Artificial Intelligence

Or, the luxury of logging out:

The idea of “analog privilege” describes how people at the apex of the social order secure manual overrides from ill-fitting, mass-produced AI products and services. Instead of dealing with one-size-fits-all AI systems, they mobilize their economic or social capital to get special personalized treatment. In the register of tailor-made clothes and ordering off menu, analog privilege spares elites from the reductive, deterministic and simplistic downsides of AI systems.

In a letter sent to Governor Gavin Newsom, Anthropic CEO Dario Amodei says the benefits of SB 1047 “likely outweigh its costs.” However, he still has concerns about government overreach and suggests maintaining a “laser focus” on catastrophic risks.

Meanwhile, OpenAI’s letter to a California senator says the bill could slow progress, and that AI regulation should be left to the federal government.

Colorado tech law professor Blake Reid has a good Bluesky thread (note: requires login) on the complicated gamesmanship behind AI content deals and copyright law. His conclusion:

At the end of the day, copyright just doesn’t give a lot of people positively and negatively impacted by copyright a seat at the table. And these deals are a powerful reminder of that.

I got Pixel Studio, Google’s AI image generation tool on the Pixel 9, to come up with some questionable things. But they’re nothing like what the folks over at Digital Trends got out of it, which feature popular cartoon characters firing AK47s, drunk driving, and donning Nazi uniforms. Google seems to have clamped down on some of these, but oof, not a good look.

The district hired AllHere to create and manage the chatbot. But the tech firm lost its CEO and furloughed most of its staff just a few months after the district launched its pilot program. Now the chatbot is no longer accessible and the district is investigating allegations that AllHere compromised its data. What a mess!

OpenAI announced it’s partnering with Condé Nast, which owns publications like The New Yorker, Vanity Fair, and Wired. OpenAI will display Condé Nast’s content in its new AI-powered search engine prototype, SearchGPT (but provided no details on if it’s using Condé’s content as training data).

These media/AI company deals are becoming more common because media execs seem to believe that accepting the money, rather than laying off staff to afford lengthy legal battles, is the best option for now. (Also, Vox Media has a partnership with OpenAI.)

Sam Altman’s crypto-adjacent Worldcoin has irritated global regulators:

It has been raided in Hong Kong, blocked in Spain, fined in Argentina and criminally investigated in Kenya. A ruling looms on whether it can keep operating in the European Union.

AMD is acquiring ZT Systems, a leading provider of AI infrastructure. AMD is calling it a “next major step” for its AI training and inferencing solutions, in a move that will clearly help it compete with Nvidia’s dominance in AI offerings. ZT Systems will join the AMD’s data center solutions group once the $4.9 billion transaction closes.

The company closed the waitlist for its “prototype” generative search product, sending out emails like the one below to signed-up users who weren’t chosen to test it.

The company has said only 10,000 users will get access at first, which could help it if its searchbot gives bad recommendations like gluing slippery cheese to pizza.

Jessica Grose writes in the NYT about educators struggling with students using AI in the classroom. One major worry expressed is that relying on it for brainstorming and writing could make students less likely to power through tough assignments on their own.

It’s almost as if the speed of available technology is making them assume that their human brains should have all the answers.

Right now, teachers have to deal with this issue on their own; some policymakers “appear to have drunk the Kool-Aid on artificial intelligence.”

Opting out of Google’s generative AI overviews means you become invisible in search — a no-go for most publishers. But keeping content in search means it can be scraped for AI Overviews. As one publisher puts it:

You drop out and you die immediately, or you partner with them and you probably just die slowly, because eventually they’re not going to need you either.”

It’s seemingly easy to make the chatbot’s new image generator spit out the few things it supposedly can’t generate — including gore and even “child pornography if given the proper prompts,” says X user Christian Montessori.

While all AI models have loopholes, Elon Musk seems unfazed by the abuse, calling it a “step for people to have some fun.”