Ingen AI / no AI.  

English below.

Nederst på min side står der, at ingen tekster eller billeder er AI-generede. Og nej, det er ikke en stavefejl. 

For hvordan kan man overhovedet bruge AI?

1. Datagrundlaget for AI – dvs. selve råmaterialet – er hentet direkte fra alt, der nogensinde har været tilgængeligt på nettet, uden hensyntagen til ophavsret. Når forskere, kunstnere, aviser m.fl. har rejst sager mod AI-selskaber har man konkret sagt, at AI-teknologien ikke er mulig, hvis man skal skele til ophavsret. Men at AI-teknologi skal eksistere er ikke en beslutning, truffet i diverse, mulige fora med folkevalgte, det er et rent kommercielt produkt, som altså bygger på stjålne data. Hvordan kan man bruge et produkt, hvor råmaterialet er tyvstjålet?
2. Datagrundlaget manipuleres, dvs. chatbotterne “trænes” til hvilke svar på givne prompts der er ønskede, og hvilke der ikke er. Nogle firmaer har f.eks. fremhævet, at man ikke ønsker, at man skal kunne bruge deres AI til at generere f.eks. børnepornografi. Men der er ikke et “brugerråd”, der tager stilling til “ønsket” og “uønsket” – det er alene tech-giganterne, der bestemmer. I Metas træningsprogram kan man f.eks. se, at efter Trumps indsættelse blev al henvisning til LGBTQ mærket “kontroversielt”. Vi har brugt årevis på at lære, i hvor høj grad sproget påvirker vores tænkning, og dermed vores kultur. Vi har luget ud i de ord, vi bruger, fordi vi ved, at ord påvirker. Men man kan jo lige rette en AI-tekst til, før man bruger den, vil nogen tænke – og det tror jeg ikke nytter meget. For hvad retter man? Konkrete fejl – hvis man ser dem, og er klar over, at det er fejl. Men man lægger måske lige netop ikke mærke til bias – og man lægger ikke mærke til det, der ikke er nævnt, det som er udeladt. Skal årtiers bevidst arbejde med sproget overlades til amerikansk tech? Hvordan kan vi risikere, at vores tekster, billeder, de informationer, vi søger, bliver manipulerede i bestemte retninger, uden at vi måske bemærker det?
3. Vi har set en stigende opmærksomhed på, at det er uheldigt, at så meget af vores teknologiske infrastruktur er amerikansk – for det giver USA meget potente våben i hænde, hvis vi i en eller anden situation ikke makker ret. Vi har f.eks. set Microsoft lukke e-mail-konti for folk, som Trump ser som “bad, bad people”. Hvordan kan vi så gøre os selv og endnu værre, offentlige servicetilbud, endnu mere afhængige af amerikansk tech?
4. Ressourceforbruget til AI er uforholdsmæssigt stort – fortrinsvis vand og strøm. En stor del af verden er i forskellig takt gået i gang med at aftale og arbejde mod klimamål, undtagen jo USA, som ikke vil være med, og som endnu værre oven i købet har valgt en præsidumt, som sætter alt ind på at øge forbruget af fossile brændstoffer og spænder ben for grøn politik i USA. Hvem har vurderet, at AI er til så stor nytte, at vi skal acceptere et energiforbrug, som techgiganterne ikke selv mener nogensinde kan løses ved grøn energi? Det har tech-giganterne. Man kan ikke bruge AI uden samtidig at latterliggøre sine daglige bestræbelser på at nedsætte CO2-udslip. Sparepærer er ligegyldige i den sammenhæng – undtagen den ene, som tech-giganterne købte mod at få lovning på, at der ikke indføres nogen restriktioner for AI. 
5. Jeg kan ikke forestille mig andet, end at brug af AI vil føre til væsentligt videns-og kompetencetab. Kan folk om 10 år overhovedet finde ud af at søge informationer, som ikke er blendede og siede af AI? Kan folk overhovedet selv skrive en tekst? Hvad sker der med vores bevidste arbejde med at luge bias ud af vores sprog og kultur? Og hvad sker der med vores opfattelse af rigtigt og forkert? En AI-genereret biografi om mig fortæller blandt andet, at jeg har været folketingsmedlem i mange år, hvilket er helt ude i hampen. Men det ved man jo kun, hvis man kender mig.
6. Overvågning – kan vi acceptere, at vores samtaler, søgninger, etc. er overvågede af udenlandske firmaer? Vi ved, at AI-udbyderne overvåger samtaler.I den sag om en canadisk dreng, hvor Sam Altman undskyldte over for den by, hvor skyderiet fandt sted, sagde han, at han beklagede, at man ikke havde reageret, selv om man var klar over at hans “samtaler” med AI handlede om skydning. “OpenAI CEO Sam Altman issued an apology letter to the people of Tumbler Ridge. The apology is necessary, and yet grossly insufficient for the devastation done to the families of Tumbler Ridge. OpenAI faced scrutiny after it admitted that the account of the 18-year-old shooter wasn’t reported to police even after staff at the company noted the link to gun violence.”.  Så der finder altså en overvågning sted – igen noget, man som bruger ingen indflydelse har på, og hvor overvågningens formål ikke er transparent.

Jeg tror, at en meget stor hurdle at komme over i AI-diskussioner er, at det samfund, vi har haft i Danmark – og f.eks. resten af Skandinavien – har gjort os “godtroende”. AI er den historisk største mulighed, vi nogen sinde har set, for infiltrering, manipulering og kontrol. 

Men den slags tror vi ikke kan ske i virkeligheden. Vi har nemlig været beskyttede af alle mulige ting, som f.eks. persondataloven, Etisk råd (hvad man så mener om det, men institutionen findes), offentlig aktindsigt, forbrugerombudsmand,  vi stoler på at vores valgsystem ikke er manipuleret, at staten ikke misbruger vores oplysninger, osv., osv.  

“Hjernevask” er sådan noget, de gør i Rusland og Kina. Vi er frie, vi er ikke udsat for mind control, vi forventer pålidelighed, transparens osv.  

Vi er simpelthen for godtroende til at tro, at nogen kan finde på at manipulere os. 

No AI.

How can we possibly even use AI?

1. The data foundation for AI – the raw material itself – is stolen directly from everything that has ever been available online, with no regard for copyright. When researchers, artists, newspapers, and others have brought legal action against AI companies, the companies have explicitly stated that AI technology would be impossible if copyright had to be taken into account. Yet, the existence of AI technology is not a decision reached through any democratic or public forum, it is a purely commercial product built upon stolen data. How can one use a product whose raw material has been stolen?
2. The data foundation is manipulated. The chatbots are “trained” to determine which responses to specific prompts are desirable and which are not. Some companies, for instance, have emphasized that they do not want their AI to be used to generate content such as child pornography. However, there is no “user council” to decide what is “desirable” or “undesirable”, this is determined solely by the tech giants. In Meta’s training program, for example, it says that following Trump’s inauguration, any reference to LGBTQ issues was labeled “controversial.” We have spent years learning the extent to which language influences our thinking, and, by extension, our culture. We have carefully curated the words we use, precisely because we know that words have an impact. One might think, “Well, you can just edit an AI-generated text before using it” – but I do not believe that helps much. Because what exactly are you editing? Concrete errors, maybe, if you spots them and recognize them as errors. But you might very well fail to notice bias – or fail to notice what has not been mentioned, what has been left out. Should decades of deliberate work on language be handed over to American tech? How can we risk having our texts, images, and the information we seek manipulated in specific directions – potentially without us even noticing?
3. There is an increasing awareness that it is problematic for so much of our technological infrastructure to be American; it places very potent weapons in the hands of the US, should we, in any given situation, fail to fall in line. For instance, we have seen Microsoft shut down the email accounts of individuals whom Trump deems “bad, bad people.” How, then, can we make ourselves – and, worse yet, public services – even more dependent on American tech?
4. The resource consumption required for AI is disproportionately high – primarily in terms of water and electricity. Much of the world has begun, at varying paces, to agree upon and work toward climate goals, with the notable exception of the US, which refuses to participate. Worse still, the US has elected a president who is fully committed to increasing the consumption of fossil fuels and actively obstructing green policy within the country. Who decided that AI is of such great value that we should accept this enormous energy demand – one that the tech giants themselves admit can never be met through green energy alone? The tech giants themselves decided that. You simply cannot use AI without simultaneously making a mockery of your  daily efforts to reduce CO2 emissions. Energy-saving light bulbs are almost ridiculous in this context – except for the one that the tech giants bought in exchange for a promise that no restrictions would be imposed on AI.
5. I cannot imagine anything other than that the use of AI will lead to a significant loss of knowledge and skills. In 10 years, will people even be able to figure out how to search for information that hasn’t been blended and filtered by AI? Will people even be able to write a text themselves? What happens to our conscious efforts to weed bias out of our language and culture? And what happens to our perception of right and wrong? An AI-generated biography about me claims, among other things, that I have been a member of the Danish Parliament for many years – which is completely absurd. But, of course, you would only know that if you knew me personally.
6. Surveillance – are we accepting that our conversations, searches, etc., are monitored by foreign companies? We know that AI providers monitor conversations. In the case involving a Canadian boy – where Sam Altman apologized to the town where the shooting took place – he expressed regret that no action was taken, even though it was known that the boy’s “conversations” with the AI ​​concerned a shooting. “OpenAI CEO Sam Altman issued an apology letter to the people of Tumbler Ridge. The apology is necessary, and yet grossly insufficient for the devastation done to the families of Tumbler Ridge. OpenAI faced scrutiny after it admitted that the account of the 18-year-old shooter wasn’t reported to police even after staff at the company noted the link to gun violence.” So, surveillance is indeed taking place – yet another aspect over which users have no influence, and where the purpose of the surveillance is not transparent.

I believe a major hurdle in discussions about AI is that the society we have had in Denmark – and the rest of Scandinavia – has made us gullible. AI represents the greatest opportunity in history – the likes of which we have never seen – for infiltration, manipulation, and control. But we do not believe such things can actually happen. After all, we have been protected by all sorts of safeguards – such as data protection laws, the Danish Council on Ethics (regardless of one’s opinion of it, the institution exists), the right to access public records, and the Consumer Ombudsman. We trust that our electoral system is not manipulated, that the state does not misuse our personal information, and so on.

Brainwashing is something that happens in places like Russia or China. We are free; we are not subjected to mind control; we expect reliability, transparency, and the like.

We are simply too gullible to believe that anyone would ever attempt to manipulate us.