Welcome back to The Prompt.
AI search startup Perplexity appears to be plagiarizing content from multiple news outlets including Forbes, Bloomberg and CNBC through “Perplexity Pages,” a feature where users can generate content on a specific topic. A post created by the Perplexity team includes a custom illustration, several sentences and details that were first reported by Forbes in an exclusive story about Eric Schmidt’s stealth drone project. These curated “pages” have already amassed tens of thousands of views and originally did not mention their source publications by name, instead making attributions in tiny, easy-to-miss footnotes. In response to tweets about the issue, CEO Aravind Srinivas said the product has “rough edges” and that it planned to incorporate feedback. After Forbes flagged the issue, Perplexity updated the layout to more prominently credit source publications at the top of these pages.
Forbes Chief Content Officer Randall Lane wrote that Perplexity’s treatment of premium journalism–ripping off paywalled Forbes content and pumping it out to its subscribers for free across a variety of formats– is “the perfect case study for this critical moment” in AI, which has wide-ranging implications across the publishing industry and beyond.
“I’m an AI bull, and in the right hands, productivity and advances and prosperity await,” Lane wrote. ”But in the hands of the likes of Srinivas — who has the reputation as being great at the PhD tech stuff and less-than-great at the basic human stuff — amorality poses existential risk.””
Now, let’s get into the headlines.
BIG PLAYS
Apple announced a legion of new and updated AI features under the umbrella term “Apple Intelligence” (which it tagged as “AI for the rest of us”) at its annual worldwide developers conference Monday. The tech giant unveiled a significant AI makeover for Siri, which it said will now be able to retrieve information from different apps, interpret commands more accurately and communicate in a more natural way.
Apple said that it is looking to deeply integrate generative AI across its products, like through AI-generated caricatures and emojis from photos of your friends. It’s also enabling summaries of news articles in Safari’s “Reader mode” and the ability to rewrite emails or notes in different tones with AI. Apple said you will also be able to ask Siri to find information from your iMessages or email, like details about a dinner reservation.
While many of these announcements mirror AI capabilities that tech giants like Google and Microsoft have already rolled out, Apple claims that its big differentiator is data privacy, stating that its generative AI features are powered on “private cloud compute” and deployed locally on the device while the models run on Apple’s own servers. Apple is also partnering with OpenAI to bring ChatGPT to the iPhone, MacBook and iPad with iOS 18, where users can opt to use OpenAI’s GPT-4o instead of Apple’s AI model to answer queries, though they will be asked for consent before sending photos or documents to OpenAI.
ETHICS+LAW
Personal details and images of Brazilian children have been included in a popular dataset that has been used to train popular AI text-to-image generators like Stable Diffusion and Midjourney, according to a Human Rights Watch report. The pictures were scraped from mommy blogs and Youtube videos. The report comes after Stanford researchers found that the dataset also contained child sexual abuse matierial.
REGULATION
The Federal Trade Commission is investigating whether Microsoft designed its deal with AI startup Inflection to evade a regulatory antitrust review of the transaction, The Wall Street Journal reported. In March, Microsoft hired Inflection CEO Mustafa Suleyman and most of the company’s employees and inked a $650 million deal to resell its technology. Although the deal was an “acquihire,” where a company’s talent is picked up rather than its products, companies are required to disclose acquisitions worth $119 million or more to antitrust enforcement agencies so they can assess a deal’s effects on the competitive landscape.
AI DEALS OF THE WEEK
Legal AI startup Harvey is in talks to raise $600 million at a $2 billion valuation, according to The Information. Part of that investment could be used to acquire vLex, a 25-year-old company that maintains a massive repository of legal and regulatory information. Harvey was cofounded in 2022 by roommates Winston Weinberg and Gabriel Pereyra. More than 10,000 lawyers use the startup’s large language models to research cases, find precedents and draft legal documents.
Additionally, French AI model maker Mistral closed a $643 million Series B funding round at a $6.2 billion valuation.
DEEP DIVE
Last year, billionaire technologist Eric Schmidt quietly founded a secretive military drone company, White Stork. Now, the stealth startup has begun testing its artificial intelligence-guided aircraft, both at the Menlo Park headquarters of Schmidt’s family office Hillspire and on the frontlines of the war in Ukraine, where the former Google CEO has continually touted himself as the country’s preeminent defense tech guru.
After Forbes revealed the venture in January, Schmidt quietly renamed it and accelerated its development, which sources in a position to know said involves using artificial intelligence to help drones home in on battlefield targets. Neighbors of Hillspire’s 25,000 square-foot office block, which sits between a row of homes, restaurants and a Caltrain station in the wealthy suburb, recently spotted individuals flying small drones from the building’s gated courtyard. And two people familiar with Schmidt’s activities in Kyiv told Forbes that his team has been testing drone prototypes with the Ukrainian Ministry of Defense and is soliciting their feedback.
White Stork’s drone development has been aided by a steady stream of notable hires. Over the past several months, it has poached at least a dozen employees from Apple, SpaceX, Google, federal agencies and the billionaire’s own philanthropic organization, Schmidt Futures, multiple sources told Forbes. Their expertise spans machine learning, aerospace, supply chains and procurement. These tactical hires have been accompanied by rank and file recruitment at universities and AI hackathons, some personally hosted by Schmidt himself.
Despite a series of public visits to Kyiv, Schmidt has labored to operate his drone project in stealth. Shortly after Forbes broke the news of his plans for White Stork, the startup rechristened itself “Project Eagle,” according to three sources familiar with the effort. Domain name records for “projecteagle.net,” the email address now used by the project’s members, show that it was newly registered in February.
Read the full story in Forbes.
AI INDEX
To carry out elaborate tasks, AI systems rely on electricity-guzzling data centers. Now a host of companies are developing alternative forms of energy like small nuclear reactors and sodium-ion batteries to satiate generative AI’s limitless power needs.
10 times
The amount of electricity ChatGPT prompts consume compared to Google searches.
160%
The increase in demand for power by 2030 driven by AI applications, according to a Goldman Sachs report.
8-10 Kilowatts
The amount of electricity that each Nvidia H100 chip uses. Cooling a data center full of them requires 30% more water than a conventional one.
YOUR WEEKLY DEMO
In April, Humane’s AI pin, a hyped wearable device that aimed to do everything a phone could with large language models, was hit with scathing reviews. On Wednesday, the company told its customers to “immediately” stop using its charging case due to a fire risk associated with its battery. Over the past few months, Humane’s cofounders, Imran Chaudhri and Bethany Bongiorno, have tried to sell the company to HP for over $1 billion, according to Bloomberg. But issues like overheating were well known internally long before the product launched publicly, The New York Times reported. Multiple former and current employees had raised concerns about the device’s technical shortcomings to the company’s founders to no avail. For instance, before demoing the gadget to prospective investors, company executives would chill it on ice packs so it would last longer. Founded in 2018 by former Apple executives, Humane is backed by prominent figures like OpenAI CEO Sam Altman and Salesforce CEO Marc Benioff.
QUIZ
The law that could ban TikTok in the U.S. amid national security concerns could also ban this ByteDance-owned popular AI video editing app:
- Gauth
- CapCut
- Captions
- InShot
Check if you got it right here.
MODEL BEHAVIOR
Tesla CEO Elon Musk appeared to have a visceral reaction to Apple’s AI announcements. After Apple announced that it plans to make ChatGPT available on its marquee devices like iPhone and MacBook, Musk threatened to ban Apple devices at its companies if the AI model is integrated at the OS level, citing privacy concerns. “Apple has no clue what’s actually going on once they hand your data over to OpenAI. They’re selling you down the river,” he wrote on X.
