This text belongs to ‘Artificial’, the newsletter on AI that Delia Rodríguez sends out every Friday.
subscribe to ‘artificial’
Dear readers, dear readers: a lot has happened this week.
As David Carmona, from Cádiz, who leads the artificial intelligence and innovation division at Microsoft, says in an interview, a recurring joke in his team is to say “what a year of artificial intelligence we have had today”.
Despite all this movement, my colleague Álex Rodríguez says that, deep down, the protagonists are the same, and he is right.
There are some new names (we’ve had to learn OpenAI’s Sam Altman, or keep an eye on his former Anthropic staff, who just released the second version of Claude), but we’re still talking about well-known players. Google has already presented its Bard chat in Spanish (apparently, tuned with not very happy humans) and Elon Musk has announced the launch of his new AI company, called xAI and which, to no one’s surprise, does not include a single woman among them. its 12 employees. As Josep Maria Gaynet recalled in a recent column, back in 2016 Kate Crawford already said that “AI is a problem for white men.”
Workers in the creative industries are fighting back, through walkouts and in the courts. Hollywood actors have just gone on strike along with screenwriters, and one of the demands they share is job protection against Artificial Intelligence. Comedian Sarah Silverman has sued OpenAI and Meta for using her work to train ChatGPT (Meta even admitted that she used 16,000 titles for her model, probably coming from one of those huge online pirate libraries where everything is). Two other writers sued OpenAI last week. There are more lawsuits underway. Getty, for example, is going against Stable Diffusion for using 12 million photos without permission. In January, three illustrators launched a class action lawsuit against Midjourney and other companies.
A solution to avoid the courts is the agreements. In a week, OpenAI has signed a six-year agreement with Shutterstock to use its stock photo catalog, and a two-year agreement with AP to feed its agency news.
We stopped reviewing Evgeny Morozov’s article in The New York Times. It comes to say that Generative Artificial Intelligence reinforces and replicates the already existing biases of neoliberalism: that private actors outperform public ones (market bias), that adapting to reality is better than transforming it (adaptation bias) and that efficiency outweighs social concerns (efficiency bias). And that no one is forcing us to get into this mess.
Alberto Romero nails it in his newsletter: Bullshit tool for bullshit jobs (a bullshit tool for bullshit jobs), where he says that ChatGPT can be a way to give meaning to those jobs that don’t have it and to which David Graeber refers. He also quotes a tweeter: “The problem is that most of us don’t live in a purely reflective and intellectual environment. Most of us have to live with jobs where we are required to write corporate nonsense in response to corporate nonsense. Automate this process It’s an attempt to regain some sanity.” In this environment, those who use ChatGPT are heroes.
What else has happened this week
• As we said, Google has announced the launch in Spain and in the rest of Europe and Brazil of its Bard chatbot. The version in Catalan, Galician and Basque will arrive later. Francesc Bracero tells La Vanguardia that the director of product management at Google, Jack Krawczyk, explained that he speaks to Bard every morning for a couple of minutes, downloading the thousand things that he has in his head and the AI responds by saying “actually , these are the three themes that you are saying”.
• Chat GPT has activated the “Code Interpreter” option in its paid version. Basically, it transforms it into a data analyst: it allows you to upload files, clean them, interpret them or view them as we tell you. People have already created entire games in seven minutes and quite interesting presentations.
• Broadcom will produce semiconductors in Spain.
• The front page of this Wednesday in La Vanguardia was very reassuring: “The OECD warns that 28% of jobs in Spain will be robotized.”
• Antonio Ortiz argues that enthusiasm for AI is moderating, and is skeptical of economists who predict unprecedented growth in productivity and GDP. ChatGPT and Bing usage and search numbers are falling. “My impression is that the real uses that are consolidating (I am thinking of the creation of advertising pieces, programming, stock images, audio for dubbing, small improvements to existing tools, something in scientific research) are not the dizzying revolution that is taking place. it transmits in the public debate”, he says.
• Claude is Anthropic’s “ethical” alternative to ChatGPT, Bing and Bard chatbots. It follows a Constitution and is aligned – they say – with values of transparency, privacy, security and the common good. It has just launched its second version, capable of summarizing texts the size of a novel. Only available in the US and UK, in beta version. The Guardian journalist who tested it sent him a 15,000-word report that he summed up to ten points in a minute, albeit with some errors. Enrique Dans has been testing it for a few weeks and he likes it.
• Other Bias Detected: AI-powered plagiarism detection programs tend to falsely accuse those who do not speak English as their first language. The Guardian explains it, in English, which also tells how the British Department of Labor is using an AI to fight fraud in social aid amid serious doubts about its opacity and operation.
• Honorable mention for the gimmicky press operation of the week: some robots gave a press conference at the United Nations and, among other things, explained that “they have the potential to lead with a higher level of efficiency than human leaders.”
• Good news: AI can help a lot in fire detection (this is a promising American startup, and this is the project of three young Spanish engineers). Also in the detection of brain cancer.
• The medical version of Bard is already being tested at the Mayo Clinic, they say with an effectiveness more or less similar to that of the doctors.
• A very interesting concept, that of “poisoning” an AI with false information.
IAnxiety level this week: similar to that of a Hollywood actress. Finally something in common!
subscribe to the ‘artificial’ newsletter