A year ago I wrote about my first weeks with ChatGPT. I was excited, but cautious: I could not tell whether it was a spectacular toy or the start of something big. A year later, the answer is fairly clear. In 2023, artificial intelligence went from a dinner-table curiosity to a tool you open in the morning, right alongside your email.
As every December, I am taking stock. This year, my review is almost entirely about AI. Not because nothing else happened, but because AI slipped into everything else: how we write, how we market, how we build products, even how we argue about the future.
From Curiosity to Daily Habit
In early February, estimates quoted everywhere suggested ChatGPT had reached around 100 million users only two months after launch. I cannot vouch for the exact number, but the feeling was right: no consumer app had ever grown that fast. Everyone, from students to bank executives, was trying the same thing in the same month.
What strikes me more than the number is the change in behaviour. At the start of the year, people asked ChatGPT for poems about cats. By the end, they were asking it to summarise contracts, write code, draft proposals and translate presentations. When a tool moves from “look what it can do” to “I can’t remember how I worked without it”, it is no longer a fad.
The Moments That Mattered
If I had to pick a few milestones of the year, my list would look roughly like this:
- March: GPT-4. A visible leap in quality. At the time I wrote that GPT-4 is not the end of jobs, and I still think so, but it changed what “good enough” means.
- Competition arrived. Google launched Bard, Anthropic came out with Claude and then, in July, Claude 2, and Meta released Llama 2, an open model anyone can run. Microsoft brought Copilot into the products we use every day. In December, Google announced Gemini.
- Images. Midjourney v5 made many generated images hard to tell apart from photographs. When I tested DALL-E 2 and Midjourney less than a year and a half ago, the results were impressive but easy to spot. Not anymore.
- November: DevDay. OpenAI launched custom GPTs: anyone can build their own assistant without writing code.
- The OpenAI drama. On 17 November, the board fired Sam Altman. By 22 November he was back. Five days in which a company that millions of businesses rely on stood on a knife edge.
What It Means for Business
The Altman episode strikes me as the most useful business lesson of the year, more useful than any product launch. Many companies had built entire processes on a single AI provider and discovered, over one weekend, how fragile that is. My conclusion: use AI as much as you can, but don’t bet your whole business on one model. Have a plan B, test alternatives, and keep your data and processes portable enough to switch providers.
Second lesson: the advantage no longer comes from having access to AI. Everyone has access, often for a few dollars a month. The advantage comes from knowing what to ask, from your own data, and from how fast you change the way your team works. Say you run an online store: anyone can have auto-generated product descriptions. What nobody can easily copy is how well you know your customers. I collected a few practical examples in 7 things a small business can automate tomorrow.
And third: for small companies, 2023 was good news. Work that used to need a copywriter, a designer and a developer can now be started by one person with a good idea and some patience. It doesn’t replace specialists, but it lowers the barrier to entry dramatically.
What It Means for Society
The year was not only about enthusiasm. In March, the Future of Life Institute’s open letter called for a six-month pause on developing the most powerful models. The pause never came, but the debate began. In May, Geoffrey Hinton, one of the field’s pioneers, left Google so he could speak freely about the risks. When someone who helped build the technology tells you to be careful, it is worth listening.
In Hollywood, writers were on strike from May to September and actors from July to November. One of the central issues was precisely AI: who owns a person’s voice, face and words, and who decides when a machine may use them. To me it looks like the first major labour dispute of the AI era, and it won’t be the last.
And governments woke up. On 30 October, the United States issued an executive order on AI. In early November, the UK hosted an AI Safety Summit at Bletchley Park. And on 8 December, the EU institutions reached a political agreement on the AI Act, the law I wrote about this summer when I looked at what Europe is preparing. Formal adoption is still to come, and the details matter enormously, but the direction is clear.
Who Wins and Who Gets Left Behind
This is the part that worries me most. AI could be the greatest equaliser in decades: an entrepreneur in a small town has, in theory, the same tools as a multinational. But “in theory” is doing a lot of work there. In practice, the winners are those who speak English, own a decent laptop, have time to experiment and the nerve to try.
Those left behind are the people who haven’t yet discovered what these tools can do, or who are afraid of them. Worse, those whose jobs are changing without anyone helping them adapt. A fair society is not one that stops technology; it is one where everyone has a real chance to use it. That means education: in schools, in companies, for people in their fifties, not just those in their twenties. I believe that is a duty for governments and for us entrepreneurs alike.
What I’m Taking with Me from 2023
- AI is no longer a separate project. It is a layer over everything we do.
- Those who don’t experiment now will pay more later, when they have to catch up.
- Depending on a single provider is a risk, not a saving.
- A person who knows what they want still matters more than a tool that knows how to answer.
What I’ll Be Watching in 2024
I’m not making predictions. This year proved that anyone trying to guess December 2023 back in January 2023 would have been wrong. Instead, I have a few questions I’m keeping an eye on:
- What will the final text of the AI Act look like, and how hard will it be for a European startup to comply?
- Will open models like Llama 2 keep pace with the closed ones?
- Will Gemini shift the balance of power among the big players?
- Will custom GPTs become a real app store, or remain a curiosity?
- And most importantly: will schools and companies start teaching people how to use AI, or will we leave it to chance?
2023 was the year everyone tried AI. I hope 2024 will be the year we learn to use it well, and together. How did you use AI this year, and what surprised you most?