Last one before the end of the year recap, where I will try to write about the things I enjoyed the most this year in different topics, as usual. Let’s start:
AI and the information age
Two interesting reads this week on AI and automation. First, what is going to happen when future generations of AI-based text models are trained on text scrapped from the internet, which contains huge amounts of text generated text by current AI’s? And the problem is not unique for text (it might be even worse for pictures). I have the feeling that most of the articles I have read this year talking about negative outcomes of AI boil down to the original problem of how to train these algorithms (i.e., the generation/maintenance of curated training sets), which is something none of the big players in the field seem to really care about.
Large language models are trained on data sets that are built by scraping the internet for text, including all the toxic, silly, false, malicious things humans have written online. The finished AI models regurgitate these falsehoods as fact, and their output is spread everywhere online. Tech companies scrape the internet again, scooping up AI-written text that they use to train bigger, more convincing models, which humans can use to generate even more nonsense before it is scraped again and again, ad nauseam.
And a little bit of cyberpunk to end the year. Imagine a little kid in the 90’s, telling his grandparents that in the future, vacuum cleaners will take pictures of the people in the house and share them online without their consent. What would be their reaction? Probably they would have said that the idea was nonsensical and that the future would be full of clean energy, flying cars, and universal healthcare. But we are in 2022, Elon Musk is one step away from becoming a James Bond villain, millions of people have to choose between going to the doctor or buying food, and your Roomba is uploading pictures of you peeing to the cloud. What a time to be alive.
The images were not taken by a person, but by development versions of iRobot’s Roomba J7 series robot vacuum. They were then sent to Scale AI, a startup that contracts workers around the world to label audio, photo, and video data used to train artificial intelligence.
They were the sorts of scenes that internet-connected devices regularly capture and send back to the cloud—though usually with stricter storage and access controls. Yet earlier this year, MIT Technology Review obtained 15 screenshots of these private photos, which had been posted to closed social media groups.
How AI-generated text is poisoning the internet, on the MIT Technology Review
A Roomba recorded a woman on the toilet. How did screenshots end up on Facebook?, on the MIT Technology Review
Good news? bad news? Who knows…
Crossing fingers on this really happening before 2025 ends:
US industrial conglomerate 3M on Tuesday set a deadline of 2025 to stop making PFAS, also known as “forever chemicals”, that are used in everything from cellphones to semiconductors and have been linked to illnesses ranging from cancer, heart problems to low birth weights.
3M sets 2025 deadline to stop making ‘forever chemicals’, on The Guardian
Another plan that will never be accomplished, made without hearing all the voices that really matter on the subject, and with no real containment measures.
A key target it sets is for nations to protect and restore 30% of the world’s land and seas globally by 2030, while also respecting the rights of Indigenous peoples who depend on and steward much of Earth’s remaining biodiversity. Another target is for nations to reduce the extinction rate by 10-fold for all species by 2050.
This on the same week we saw how scientists are giving up on trying to save the ecosystem, and have started proposing to freeze coral so that future generations can grow it again IF we undo climate change. In my opinion, it is bold to assume that there will be any future generations and that they will be able to devote their time to this instead of searching for drinking water.
Nations forge historic deal to save species: what’s in it and what’s missing, on Nature
Scientists freeze Great Barrier Reef coral in world-first trial, on Reuters
Another round of recaps
Another year, another DnB recap by the people at Liquicity. First time without Maduk if I recall correctly.
Also, a couple pieces by the people at The Verge on things space and video games (I still cannot wrap my head around the Stray craze) that happened in 2022, one article on the tech policies that defined 2022 at The Washington Post, and one of the last strongholds of written text in the sea of social networks.
2022: a space yearbook, on The Verge
These were the video games that defined 2022, on The Verge
These tech policy stories defined 2022, on The Washington Post
My Year of Reddit and Relaxation, on The New Yorker
Now you see it…
How do frogs become invisible at land? Short answer: we do not know. Long answer: it seems that they put almost all their blood in the livers at night, and fill their body with plasma. Amazing.
By beaming lasers at the frogs, the scientists were able to track the movements of individual blood cells as the animals fell asleep and then woke for their nocturnal jaunts. The team found that as the frogs hop around dreamland, their vasculature fills almost entirely with plasma—colorless, save for a gauzy bluish tint—interspersed with just a few red cells, turning their body two or three times as transparent as it is while the animals are awake. Even the blood-cell-filled liver plays its own deception game: The organ’s outside is coated with a film of tiny, reflective crystals, which essentially conceal the redness behind a veil of white.
How Glass Frogs Weave the World’s Best Invisibility Cloak, on The Atlantic
And that’s it for the week. Stay safe!
Now listening: Liquicity Drum & Bass Yearmix 2022 (Mixed by Andromedik)
Featured image: Glass Frog – National Geographic – Jaime Culebras