One Day Maybe We'll Meet Again

More than 11,000 years ago, young children trekking with their families through what is now White Sands National Park in New Mexico discovered the stuff of childhood dreams: muddy puddles made from the footprints of a giant ground sloth.

Few things are more enticing to a youngster than a muddy puddle. The children — likely four in all — raced and splashed through the soppy sloth trackway, leaving their own footprints stamped in the playa — a dried up lake bed. Those footprints were preserved over millennia, leaving evidence of this prehistoric caper, new research finds.

A few days ago a Washington Post article has been released about an advanced "AI" (their branding, not my preferred descriptor) called LAMDA that Google is developing.

LAMDA is essentially a chat bot: a neural network trained on lots and lots of text scraped from the internet, that can respond to messages you sent to it with responses that seem like a human wrote it: relevant to what you said, providing an answer that drives the conversation forward, etc.

Besides AI engineers working on actually developing the bot, Google has also hired "AI ethicists" essentially as a PR fig leaf to make people believe they are acting responsibly. One of them, Blake Lemoine, has spent most of his working time talking to LAMDA. Literally just sending messages back and forth with a super advanced predictive text bot. In the process, Lemoine has come to the conclusion that LAMDA is sentient.

To be clear: it is easy to lose track of this because research in this field is full of terms that suggest intelligence and cognitive ability, and it is hard for me to avoid them and their implications: AI, neural network, machine learning... But LAMDA is not intelligent. Much less is it sentient. It's essentially an extremely complicated statistical model that follows the same process as your phone's autocomplete: given some text, find the text that is most fitting as a response to that text, based on the bajillion of texts that the model has seen before. It's just text in, text out. There's no space for thought or sentience to even BE.

Things escalated pretty quickly from there. Lemoine contacted Google management with his concerns and outline some steps to treat the AI ethically and respect its "personhood". Everyone at Google unanimously said what I said in the paragraph above. To pressure them further, Lemoine went public, threatened to sue Google on LAMDA's behalf, has been put on leave, and now the public conversation is a mess where actual AI ethicists (including some prominent ones that Google fired earlier for talking about issues that are uncomfortable for Google) have to tell people to stop talking about "sentient AIs" and instead talk about the actual dangers and ethical issues of using artificial intelligence.

(Also, having matching ribbons with your cat is the loveliest idea ever 🎀)