Newsletter 'Artificial' | The devil is in the details

This text belongs to 'Artificial', the newsletter that Delia Rodríguez sends to La Vanguardia subscribers every Friday.

Oliver Thansan
Oliver Thansan
22 May 2023 Monday 13:39
4 Reads
Newsletter 'Artificial' | The devil is in the details

This text belongs to 'Artificial', the newsletter that Delia Rodríguez sends to La Vanguardia subscribers every Friday. If you want to receive it, sign up here.

Dear readers, dear readers:

This is the second edition of Artificial, La Vanguardia's weekly AI newsletter. Thank you very much for receiving the first letter, we are already several thousand! I take note of the comments that you have sent me, especially the one that advises me not to use so many links, because they are overwhelming, and I will proceed to not pay too much attention to them: this is what this newsletter is about, to cope with IAnxiety among all of us. Let's accept that there is always something more important to read. Let us begin.

This week the big topic is REGULATION, and it is better to start learning about the details because they will be important in the coming years.

Sam Altman, the CEO of OpenAI (the developer of ChatGPT), has testified before the US Senate and has called for regulating AI because it can be dangerous. The chronicles highlight the good atmosphere in his appearance, contrary to what happened on other occasions with Mark Zuckerberg, for example.

It's strange for a manager to ask for laws limiting his own activity in a conversation blithely citing nuclear weapons controls as an example, so we have two options: either he's a sensible person , a leader at the level of History concerned with the good of Humanity or, ahem, there may be other interests, such as creating a framework of danger and science fiction around the AI ​​that inflates its value and serves as a diversionary maneuver on "small" problems that already exist (discrimination, opacity or misinformation, for example); that it be an attempt to leave out other actors from the beginning, positioning themselves as preferred interlocutors and hoping to write the future fine print; Or that it is mere cynicism, because while time is being gained by calling for regulation, developments continue to advance at full speed, de facto ruling out a pause in their development. The devil is in the details, as an expert consulted by The New York Times says. Francesc Bracero, in La Vanguardia. In English, in The Washington Post.

While the United States begins to consider the issue, Europe and China advance their respective legislation in a global race for regulation (Foreign Policy, in English). Being first matters, because -again- the framework is created: as Axios says, “if China is first in AI governance, it can project its standards and regulation globally”.

Behind the laws, as always, fears that they could slow down innovation or competition from startups that cannot risk, for example, the fines that European regulations are raising. Read about it, for example, the opinion about it from my boss. Faced with these reservations, I can only think of a very exaggerated and very true tweet. I translate: “What is your favorite technological innovation? The illegal taxi company, the illegal hotel chain, the counterfeit money for criminals or the plagiarism machine? Regarding the need to regulate, I recommend the interview that Miquel Molina did a few weeks ago with the digital policy expert Francesca Bría.

What else has happened this week

- A 23-year-old influencer, Caryn Marjorie, has created a virtual version of herself by feeding GPT-4 with 2,000 of her YouTube videos. Through her Telegram, her fans can chat with “her” for a dollar a minute. She made $100,000 in one week. The problem is that the app began to generate sexual conversations (Fortune or The Washington Post, in English).

- Claude, the Anthropic chat, is already able to assimilate a context of 75,000 words, like a book, much more than its competitors. By Mayte Ríus in La Vanguardia, which she also explains how a "constitution" follows, a series of principles and values. The company has also reached an agreement with Zoom, which Claude will apply in virtual meetings.

- I have tried Luzía, a Spanish artificial intelligence application that was born a month ago and is being introduced to society. It is added as a contact to WhatsApp and serves, among other things, to transcribe the audios that are sent to us. It works very well. One of the founders says that it is named after a classmate who had the best notes at the university (this sounds familiar to me). In The Vanguard. On the occupational risk that translators and interpreters run, this interesting article by Carina Farreras.

- Not only have I tried to work less by delegating the transcriptions of the audios. I have contracted the paid version of ChatGPT-4, with an internet connection, and I have asked him to do the same thing as the tweeter Antonio Ortiz talked about in Monos Estocásticos: to summarize the AI ​​news of the week and cite his sources. Without hesitation he selected Altman's appearance, as did I, and provided me with a good description and a respectable list of links. It doesn't have much merit - this week the main topic was clear for any journalist - but it has left me with some concern (Level of IAnxiety: 6/10, we will have to keep trying). Ramón Peco has also tested the connected ChatGPT thanks to the integration of plug-ins.

- Sundai Pichar, the CEO of Google, gave an interview to The Verge where he said that the company has been fully focused on AI for seven years, that ChatGPT found the moment when the public was ready for it, which does not seem like a good idea to him that an AI sends mails and another AI answers them in an endless loop and many other things, including the future of searches. To read or listen calmly, in English.

- Meta, after investing billions in LLaMA has decided to release its code making it available to anyone, in a move opposed to that of Google with Bard and Open AI with ChatGPT, very, very obscure proprietary systems. The intention behind such generosity: to end up leading the market thanks to the traditional virtues of open source, an innovation accelerator. Thrilling. In The New York Times, in English.

- OpenAI has released the ChatGPT app for iOS in the US. It will soon reach other countries and Android, they say.

- We keep asking who the best informed follow. Platformer founder and co-host of The Hard Fork Casey Newton recommends subscribing to Jack Clark's Import AI. “He is the co-founder of Anthropic, one of the most interesting AI startups. Jack was a journalist and his weekly AI newsletter always tells me something interesting,” he says.

*Image: I have also asked ChatGPT for ideas to illustrate this text. It suggests the scene of the Blade Runner 2049 test, something interesting because in the first installment I chose Sean Young behind the scenes of the original film and today we have talked about an interrogation, an appearance. I find it slightly disturbing and inappropriate and go with the image of a small monkey generated in Midjourney by someone called Cakedroid.