The week in AI: Nightshade, the data poisoning tool helping artists fight back against generative AI

Presented by

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each Thursday, we aggregate the major developments in artificial intelligence; we pass along the news, useful resources, tools or services, and exciting projects in open source. Even if you aren’t an engineer, we’ll keep you in touch with what’s going on in AI.

And as the article notes: generative AI models are excellent at making connections between words, which helps the poison spread. Nightshade infects not only the word “dog” but all similar concepts, such as “puppy,” “husky,” and “wolf.” 

The tool could help put some power back in artist’s hands. Most generative AI services now offer opt out policies to ‘help’ artists, authors, etc. ensure their works aren’t being scraped without their consent to train an AI model - but requiring creators to opt out of an increasing number of individual AI systems is unreasonable. Some companies like Shutterstock have avoided this thorny issue by only training their models with content that they actually have rights to. That was a smart move: there are more AI/data-scraping lawsuits open than you can shake a stick at.

And clearly, some artists aren’t interested in waiting for the courts to fight their battles.

Microsoft released a glowing earnings report on Tuesday, and Forbes is citing AI as the driving force behind an all-time record high quarterly revenue. Microsoft started 2023 with a $1.7T market cap; they’re now over $2.5T and closing in on Apple ($2.7T) as the world’s most valuable company - a title Apple hasn’t forfeited since obtaining it in 2011.

A few days prior to the earnings call, Microsoft CEO Satya Nadella wrote his annual letter - and unsurprisingly, it is almost entirely about AI. It’s a good read to get some insight into just how deeply the company has already successfully integrated the technology across their platforms and multiple industries - and how they are continuing to double down on it.

The problem of misinformation will likely get even worse as AI text and image generators become more capable/accessible and spread false or manipulative content at scale. Lessig argues that the First Amendment puts real limits on existing legal solutions like content moderation or direct speech regulations. Instead, he advocates finding ways to foster healthy democratic deliberation in more protected spaces, outside the reach of engagement-obsessed, profit-driven algorithms. Lessig points to the rise of "citizen assemblies" in places like France, Ireland, and Japan as a promising democratic innovation - representative groups of citizens who come together in person for informed, substantive debate on major issues.

Bloomberg took a deep dive this week into what Apple is currently doing in AI. Apple has been relatively quiet (compared to other tech giants) on the AI front, but it’s clear from the article just how much of a company-wide focus there has been to ‘catch up’ with the AI frenzy. The article examines the company’s ongoing debate about how to deploy generative AI (on-device vs. cloud vs. something in between) and what’s going on with major product lines like the Vision Pro, iMac, MacBook Pro, etc.

Apple just announced a mysterious “Scary Fast” online event, livestreaming October 30th just ahead of their Nov. 2 earnings call. Given the title, we're expecting some AI-in-mind announcements touched upon in the Bloomberg article.

The UK's massive upcoming global AI summit, spearheaded by Prime Minister Rishi Sunak, is drawing criticism for its narrow focus on speculative dangers of advanced AI rather than practical governance. While Sunak touts the summit as cementing Britain's leadership in AI safety, UK experts argue it ignores immediate real-world AI policy needs and sidelines the country's own AI industry. With likely minimal participation from key global players like the US and China, limited concrete proposals, and a questionable choice of venue, the summit seems unlikely to advance substantive international AI governance. Critics contend it is more political theater than meaningful progress.

More in AI this week:

• Israeli “war room” mobilizes AI experts to locate Hamas hostages

• Thanks to AI, the future of programming may involve YELLING IN ALL CAPS

• Ethan Mollick on the “Best Available Human” standard for AI

• Music industry sues Anthropic (creators of popular chatbot Claude) for alleged ‘theft of lyrics’

• US surprises Nvidia by speeding up new AI chip export ban

• Meta’s AI Chief Yann LeCun says, “Don’t regulate us, AI just isn’t that smart”

• AI tidies up Wikipedia’s references - and boosts reliability

• UK officials use AI to decide on issues from benefits to marriage licenses

• Inside Sequoia Capital’s growing AI portfolio

• Microsoft to spend $3.2b in Australia as AI regulation looms

The AI team you didn’t know your company needed — until now

Hire world-class AI experts from Harvard, Stanford and MIT

Not sure how to implement the right AI strategy for your product? Hire AE Studio's world class team of software builders to craft and implement the optimal AI solution for your business.

Our development, data science and design studio work closely with founders and executives to create custom software, machine learning and BCI solutions.

From custom-built MVPs to bespoke AI/ML solutions, see how you can leverage AI to achieve your business objectives. 

Schedule a Free Scoping Session.

That’s it for AI this week. We’ll see you next Thursday!