Breaking
EU Commission issues new nitrogen compliance ultimatumFrisian farmers vow to resist Brussels directiveNew fierljeppen record set in WinsumWetterskip Fryslân warns of coastal flooding riskLeeuwarden named top cycling city in the NetherlandsEU Commission issues new nitrogen compliance ultimatumFrisian farmers vow to resist Brussels directiveNew fierljeppen record set in WinsumWetterskip Fryslân warns of coastal flooding riskLeeuwarden named top cycling city in the Netherlands
Tuesday, 20 May 2026  ·  Ljouwert, FryslânEst. 2026

FRISIAN NEWS

Nijs fan de Wrâld  ·  World News  ·  Frisian Perspective

What Artificial Intelligence Cannot Do, No Matter What CEOs Say
World

Wat keunstmjittige yntelliginsje net dwaan kin, wat direkteuren ek sizze

December 23, 2025 · Frisian News

Tech executives promise AI will solve medicine, poverty, and climate change. The machines cannot think, understand causation, or take responsibility for their mistakes.

Frisian flagFrysk

Ferline wike sei de direkteur fan in grut AI-bedriuw tsjin ynvestearders dat syn modellen ynkoarten kanker genêze soene, earmoed beëinigje soene en de opwaarming fan de ierde keare soene. Hy joech gjin tiidline, gjin budzjet, gjin proefprojekt. Allinnich in belofte dat keunstmjittige yntelliginsje oplosse soe wat minsken tsientallen jierren net oplost hawwe. Dit is gjin foarsizzing. Dit is ferkeappraat fermomd as fyzje.

AI-systemen dogge ien ding goed: se kopiearje patroanen út trainingsgegevens en produsearje tekst dy't oannimlik klinkt. Se kinne net redenearje út earste begjinsels. Se kinne in hypoteze net tsjin de wurklikheid toetse. As jo in AI-model freegje wêrom't wat bart, riedt it. As it antwurd ferkeard is, hat it gjin wize om te witten dat it in flater makke. In dokter kin in flater tajaan. In model produsearret gewoan it folgjende wierskynlike wurd.

De masines misse oarsaak en gefolch. Se sjogge korrelasje en neame it wet. Jou in AI genôch histoaryske waarsgegevens en it sil mei fertrouwen rein foar kommende tiisdei foarsizze, sels as syn logika ûnsin is. It begrypt atmosfearike druk, wynferskowing of wêrom't dy dingen fan belang binne net. It fûn in patroan en ekstrapolearre. Dit wurket foar guon taken. Foar genêskunde, engineering en belied is it gefaarlik.

Dêr't AI echte winst opsmyt, dogge minsken noch altyd it swiere wurk. As modellen helpe sykte te diagnostisearjen, beoardielet in radioloog de útfier en nimt it beslút. As algoritmen belied suggerearje, tinke keazen lieders (soms) nei oer de gefolgen. De masine helpt. Minsken beslisse. Helje de minsken derút en jo hawwe in systeem dat tûk klinkt mar tinkt as in hiel rappe idioat.

Direkteurs ferkeapje AI as in middel dat alles genêze kin om't ynvestearders it leauwe wolle en om't it tajaan fan de grinzen fan harren technology harren jild kostet. Se sille folgjend jier deselde beloften dwaan. De masines sille kanker, earmoed of klimaatferoaring net genêze. Mar dat sizze kostet gjin oandielen.

English

Last week, the CEO of a major AI firm told investors his company's models would soon cure cancer, end poverty, and reverse global warming. He offered no timeline, no budget, no pilot program. Just a promise that artificial intelligence would solve what humans have failed to solve for decades. This is not prediction. This is sales pitch dressed as vision.

AI systems do one thing well: they copy patterns from training data and output text that sounds plausible. They cannot reason from first principles. They cannot test a hypothesis against reality. When you ask an AI model why something happens, it guesses. When it gets the answer wrong, it has no way to know it made a mistake. A doctor can admit error. A model simply outputs the next likely word.

The machines lack cause and effect. They see correlation and call it law. Feed an AI enough historical weather data and it will predict rain next Tuesday with confidence, even when its logic is nonsense. It has no understanding of atmospheric pressure, wind shear, or why those things matter. It found a pattern and extrapolated. This works for some tasks. For medicine, engineering, and policy, it is dangerous.

Where AI delivers real gains, humans still do the hard work. When models help diagnose disease, a radiologist reviews the output and makes the call. When algorithms suggest policy, elected officials (sometimes) think through the consequences. The machine assists. Humans decide. Strip away the humans, and you have a system that sounds smart but thinks like a very fast idiot.

CEOs sell AI as the cure-all because investors want to believe it, and because admitting the limits of their technology costs them money. They will keep making promises next year too. The machines will not cure cancer by then, or poverty, or climate change. But saying so does not sell stock.


Published December 23, 2025 · Frisian News · Ljouwert, Fryslân