Welcome to Weeklies, where I recap some of the weird links I’ve shared on the homepage over the past week!
Doomsday Asteroid Update
Bad news for anyone who might have been looking forward to a light show in 2032: Asteroid 2024 YR4 has had its impact probability updated, and has dropped from around a 3.1% chance of hitting Earth to a measly 0.28% chance. This is likely to go down even further.
“As more observations are collected, and that data is added to the orbit calculations, it is likely this probability will drop to zero.”
NASA
According to CBS News, when an asteroid reaches a 1% chance of hitting Earth, agencies are required to notify the public. The probability is expected to fluctuate.
However, despite the low odds, space orgs are still taking the idea seriously. Asteroid 2024 YR4 is a 130 to 300 ft diameter asteroid that could do actual damage if it collided with Earth, equivalent to about 500 Little Boy bombs. These probabilities have actually been very specific, as well. They’re the odds that the asteroid will impact Earth specifically on December 22, 2032 at about 14:02 UT.
Social Somethings
Heard of any good Bigfoot sightings lately? Here’s one: Saxsquatch was spotted out in the woods with a synth playing “Carry On Wayward Son” about a month ago. Put your blurry cameras away; the guy just wants to jam.
Over in a North Dakota cemetery, meanwhile, a small group of turkeys were spotted suspiciously circling a gravestone. “It was quite bizarre” said the witness. What were the turkeys up to, and should we be concerned? (This is actually fairly normal turkey behavior, believe it or not.)
Do Older AIs Suffer Cognitive Decline?
A recent and somewhat controversial study looked at how older Large Language Models show symptoms of cognitive decline. Researchers used a cognition test designed for humans, known as the Montreal Cognitive Assessment or MoCA test, and applied it to some popular AI chatbots, including Google’s Gemini and OpenAI’s ChatGPT.
What they found was that LLMs, particularly older versions, score poorly on visual/spatial skills and executive tasks. This might not be too surprising – the study has met some criticism from those who believe you can’t just take a human cognition test and slap an AI with it. An old version of Google’s Gemini performed particularly poorly, scoring only 16 out of 30. According to Live Science, 26 out of 30 is considered a passing score. A recent version of ChatGPT scored the highest with 26.
Other Links of Interest
Did Isaac Newton calculate the end of the world? Maybe!
Bipedal Musculoskeletal Android V1 motions menacingly.
Ghosts at London Underground Station?
Has NASA Discovered a ‘Parallel Universe’ In Antarctica? Maybe not.
Why can’t we remember our lives as babies or toddlers?
2,000-year-old spoon from Isle of Man may have been used in blood rituals for fortune telling.
Does end-of-life brain activity prove humans have souls?
Scientists use crystals to store terabytes of data in millimeter-sized memory.
That’s it for this week’s edition of Weeklies. Check the homepage daily for new news, or catch last week’s Weeklies for slightly older news!
