Why this technology is reaching a tipping point now—and what that means for the philanthropic sector
By Marc Broers, Co-founder and CTO FindNFund
When I speak with board members of foundations, I often hear the same sentence: “AI suddenly seems to be everywhere.” That's true—and not entirely true. The technology we now call ChatGPT, Copilot, or Claude stands on the shoulders of seven decades of research, experimentation, and application.
In this article, I'll take you through the history of AI in broad strokes. Not as a history lesson, but as a framework to understand where we stand—and why the philanthropic sector cannot afford to ignore it.
1. The dream (1950-1970): can a machine think?
The term “artificial intelligence” was coined in 1956 at a summer workshop in Dartmouth. A small group of scientists believed that intelligence could be broken down into logical rules —and that a computer could, in principle, do anything a human can.
The first successes were small but remarkable: programs that solved chess and checkers problems, parsed English sentences, or reproduced simple mathematical proofs. Expectations were sky-high. Researchers openly predicted that “within ten years,” a machine would become world chess champion and translate fluently.
The definition I personally find most fitting: AI is a system that performs tasks that normally require human intelligence.
“The definition I personally find most fitting: AI is a system that performs tasks that normally require human intelligence.”
2. The AI winters (1970-2000): promises unfulfilled
Reality proved more stubborn. Computing power was limited, data was scarce, and the approach, capturing everything in rules, turned out to be far too rigid for the messiness of the real world. Funding dried up, and researchers left the field.
Yet behind the scenes, progress continued. So-called “expert systems” found their way into hospitals and banks, supporting tasks like risk analysis. Neural networks, one of the core techniques behind modern AI, evolved from concept to proven method, thanks to the perceptron algorithm. Speech recognition and machine translation gradually became usable.
Simply put: the foundations for scaling AI were laid.
3. The renaissance (2000-2017): data, GPUs, and deep learning
Three developments changed the game. First, the widespread adoption of computers and the internet led to an explosion of data. Second, graphics processing units (GPUs), originally built for gaming, turned out to be ideal for the parallel computations required by neural networks. Third, researchers learned how to train much deeper neural networks to recognize patterns in data—so-called “deep learning.”
Around 2012, a deep learning model won a major image recognition competition by a wide margin. Suddenly, computers could distinguish cats from dogs better than many humans. Within a few years, breakthroughs followed in speech recognition, machine translation, and image classification—not based on human-written rules, but by allowing models to discover patterns in data themselves.
4. The breakthrough (2017-2022): the transformer architecture
In 2017, a group of Google researchers published a paper with the understated title “Attention Is All You Need.” It introduced a new architecture: the transformer. It proved exceptionally good at understanding context in language — and highly scalable: the more data and computing power you feed it, the better it performs.
In the years that followed, an arms race emerged around increasingly large language models. In 2022, OpenAI opened the door to the general public with ChatGPT. Within five days, it reached one million users. For the first time, a broad audience experienced that a machine could not only calculate or search — but also write, reason, and explain.
5. Where we are now (2023-2026): from novelty to infrastructure
Over the past three years, AI has evolved from headline-grabbing demos to serious infrastructure. Models have become multimodal: combining text, images, audio, and code in a single system. They are faster, cheaper, and safer. Organisations are building their own assistants on top of these models, trained on their own knowledge and context.
Where I was still explaining what a “prompt” was in 2023, we are now helping foundations answer a different question: how do we integrate this safely and responsibly into our work? That's a different category of question. That's the question of a mature technology.
The question is no longer whether AI will impact the sector, but how we will use it to create even greater impact together.
“The question is no longer whether AI will impact the sector, but how we will use it to create even greater impact together.”
What does this mean for the philanthropic sector?
For our sector, this history points to three things:
- This is not a hype cycle that will simply fade away. The underlying technology is improving month by month — not declining.
- The question is shifting from “what can it do?” to “how do we use it well?” That is a fundamentally different conversation.
- The foundations that start learning, experimenting, and setting frameworks now will build an advantage that, in three years, will determine whether they are catching up — or leading the way.
At FindNFund, we believe the best technology should reach the sector that deserves it most. That does not mean chasing every hype. It means staying critical, learning together, and applying it responsibly.
