AI leaps six months in four years: hype or reality?

If a well-known AI communicator’s recent podcast claim is true, the last six months have rendered everything you thought you knew about AI obsolete overnight. The speaker, a photographer turned tech evangelist, argues that AI has pogre more in this half-year than in the previous four years. This statement carries weight: we are either facing a historic inflection point or witnessing the decade's biggest marketing exercise.

Image generation: where change is visible and tangible

Image generation is the most visible example. Two years ago, a paid Midjourney account produced curious results but often deviated from requests. Now, free tools like Gemini’s image editor, nicknamed “nano banana,” interpret complex instructions with unsettling precision. Users can iterate on corrections to achieve nearly perfect results. However, not all is gold: consistency between scenes remains poor, video stalls with more than four people, and models apply only one in seven instructions when numerous. Yet, the improvement compared to two years ago is objective.

The hype business: courses, subscriptions, and Romuald Fons’ shadow

Economically, another side emerges. The communicator is not just an evangelist: he sells AI courses for around €2,000, has advertising deals with sector companies, and his independence is, at best, questionable. Some recall that his relationship with Romuald Fons, partner of a known training agency, coincided with a shift toward direct-response TV-style sales. The suspicion is clear: are you watching an analysis or a sales funnel? The same logic applies to tools: paid models are capped for the general public, and when one starts working well, it is placed behind a subscription wall. The gap between what ordinary users see and what certain institutions use, reportedly, is abyssal.

AGI, employment, and the looming bubble

Behind the noise lies the crucial question: when will AGI arrive and what will it do to jobs? Estimates range from optimists predicting 2 to 4 years, to skeptics saying 40 or never. Skepticism points out that current models are statistical parrots—fast but lacking real comprehension—and that massive investment suspiciously resembles the subprime crisis. Others highlight that capability improvements are visible every three months and that the bottleneck is computing power, not ideas. Meanwhile, layoffs in cognitively intensive sectors accelerate, and everyone talks about universal basic income as if machines were going to row for us.

With this landscape, the communicator’s claim gains traction among daily technology users, but distrust regarding his commercial interests persists. Are we facing the singularity or the decade's biggest advertorial? Machines still don’t row. But if they ever do, we must watch who holds the oars.