"To plumb the ocean of this Jovian moon, engineers must [. ] get missions to survive a never-ending rain of radiation that fries electronic circuits. They must also plow through an ice shell that’s at least twice as thick as Mount Everest is tall.".

A new database aims to compile a comprehensive list of the risks posed by AI, including details about:. * When in the pipeline risks occur. * What type of harm (bais vs misinformation vs compromised privacy . ).* The root cause. * Intentionality.

Join millions who have switched to Grok.

It's the little things: at least we can agree not to give AI the nuclear codes.

Hmm. I wonder which proposed cabinet member asked for this change.

Interesting take on the splintering of social media. Will the monolithic firms that dominated the 2010s (Facebook and Twitter) be a historical aberration? This writer thinks so:.

Algorithmic advertising — like many software systems — is a massive tangle. It's hard to know where all the endpoints are and what "emergent properties" there might be. Be wary when rolling out new features, or you'll end up like BuzzFeed.

Google is attempting to bake fact-checking into its AI models, with some early signs of success. Not infallable, but much better than the raw model.

Driven in part by infinite patience, LLM Chatbots effectively talked people out of believing certain conspiracy theories.

A look behind the "creative accounting" that data center operators use to claim "carbon neutrality." ML models are driving major spikes in data center construction and energy use.

It appears that arresting the CEO of Telegram has resulted in a greater willingness to cooperate with law enforcement. I — for one — am shocked. 😮 .

Gary Marcus — a prominent Gen AI critic — gave an interesting interview to IEEE Spectrum. A snip:.

A new ML model for robotics is attempting to power many embodiments with a single model, with some decent success. The model is too large to run on-device, though, which is a serious limitation for many robotics applications.

PSA: For the love of all that is holy, do not just trust AI to identify or accurately depict a mushroom that you are going to eat.

The California AI Safety bill is pretty controversial among experts. Anecdotally, I've noticed the more someone believes AGI is just around the corner, the more they support the bill, and vice versa.

How much will "AI Scientists" help drive progress in science?.

In some ways this is a little silly, but in others it's a real condemnation of the biases inherent in training models with data generated by society writ large: AI Image Generators can't make a realistic Kamala Harris.

There are currently 120+ AI-focused bills in Congress. Most won't pass, and many won't even make it out of the committee, but it shows that lawmakers actually have a sense of urgency about these products and tools.

I teach debate in a high school, I teach programming and AI/ML to professionals, and I'm reasonably online: "Is that AI, or does it just suck?" is a question I ask myself far too often. On the plus side, this article will give you a laugh.

Russian agents and state-operated media buying millions of dollars worth of video spots from very-online right-wing YouTubers is funny, totally predictabl, and probably not so healthy for our media ecosystem.

I think it's probably for the best that teenagers are being shoe-horned into a more limited, private experience on Instagram.