own right. It is about two things that are often confused: cybersecurity with AI, and safety in AI systems.
AI is already used to detect attacks, identify vulnerabilities, and protect networks faster than any human analyst can. That is genuine progress. But the same capabilities are used to attack. To generate phishing emails that are impossible to distinguish from real ones. To find holes in systems at a scale that was not possible before.
And then there is the question of AI systems' own security. What happens when we give AI systems control over critical infrastructure: power grids, water supply, financial systems? What happens when they fail? What happens when they are manipulated? We do not actually know with certainty. And that is a problem.
Research in AI safety, what the field calls alignment, is about ensuring that AI systems do what we actually want, not just what we asked them to do. It is one of the serious scientific problems of our time, and it is not solved.
AI and war
This is the darkest chapter. And the shortest, not because it is insignificant, but because it is heavy enough to carry its own article.
Autonomous weapons systems exist. Drones that choose and attack targets on their own. Systems that analyse battlefield data and recommend, or make, decisions about life and death. This is not the future. It is the present. Israel, the USA, Russia, China, Turkey: all are investing heavily.
No international agreement bans autonomous lethal weapons systems. The UN is discussing it. While they discuss, they are being built. It may be the most pressing ethical and geopolitical challenge AI puts before us.
Who is responsible when an autonomous drone kills the wrong person? Nobody can answer clearly. And that should keep us awake at night.
How we think about it
At Windfeld Jensen Co., we treat AI as one tool among many. Sometimes it belongs in the product. Sometimes it belongs in the workflow. Sometimes it does not belong at all.
We start with the work, the people doing it, and the outcome the business needs. Then we decide whether AI adds clarity or only adds complexity.
That is how we avoid building impressive demos that nobody wants to use six months later.
Is AI the future?
Yes. But that does not say very much. Electricity was the future. The internet was the future. The interesting question is not whether, but which future.
AI is not a force with its own will. It is technology that reflects the choices we make: about who owns it, who regulates it, who has access to it, and what we use it for. The future AI creates is the future we choose to create with it.
I am an optimist of the hard kind. The kind that knows technology does not automatically make the world better, but believes it can, if we are conscious, discerning, and politically engaged enough to insist on it.
AI can help us fight climate change. It can democratise access to education and healthcare. It can free people from tedious tasks and give them time for what is human. It can make us smarter, faster, more creative.
But it can also concentrate power in the hands of the few. It can poison public discourse. It can be used as a weapon. It can erode the foundation of trust that democracies rest on.
The future with AI is not written yet. It is being written now, by the engineers building the systems, the politicians regulating them, the journalists asking questions of them, and the citizens using them, or rejecting them. That choice is still ours. Use it.
