Robots are not all-powerful: where AI still loses to humans — expert analysis
April 2026 marked a historic event: the Sony Ace robot, equipped with advanced artificial intelligence, defeated professional table tennis player Mia Kihara. The match was played under the official rules of the International Table Tennis Federation (ITTF), and Sony engineers have already called this moment a landmark milestone. For the first time, a machine achieved expert-level performance in a real competitive sport.
This event is just the tip of the iceberg. History knows many cases where AI surpassed humans in strictly defined disciplines. Let's look at the key milestones:
- 1997: Deep Blue defeats Garry Kasparov in chess — the first victory over a world champion in a classical match.
- 2011: IBM Watson crushes the best players on Jeopardy!, demonstrating the ability to process complex speech structures.
- 2016: AlphaGo beats Lee Sedol at the game of Go — overcoming the barrier of a game with an astronomical number of moves.
- 2017: Libratus wins over $1.7 million in poker — AI learned to bluff under conditions of incomplete information.
- 2019: OpenAI Five defeats world champions in Dota 2 — the machine surpassed humans in a team esports discipline.
However, despite these impressive achievements, it is still premature to talk about total machine dominance. May 2026 clearly demonstrated that in some areas, humans still hold the lead.
Physical labor: humans pull ahead
A live broadcast of a 10-hour showdown between the humanoid robot F.03 from Figure AI and an ordinary intern named Aime showed an unexpected result. The task was simple: scan a barcode, lift a box, and place it label-side down on a conveyor belt.
Aime processed 12,924 packages, spending an average of 2.79 seconds per item. The F.03 robot handled 12,732 packages, with an average time of 2.83 seconds. The difference is minimal, but the nuances matter: the human had legitimate breaks for rest and lunch, while the AI only pulled ahead in the fifth hour when the intern stepped away. By the end, Aime had developed calluses and his hand was very tired. The robot can work non-stop, but over a short distance, the human proved more efficient.
Economic reality: humans are cheaper than machines
There is also a compelling economic argument. Employers widely acknowledge that hiring people is often more profitable than maintaining AI. Corporate spending on technology is growing too fast. Microsoft is already limiting internal licenses for Claude Code for staff due to high token costs, and Uber exhausted its entire AI budget for 2026 in four months. Per-minute billing for computing power often eats up all the benefits from workforce optimization.
My conclusion as an analyst: Machines win where algorithms are clearly measurable — in games, sports with fixed rules, and data analysis. But in areas requiring adaptability to physical conditions, fine motor skills, and, critically, cost-effectiveness, humans remain irreplaceable for now. The AI boom does not mean the end of human labor — rather, it redefines its value. And in this new reality, human flexibility and endurance turn out to be their main trump card.