Puppugeneraattori

Kari Lilja, TkT, Erikoistutkija; Sirpa Sandelin, TkT, Yliopettaja

English version: Click the link

Kun Katariina Järvinen, psykoterapeutti ja luennoitsija, ja Laura Kolbe, Euroopan historian professori emerita, vuonna 2007 ilmestyneessä yhteisessä kirjassaan Luokkaretkellä hyvinvointiyhteiskunnassa kirjoittivat puppugeneraattorista, he tuskin arvasivat, että vielä heidän elinaikanaan olisi olemassa kone, tai oikeammin sovellus, joka todella toimisi puppugeneraattorina: Tekoäly tai AI, ChatGPT, Copilot, millä nimellä kukin tuon tunteekaan. Heidän terminologiassaan puppugeneraattori oli valikoima tekstejä, jotka aloittelevien tutkijoiden piti professorinsa ohjeen mukaisesti kirjoittaa valmiiksi pöytälaatikkoon niin, että aina tarpeen tullen oli olemassa valmis teksti, johon vain tehtiin tilanteen, aiheen ja uuden tekstin tarkoituksen vaatimat muutokset. Näin saatiin suhteellisen lyhyessä ajassa valmiiksi suuri määrä julkaisuja, ja julkaisujen määrä oli tuolloin – 60- luvun lopulla ja 70-luvun alussa – se, jolla päästiin etenemään akateemisessa hierarkiassa. Sisältöä ja laatua tuskin kukaan tarkasteli. Suomalainen yhteiskunnallinen keskustelu kun ei ollut silloin in kansainvälisissä tutkijapiireissä.

Kolben ja Järvisen tarina on viimeisten kuukausien aikana tullut toistuvasti mieleeni, kun olen yrittänyt käydä järkevää keskustelua pankkieni kanssa. Yksi pankeista vaihtoi tai päivitti verkkopankkisovellustaan, ja muutoksessa ei kaikki mennyt niin kuin piti. Kun ilmoitin heille havaitsemistani virheistä, esimerkiksi siitä, ettei ohjelma hyväksynyt vastaanottajan nimeen heittomerkkiä, vaikka pankin tilitiedoissa kyseinen heittomerkki oli, ja saamistani virheilmoituksista, kun EU:n uusi ohjeistus vaatii pankkia tarkistamaan vastaanottajan nimen, sain vastaukseksi kiitoksen asemesta pitkän kopion tilin käyttöehdoista ja verkkopankin käyttöohjeista ja kehotuksen lukea ne ennen käytön jatkamista.

Sen enempää ohjeissa kuin ehdoissakaan ei sanottu sanaakaan siitä, miten vastaanottajan nimi pitäisi kirjoittaa eikä kielletyistä merkeistä nimessä.

Lampaat vastakkain...
Mihinkähän tämäkin keskustelu johtaa? Image from Pixabay

Kaiken kukkuraksi tuon noin 5 a4-sivua pitkän tekstin keskellä oli huomaamaton kehotus käydä päivittämässä omat tuntemistiedot. Tuollainen pieni huomautus keskellä muuten puisevaa ja tautologista tekstiä jää monelta huomaamatta. Joissakin tapauksissa olen yrittänyt saada vastausta yksinkertaiseen kysymykseen, johon olisi vastaukseksi riittänyt parin lauseen virke. Vastaus on silloinkin ollut monisivuinen litania tiliehtoja, maksupalvelujen ehtoja jne., mutta varsinainen ongelma on jäänyt ratkaisematta. Tämän tyyppisiä vastauksia ei yleensä ole allekirjoittanut henkilö, vaan allekirjoituksessa on vain organisaation nimi.

Berliinin muuri, "takana länsi, edessä itä" (lainaus Yrjö Jylhä, Taipaleen joki, Tuonelan joki)
Joskus tuntuu siltä kuin olisimme aivan eri maailmoissa. Image by Gerd Altmann from Pixabay.

Pahimmillaan asiakas tuntee itsensä tuollaisen kirjeenvaihdon jälkeen aliarvioiduksi, väheksytyksi, jopa loukatuksi. Näin etenkin, jos hän ei tajua, että vastapuolella ei ole enää sitä tuttua hymyilevää pankkineitiä – olettehan huomanneet, että häntä ei ole enää pankin mainoksissakaan – vaan kylmä ja tunteeton generatiivisen tekoälyn algoritmi. Asiakkaalle, joka on käyttänyt pankin sähköisiä palveluita niiden esiversioista alkaen, itse koodannut yrityksen reskontrien ja pankkiliikenneohjelman välisen rajapinnan ja liittymän vuonna 1984, ja tottunut keskustelemaan pankkiasioista ja muista ongelmista ihmisten, ei algoritmien kanssa, tuollainen on suora kehotus vaihtaa pankkia.

Kultaisen noutajan pentu
Mainosten säihkysilmät on korvattu nappisilmillä. Image by Gabru PawPixels from Pixabay.

Vaan mihin vaihdat, kun vastassa on jälleen uusi puppugeneraattori, jonka kanssa ei voi neuvotella koroista, tiliehdoista tai muista asiakkaan kannalta tärkeistä asioista, ja joka ei osaa vastata lyhyesti edes yksinkertaisiin kysymyksiin.

EU-lainsäädäntö velvoittaa työnantajat huolehtimaan siitä, että jokaisen organisaation jokaisella työntekijällä on riittävä tekoälylukutaito, kuten termi suomennettuna kuuluu. Paitsi että termi sinällään on älytön – miten tekoälyä voi lukea, tekoälyn tuotoksia valitettavasti joudumme lukemaan yhä enemmän – itse lainsäädännön soveltaminen lähestyy asiaa väärästä päästä. EU:n tekoälydirektiivi (EU AI Act), joka astui voimaan 1. elokuuta 2024, toi mukanaan lakisääteisen velvollisuuden kouluttaa henkilöstöä tekoälyn käyttöön ja sen riskeihin. Tätä on tulkittu niin, että koulun penkille pannaan suorittava työntekijäporras työnjohtajineen, ja heille opetetaan ChatGPT:n ja Copilotin peruskäyttöä. Opettajina näillä kursseilla on usein intohimoisia AI-Guruja, joiden mielet aktiivisesti sulkevat pois kaikki mahdolliset negatiiviset seuraukset, ja joilla ei ole edes perusymmärrystä käytettävyydestä ja tuotetun tekstin luettavuudesta ja ymmärrettävyydestä tai organisaation liiketoiminnasta ja asiakaspalvelutilanteista.

Tekoälyasetuksen tavoitteena on varmistaa, että tekoälyjärjestelmät eivät vaaranna ihmisten terveyttä, turvallisuutta tai perusoikeuksia. Koulutusvaatimus kohdistuu käyttöön ja riskeihin. Riskit, joita EU-tason lainsäätäjä on tekoälyn käyttöönotossa nähnyt, ovat siten organisaation päätöksentekohierarkiassa aivan eri tasolla kuin siivooja, joka joutuu miettimään, käyttääkö hän tänään moppia vai lastaa. Tekoälyn käytön riskien realisoituminen on ensi sijassa kiinni organisaation strategiasta ja johdon tekemistä päätöksistä. Riskien välttäminen ja torjuminen vaatii toki johdolta koulutusta ja ymmärrystä algoritmien luonteesta ja toiminnasta, mutta myös vahvaa eettistä ymmärrystä ja moraalista sitoutumista.

Ainakaan tähän mennessä en ole kuullut, että yksikään taho olisi tarjonnut tämän kaltaista tekoälykoulutusta organisaatioille.

Mäyräkoira remmin päässä.
Lauman johtajakin joutuu joskus koulun penkille… Image by Sabrina St. from Pixabay.

English version

The rubbish generator

When Katariina Järvinen, a psychotherapist and lecturer, and Laura Kolbe, professor emerita of European history, wrote about the rubbish generator in their joint book, “Luokkaretkellä hyvinvointiyhteiskunnassa” – “Overriding the class borders in welfare society”, published in 2007, they could hardly have imagined that in their lifetimes there would be a machine, or rather an application, that would actually function as a rubbish generator: Artificial Intelligence or AI, ChatGPT, Copilot, by whatever name you know it. In their terminology, the rubbish generator was a selection of texts that novice researchers had to write in a drawer according to their professor’s instructions so that whenever necessary, there was a ready-made text, to which only the changes required by the situation, topic and purpose of the new text were made. In this way, a large number of publications were completed in a relatively short time, and it was at that time – in the late 60s and early 70s – the number of publications that allowed one to advance in the academic hierarchy. Hardly anyone checked the content and quality. Finnish social debate was not in in international academic circles.

The story of Kolbe and Järvinen has repeatedly come to mind in recent months when I have tried to have a reasonable discussion with my banks. One of the banks changed or updated its online banking application, and not everything went as expected during the change. When I informed them of the errors I had noticed, for example, that the program did not accept an apostrophe in the recipient’s name, even though the bank’s account information contained the apostrophe, and of the error messages I received when the new EU guidelines require the bank to check the recipient’s name, I received in response, instead of thanks, a long copy of the account terms and conditions and the online bank’s user manual, and an invitation to read them before continuing to use them.

Neither the instructions nor the terms and conditions said a word about how the recipient’s name should be written or about prohibited characters in the name.

Two sheeps against each other
What is this discussion leading to? Image from Pixabay

To top it all off, in the middle of that approximately 5 A4-page long text was an inconspicuous invitation to update your contact information. Such a small note in the middle of an otherwise dull and tautological text remains unnoticed by many. In some cases, I have tried to get an answer to a simple question that would have been answered in a sentence or two. Even then, the answer has been a multi-page litany of account terms, payment service terms, etc., but the actual problem has remained unresolved. These types of answers are usually not signed by the person, but only by the name of the organization.

The great Berlin Wall between west and east...
Sometimes it feels like we’re in completely different worlds. Image by Gerd Altmann from Pixabay.

At worst, the customer feels underestimated, belittled, even insulted after such correspondence. This is especially true if they do not realize that the other party is no longer the familiar smiling bank lady – you may have noticed that she exists no longer in the bank’s advertisements – but a cold and emotionless generative artificial intelligence algorithm. For a customer who has used the bank’s electronic services since their pre-release versions, who coded the interface and connection between the company’s systems and the banking program himself in 1984, and who is used to discussing banking matters and other problems with people, not algorithms, this is a direct invitation to change the bank.

Golden retriever puppy watching at you.
The glitter eyes in the ads have been replaced with button eyes. Image by Gabru PawPixels from Pixabay.

But which bank could you change to if you are everywhere faced with yet another rubbish generator with whom you cannot negotiate interest rates, account terms, or other matters important to the customer, and who cannot answer even simple questions briefly?

EU legislation requires employers to ensure that every employee in every organization has sufficient AI literacy. The term has been translated into Finnish “tekoälylukutaito” – the skill to read AI. Not only is the term itself unintuitive – how can you read AI, and, unfortunately, we are increasingly forced to read products of AI – but the application of the legislation itself approaches the issue from the wrong end. The AI Directive of the EU, which entered into force on 1 August 2024, brought with it a statutory obligation to train personnel in the use of AI and its risks. This has been interpreted as putting grassroot employee and their supervisors in school and teaching them the basic use of ChatGPT and Copilot. These courses are often taught by passionate AI Gurus whose minds actively exclude all possible negative consequences, and who do not even have a basic understanding of usability, readability and comprehensibility of the produced text, or of the organization’s business and customer service situations.

The aim of the AI Regulation is to ensure that AI systems do not endanger human health, safety or fundamental rights. The training requirement focuses on use and risks. The risks that the EU level legislators have seen in the introduction of AI are therefore at a completely different level in the decision-making hierarchy of the organization than a cleaner whose only decision to be made is whether to use a mop or a bucket today. The realization of the risks of using AI depends primarily on the organization’s strategy and the decisions made by management. Avoiding and combating risks certainly requires management to have training and understanding of the nature and functioning of algorithms, but also a strong ethical understanding and moral commitment.

At least so far, I have not heard of any entity offering this kind of AI training to organizations.

Dachshund having a look forward
Even the leader of the pack sometimes has to go to school… Image by Sabrina St. from Pixabay.

Tämä blogikirjoitus on kirjoitettu Euroopan unionin osarahoittaman Win4SMEs-hankkeen puitteissa. Näkemykset ja mielipiteet ovat kuitenkin ainoastaan ​​kirjoittajan/kirjoittajien omia eivätkä välttämättä vastaa Euroopan unionin tai Euroopan koulutuksen ja kulttuurin toimeenpanoviraston (EACEA) näkemyksiä. Euroopan unionia tai EACEA:ta ei voida pitää niistä vastuussa.

This blog text has been written within the framework of project WIN4SMEs, co-funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Education and Culture Executive Agency (EACEA). Neither the European Union nor EACEA can be held responsible for them.