(English below)
Vi lever i en tid, hvor vi i flere og flere situationer skal acceptere, at noget ikke er optimalt, men at det er sådan nogenlunde acceptabelt.
Vi ser billeder af pingviner i Grønland, og så griner vi lidt af den, der har promptet AI-generatoren. Men havde man nu bedt en fotograf om et billede med pingviner i Grønland, så havde han eller hun nok svaret, at det ikke kan lade sig gøre, med mindre man selv tager en flok pingviner med. Nu er det jo et eksempel, der først og fremmest latterliggør Trump, og viser, hvor lidt han ved – men i mange andre sammenhænge ser vi billeder, der har fejl – men hvor man har vurderet at det er “godt nok”. En pige på en slæde, der “trækkes” af 6 slædehunde – men de løb alle sammen bag slæden. Eller et billede på forsiden af NYtimes forleden, som var en test – kan du se, om billedet er AI-genereret? Det var lavet med den sidste nye, formidable billedgenerator. Og ja, så lægger man f.eks. mærke til, at den mand, der på billedet står med en kasse i hænderne, tydeligt har hvide hænder, men på det ID-kort, han har hængende om halsen, er der et billede af en sort mand. Osv.. men ja, der er fejl, men det kan vel være godt nok til nogle formål. Men hvorfor skal vi præsenteres for noget, der kun lige akkurat er godt nok?
For nylig skulle jeg bestille fragt af et lille bord, og gik ind på et fragtfirmas hjemmeside, for at udfylde de oplysninger, der skal til for at en pris kan beregnes. Alle felter blev forudfyldt af AI, og der stod så nedenunder, at man kunne redigere i AI-teksten. Jeg havde brug for at skrive 3 mål – højde, bredde og længde – og så vægten. Det ville have taget mig 30 sekunder, og det ville have taget en fragtmand ca. den samme tid at læse. Men der kom en halv side frem med AI-genereret sludder, og jeg kunne ikke redigere. Så jeg afsluttede, og sendte besked til firmaet, at jeg ville finde et andet fragtfirma, der ikke brugte AI. Næste dag blev jeg ringet op af en leder derfra, som godt kunne se problemet, og godt kunne se, at den måde, de havde sat deres system op på, i virkeligheden ikke tillod redigeringer – så snart man skrev sine ændringer, blev de slettede, og erstattet med en ny AI-tekst. Jeg spurgte ham så, hvorfor de overhovedet brugte det makværk, og fik svaret, at ja, men det er jo en ny teknologi, og ja, der er fejl, men mange gange er det jo godt nok…
Igen – hvorfor skal vi pludselig skrue ambitionerne ned til “godt nok”?
Hvis folk søger oplysninger ved at prompte en AI-bot, så ved man godt, at svarene kan være fejlbehæftede, og så er det jo ubrugeligt, netop fordi folk så ikke kan vide, om lige præcis det svar, de så får, er fejlbehæftet. Det er altså ikke engang “godt nok” – det er ubrugeligt. I forvejen har vi allerede mistet evne til at finde ordentlige informationer, for i mange tilfælde stiller vi os bare ud på Google og spørger om et eller andet – f.eks. “kan man få skoldkopper igen, hvis man har haft det som barn?” Eller “hvad koster det at få fornyet sit pas?” Og så får folk en række links, de kan klikke på, og de øverste – og de fleste – er oftest AI-genererede. I gamle dage vidste vi godt, at hvis man ville vide noget om sygdom, så gik man direkte ind på “sundhed.dk”, eller “ssi.dk”, og hvis man ville vide noget om f.eks. pas, så gik man ind på “borger.dk”. Men nu står folk og råber ud i mennesketomheden på Google, og så bruger de de svar, der har fået af en AI-bot. Og svarene er somme tider brugelige, andre gange helt ubrugelige, men det må så være godt nok.
De automatoversættelser, vi ser på brugsanvisninger og den slags – eller denne side – er ofte elendige. Og vi vænner os langsomt til, at sådan er det – at man skal gætte sig til en del. I ét af mine opslag lagde jeg mærke til, at “Turbonissen” på de få linjer tekst, der er, er oversat til både “Turbo Santa”, “Turbo the Gnome”, “Turbo Elf” og endelig simpelthen “Turbonissen”. Men det er vel godt nok…
Er det fordi, alting så bliver billigere, at vi skal lære at nøjes i flere og flere situationer? Det tror jeg, de færreste forbrugere oplever, men Tech-giganterne forventer stadig, at de kan skabe arbejdsløshed i helt uset omfang, fordi AI vil kunne erstatte mennesker, så hvis det skulle være tilfældet, så skal rigtig, rigtig mange mennesker på overførselsindkomst, og deres skatteindbetalinger til fælleskassen bliver samtidigt betragteligt mindre. Desuden kender vi ikke regningen, når tech-giganterne ikke længere selv betaler for det enorme energiforbrug – når diverse virksomheder har gjort sig selv afhængige af AI. Og her skal vi måske bare acceptere, at når det ene datacenter efter det andet skyder i vejret, så må vi skrotte vores ambition om at dække hele vores elforbrug med sol, vind og vand – og så må vi nøjes med en lavere ambition. Tech-giganterne siger selv, at de ikke mener at det er realistisk, at deres elforbrug nogen sinde kan dækkes af vedvarende kilder, og den præsident, de selv har købt, skruer også bevidst op for kul- og olieforbrug. Det sorte kul er pludselig smukt og rent, og olien er flydende guld lige under vore fødder. Det viser sig så nu, at vi er så langt fra Paris-aftalens mål om en maksimal temperaturstigning på 1,5 grad, at vi mere eller mindre må nøjes med at tænke, at det må gå, som det går, og så må det være godt nok.
Vi diskuterer også, hvad vores børn egentlig skal lære, nu hvor de bare kan spørge AI om det ene og det andet – og at man overhovedet kan diskutere det, viser at vi allerede har svært ved at skelne mellem viden og løse stumper af information, som måske er korrekt, og måske ikke er. Og jo længere, vi kommer ud af dette spor, jo dummere bliver vi, og jo sværere får vi ved selv at se manipulation og bias.
Heldigvis er der en stigende tendens, måske særligt i yngre generationer, at det at bruge AI er kikset og pinligt. Jeg håber virkelig, at den tendens breder sig endnu mere, for det liv, vi er ved at skabe for de næste generationer, det er bare ikke godt nok.
Good enough ..
We live in an era where, in an increasing number of situations, we have to accept that things aren’t optimal—they are merely “good enough.”
We see images of penguins in Greenland, and we might chuckle at the person who prompted the AI generator. Yet, if you had asked a photographer for a picture of penguins in Greenland, he or she would likely have told you it’s impossible—unless you brought a flock of penguins along yourself. While this specific example primarily serves to mock Trump and highlight his lack of knowledge, we frequently encounter images in other contexts that contain errors – yet are considered just “good enough.” Like a girl on a sled being “pulled” by six sled dogs – only for the dogs were running *behind* the sled. Or take an image that appeared recently in *The New York Times* as a test: can you tell if it’s AI-generated? It was created using the latest, highly advanced image generator. You might notice, for instance, that the man holding a box clearly has white hands, yet the ID card hanging around his neck features a photo of a black man. And so on. Yes, there are flaws, but they considered just acceptable enough for purposes like advertising. But why should we be presented with something that is only *just* good enough?
Recently, I needed to arrange shipping for a small table, so I visited a shipping company’s website to enter the details required to calculate a price. All the fields were pre-filled by AI, and a note below indicated that the AI-generated text could be edited.
I simply needed to enter three dimensions – height, width, and length – and the weight. That would have taken me 30 seconds, and it would have taken a carrier about the same time to read. Instead, half a page of AI-generated nonsense appeared, and I couldn’t edit it. So, I quit the process and sent a message to the company, saying I would find another carrier – one that didn’t use AI. The next day, I received a call from a manager, who acknowledged the problem; he realized that the way their system was set up actually prevented any editing – as soon as you typed in your changes, they were deleted and replaced by new AI-generated text. I asked him why they were using that crap in the first place. His response was, “Well, it’s a new technology – sure, there are errors, but often it’s ‘good enough’…”
Once again—why should we suddenly lower our standards to “good enough”?
If people seek information by prompting an AI bot, they know – at least, I hope they do – that the answers might be flawed… which makes the answers useless – because people have no way of knowing whether the specific answer they receive is flawed. It’s not even a matter of being “good enough”—it’s simply useless. We have already lost the ability to find reliable information ourselves – in many cases, we just throw a question out into the void of Google – for instance, “Can you get chickenpox again if you had it as a child?” or “How much does it cost to renew a passport?” People are then presented with a list of links, and the top results – in fact, the majority of them – are usually AI-generated. In the past, we knew that if we wanted information about a medical condition, we would go directly to “sundhed.dk” or “ssi.dk,” and if we needed information about something like a passport, we would visit “borger.dk.” But now, people just shout their questions out on Google and rely on the answers provided by an AI bot. Sometimes these answers are useful, other times they are completely useless – but apparently, that’s considered “good enough.”
The machine translations we see in user manuals and similar materials – or even on this very site – are often abysmal. And we are slowly becoming accustomed to this reality: that you often have to guess at the meaning. In one of my posts, I noticed that the term “Turbonissen” was translated—within just a few lines of text—as “Turbo Santa,” “Turbo the Gnome,” “Turbo Elf,” and finally, simply as “Turbonissen.” And poor Laundry Nisse – he was named a scumbag..
But I suppose that’s good enough…
Is it because everything will become cheaper that we are expected to learn to settle for less in more and more situations? I suspect very few consumers perceive it that way; yet, the tech giants anticipate creating unemployment on an unprecedented scale, as AI will be capable of replacing human workers. If that were to happen, a vast number of people would end up relying on social welfare payments, while their tax contributions to the public purse would shrink significantly. Furthermore, we do not yet know the true cost – the bill that will come once the tech giants no longer shoulder the expense of their own massive energy consumption, and once various businesses have become entirely dependent on AI. In this scenario, we may simply have to accept that as one data center after another springs up, we must abandon our ambition to meet our entire electricity demand through solar, wind, and hydropower – settling, instead, for a more modest goal. The tech giants themselves state that they do not believe it is realistic for their electricity consumption to ever be met by renewable sources; meanwhile, their own president bought fair and square with their own money, is deliberately ramping up the use of coal and oil. Suddenly, black coal is portrayed as beautiful and clean, while oil is described as liquid gold lying just beneath our feet. It now appears that we are so far from the Paris Agreement’s target – limiting the global temperature rise to a maximum of 1.5 degrees – that we are left with little choice but to accept that things will unfold as they will, and simply deem that “good enough.”
We are also discussing what our children should actually learn, now that they can simply ask AI about anything and everything – and the very fact that this is up for debate shows that we are already struggling to distinguish between genuine knowledge and scattered bits of information, which may or may not be accurate. The further we go down this path, the dumber we will be, and the harder it gets for us to spot manipulation and bias on our own.
Fortunately, there is a growing trend – perhaps mostly among younger generations – where using AI is seen as uncool and embarrassing. I truly hope this trend spreads even further, because the life we are currently creating for future generations is simply not good enough.
Skriv et svar