AI bootcamp for EU affairs, day 2: your AI dictionary
AGI may arrive inside this Commission’s mandate. Before you can judge the hype or the risk, you need the basics: narrow and general AI, the layers inside artificial intelligence, and why every model comes down to prediction.
Sebastián Rodríguez Pérez is a European campaign strategist with direct experience running and advising campaigns for some of Europe's largest pro-European organisations. He founded the European Campaign Playbook to codify practitioner knowledge across the EU's 27 political environments and make it accessible to campaign professionals, advocacy teams, and civic organisations working at the European level. His work spans electoral strategy, EU public affairs, digital campaigning, and the integration of AI tools into political communication workflows.
Disclosure
Sebastián Rodríguez Pérez is the founder of the European Campaign Playbook and Campaign Intelligence Library. He has worked with pro-European organisations across multiple EU member states. He maintains editorial non-partisanship in all published content; where articles relate to campaigns or organisations he has directly worked with, this is disclosed within the article. He holds no current positions in active electoral campaigns.
What it covers
The difference between narrow and general AI, the layers inside artificial intelligence from machine learning to large language models, how models learn, where their bias comes from, and why prompting matters.
Who it is for
Anyone in a small public affairs, policy or communications team who uses AI tools and wants the vocabulary to follow, and shape, the debate about them.
When to use it
Before any serious conversation about AI risk, AI regulation or AI tools at work, and whenever someone uses “AI” to mean five different things in one sentence.
Key takeaway
Today’s tools are narrow AI, and at their core they are prediction machines: very good at guessing the likely next word, pixel or answer, with no built-in sense of what is true. Knowing that is what lets you use them well and argue about them sensibly.
In brief
- A former White House AI adviser expects extraordinarily capable AI within a couple of years, inside this Commission’s mandate and Trump’s second term.
- The “AI could end humanity by 2030” talk comes mostly from one widely read scenario, AI 2027, whose own authors have since pushed their timelines back.
- Narrow AI is what you use today. General AI, which would learn and reason across domains, does not exist yet.
- AI contains machine learning, which contains deep learning, which contains generative AI and the large language models behind tools like ChatGPT.
- Models learn from huge labelled datasets and then predict. Their training data is overwhelmingly WEIRD, so their answers reflect some humans more than others.
General AI may arrive within Trump’s current presidency
General AI may arrive within Trump’s current presidency. Sounds dramatic, right?
But it’s not just speculation.
In the first series, I already touched upon the exponential growth of these tools and how our brains can’t really understand it.
In a recent New York Times podcast, the former White House chief of AI said that AGI (artificial general intelligence) could emerge within the next 2–3 years.
That’s not decades away.
That’s inside this European Commission’s mandate (2024–2029).
That’s during Donald Trump’s current presidency.
And we’re still trying to regulate basic algorithmic transparency.
Source: The Ezra Klein Show, The New York Times, 4 March 2025, with Ben Buchanan, former White House special adviser for artificial intelligence.
Why everyone is talking about AI ending humanity by 2030
Much of it traces back to one document. In April 2025, Daniel Kokotajlo, a former OpenAI researcher, and a small team published AI 2027, a detailed, month-by-month scenario. In it, AI systems learn to do AI research themselves, progress compounds faster than any government can react, and in the ending the authors considered more likely, a misaligned superintelligence wipes out humanity by 2030.
It spread because it is vivid, and because the people worried about this sit far from the fringe. In 2023, hundreds of researchers and the heads of OpenAI, Google DeepMind and Anthropic signed a one-sentence statement putting the risk of extinction from AI alongside pandemics and nuclear war. Geoffrey Hinton, who shared the 2024 Nobel Prize in physics for foundational work on neural networks, has put the odds of AI wiping out humanity at 10 to 20 per cent within 30 years.
The worry follows a simple chain: if a system can outperform humans at most cognitive work, including the work of improving AI itself, it could become very capable very quickly, and nobody yet knows how to guarantee that such a system pursues the goals we intended. Researchers call that the alignment problem.
Keep it in proportion too. Scenarios are stories, and the AI 2027 authors have since pushed their own timelines back by several years. Plenty of leading researchers think today’s language models are nowhere near general intelligence. And a fixation on extinction can crowd out the harms already here: bias, disinformation and power concentrating in a handful of companies.
So, to understand what all this is about, let’s talk basics
Everything below is the vocabulary. By the end you will know what people actually mean when they say AI, and why one word, predictions, matters more than any other.