This text belongs to 'Artificial', the AI newsletter that Delia Rodríguez sends every Friday.
Dear readers, dear readers: Sometimes we underestimate the effect of a technology thinking that “that” could also be done before. But sometimes innovation consists precisely in improving processes and making everything simpler, faster, cheaper, more scalable, more easily distributed than in the past.
Since the invention of photography, false images of another person can be created. Photoshop made it incredibly easy, and the Internet added the possibility of mass distribution. The Artificial Intelligence tools popularized in recent months add a fascinating innovation: they allow us to create any image that crosses our mind in a second.
Little by little we are realizing what that really means. While we are still waiting for a political deepfake video that deceives all of us voters or a false image that destabilizes the stock price of a company, Scarlett Johansson, Rosalía or Laura Escanes are already tired of defending themselves against montages of their face with a naked body. A common phrase on the Internet is that the porn industry is always the first to adopt new developments, and that you have to pay attention because the rest of the content usually follows behind. In other words, we must look at the innovative ways of exploiting young women. Perhaps the La Vanguardia article that I have linked the most times in this newsletter is Artificial porn without consent: what to do if they use your image without permission?, by Susana Pérez Soler.
This week we have once again been pioneers in Spain with one of the first major cases of the creation of child pornography without consent and its distribution through WhatsApp. A dozen boys under 14 have used a fake application to undress around twenty of their friends, neighbors and colleagues, also minors. The scandal has been tremendous. I don't know what tool was used in this case, but there are many capable of doing it, not particularly difficult to find despite being blocked in official stores.
There are many nuances. The use of nudity without consent in porn is not the same as child pornography, nor is extortion or public dissemination the same as private use between adults to feed fantasies. Justice is quite clear.
In the case of child pornography, experts are warning that pedophile material in which children's faces are superimposed on adult bodies is increasing. In a year, researchers expect such a boom in these prohibited images that it will be very difficult to identify which photos are real and they assume that a boy or girl is in danger. Although major imaging tools such as Dall-E, Midjourney and Stable Diffussion take steps to prevent pornographic use, pedophiles have found ways to bypass the restrictions, for example by using an old open source version of Stable Diffussion. and tweaking their code or using languages other than English, much more closely guarded. The Washington Post says that in a pedophile forum they did a survey among 3,000 users and 80% had used or intended to use these tools. The development of AI tools to combat the generation of abusive images with AI is also going at full speed.
Once a technology is released, it is difficult to go back, just as it is difficult to unlearn what is known, and that is why it is important that large companies take responsibility both for launching products that are not sufficiently tested and for their dark side: we have been on the Internet for long enough to to know how important moderation is and what happens in communities when it is not applied.
A perverse effect of the Almendralejo case is that the download of deepnude applications has probably multiplied, exactly the same effect that we saw last week with the popularization of dubbing tools for humorous use. We are simultaneously learning about technology and the worst of its uses.
What else happened this week
One more demand. RR Martin, John Grisham, Jonathan Franzen and other authors have denounced OpenAI for the “systematic theft” of their works.
Amazon is enhancing Alexa with generative AI so we can have more human-like conversations with it. We will be able to chat with 25 characters, from Socrates to Einstein.
A piece of news that is better understood by reading the previous two: Stephen Fry has said that his voice was stolen. Someone fed an artificial intelligence his Harry Potter audiobooks and used it to narrate a historical documentary without his permission.
Xavi Ayén asks Núria Cabutí, CEO of Penguin Random House, about the use of AI in publishing houses: “We think that human creativity is not going to be replaced by artificial intelligence and what adds value to a publishing house is that creativity. Copying is something that machines will surely do well, but in the end what we want are new things, and only human beings have that.”
OpenAI has announced that in October it will release the third version of its DALL-E image generator, which will be natively integrated into ChatGPT. Apparently it is noticeably better than the previous version. This means that, in practice, ChatGPT will be able to create images.
A warning from UN Secretary General António Guterres: “We are hurtling towards a major fracture in economic and financial systems and trade relations, one that threatens a single, open internet, with divergent strategies in technology and artificial intelligence, and frameworks.” potentially conflictive security issues.
What does Barcelona look like in AI? asks Josep María Ganyet in an opinion article. Two things, he says. The first, the Mare Nostrum supercomputer. The second, the scientists: “Mateo Valero in supercomputing, Alfonso Valencia in life sciences, Ramon López de Mántaras in machine learning (ML), Carme Torras in social robotics, Karina Gibert in ML, ethics and explainability and Ulises Cortés representing of knowledge".
How to seduce machines and humans when we are looking for a job, by Raquel Quelart.
Photographer Joan Fontcuberta has centuries of conceptual advantage over those who have just discovered the artistic potential of fictional images. His latest work, exhibited at the Àngels Barcelona gallery, invents hypnotic artificial plants created with Stable Diffusion. By Miquel Molina.
The biography of Elon Musk written by Walter Isaacson is still alive and well. He explained to Alexis Rodríguez-Rata that he sees himself as a superhero capable of taking humans to Mars and making AI safe. “He separated from Sam Altman and OpenAI because he felt that it should be more open and go faster in order to protect the human species from uncontrolled robots, and that is why he founded his own company, called xAI, which will make artificial intelligence robots that They will compete with ChatGPT, but also self-driving cars and other real-world applications. The most interesting of all is Neuralink, where he implants chips in the human brain that will connect directly to computers.”
The Madrid Health Service launches a pilot project to use AI in the diagnosis of rare diseases.
Much attention to DeepMind, Google's AI research laboratory located in London. Some of its researchers have won the Lasker Prize for their “revolutionary technology” for predicting the shape of proteins. It is not strange that Lasker winners later win a Nobel. They have also built a tool to predict whether millions of DNA mutations can be harmful and generate a rare disease.
Google has connected its Bard chatbot to Gmail, YouTube and other tools, but it is not yet available in Spanish.
Following complaints about how Amazon is being flooded with books created with AI, the platform has taken serious measures: it prohibits authors from publishing more than three books a day.
Some American public universities have a problem with the low rate of students graduating. At John Jay College they have managed to increase it by 32% thanks to an algorithm that identifies which students are falling behind.
Level of Anxiety this week: that corresponding to “I could see it coming”.
