Google continues to press ahead with its AI-enhanced vision of the future. This week, the company announced new versions of its Pixel hardware — including its smartphone, watch and earbuds — that feature generative AI enhancements powered by its Gemini chatbot.
CNET’s Abrar Al-Heeti summed up last week’s Google news this way: “When Tuesday’s Made by Google keynote rolled around, the tech giant dedicated the first half hour to talking up Gemini, even going so far as to demo AI updates on other Android devices like the Samsung Galaxy S24 Ultra — before finally unveiling the Pixel 9 lineup. Welcome to hardware reveals in the age of AI, where highly anticipated gadgets take a back seat at their own party.”
The Pixel 9 update includes a new processor, more memory, a higher-resolution ultrawide camera, a bigger battery, and a slightly larger screen. It also has an AI-powered Android 14 operating system that includes everything “from AI-generated weather summaries to a new image generation app called Pixel Studio for creating pictures on the fly,” wrote CNET mobile reviewer Lisa Eadicicco.
“Among the standout new features is an app called Pixel Screenshots, which makes it possible to ask questions about screenshots stored on your phone. That might not sound too exciting, but think about how many times you may have screenshotted something for later reference — perhaps a Wi-Fi password to your Airbnb or an address from a text message,” Eadicicco added. “The Pixel Screenshots app lets you search for those individual bits of information as you need them.”
She also pointed out a new phone feature that many of us may find interesting. Called Add Me, it’s aimed at helping you take group photos. You actually take two, and in the second one, the original photographer swaps with someone else so they can join in the group photo. An AR overlay shows the person taking the shot where the original photographer should stand so when the two images are merged, it all looks natural. Pretty cool, right?
Google’s Pixel phones (including the 9, 9 Pro and 9 Pro XL) enter the market as the company’s rivals also prepare to step up their AI offerings. Market researcher IDC predicts that shipments of gen AI phones, or phones with specific specs for running on-device generative AI models, will surge 364% year-over-year in 2024.
In July, Samsung, the world’s largest maker of smartphones, added AI to its popular Galaxy Z phones as part of a suite of services and features called Galaxy AI. Apple’s next model of the popular iPhone, typically released in September, will have AI integrated into the iOS operating system, the company said in June, as part of a new AI push it’s calling Apple Intelligence.
Google’s Pixel, meanwhile, remains “a niche player in the smartphone market,” with less than a 5% market share, according to research firm Statista.
“We’re obsessed with the idea that AI can make life easier and more productive for people,” the Google exec who oversees the Pixel lineup, Rick Osterloh, said at the Made by Google event. “The most important place to get this right is right in the devices we carry with us every day.”
We’ll see if consumers agree.
Here are the other doings in AI worth your attention.
Elon Musk’s Grok image generator spawns disturbing deepfakes
Elon Musk’s social media platform, X, was inundated with fake images of politicians, including former President Donald Trump and current Vice President Kamala Harris, after Grok subscribers began using a beta version of a new AI image generator called Flux. Flux, which was released as part of Grok-2 and can post images directly to X, is available to users who pay $16 a month for Grok.
“Unlike other mainstream AI photo tools, Grok, created by Musk’s artificial intelligence startup xAI, appears to have few guardrails,” CNN reported. The news site found in its own tests that it was “easily able to get Grok to generate fake, photorealistic images of politicians and political candidates that, taken out of context, could be misleading to voters.”
CNN also found that X users created deepfakes “with Grok showing prominent figures consuming drugs, cartoon characters committing violent murders and sexualized images of women in bikinis. In one post viewed nearly 400,000 times, a user shared an image created by Grok of Trump leaning out of the top of a truck, firing a rifle. CNN tests confirmed the tool is capable of creating such images.”
Other makers of AI tools are trying to stop fraudsters from using their AI image generators to create political misinformation ahead of the US elections in November and have adopted guardrails such as not allowing their tools or platforms to be used to impersonate real people or share hate speech. Some have watermarks that show an image has been AI-generated.
But none of those methods will truly be able to stop the creation of deepfakes. Last week, San Francisco City Attorney David Chiu filed a lawsuit that aims to shut down 16 websites that “undress” people. That is, the sites use AI to let people turn images of real girls and women into deepfake pornography, which Chiu called “sexual abuse.”
“While Grok claims that it will not make deepfakes, it’s quite clear that this was not the case on day one,” New York Magazine’s Intelligencer reported. “X users went wild posting some of the unseemly images that Grok will make for you.”
“Users also shared examples of Grok-2 generating images of copyrighted products and brands including the Simpsons, Nike sneakers and Tom and Jerry sharing a Coca-Cola,” Forbes reported.
One user on X praised the AI tool for being the most “uncensored model of its class yet. Glad @elonmusk is ensuring freedom of speech for humans and machines alike.” Musk responded in a post that “Grok is the most fun AI in the world!”
For the record, “freedom of speech” in the US Constitution gives citizens the right to share their opinions and ideas without interference or punishment by the government. It doesn’t give people the right to “say whatever they wish, wherever they wish” — and it definitely doesn’t support a right to give people tools that enable them to spread misinformation, deepfakes (or worse) on privately owned, for-profit social media platforms like X.
Grok-2 seems to have added some guardrails after all the hoopla, with CNN noting that xAI now won’t create images of political candidates or copyrighted cartoon characters “committing acts of violence or alongside hate speech symbols.”
Release of the Grok image editor came a week after five secretaries of state sent an open letter, on Aug. 5, to Musk, “urging him to ‘immediately implement changes’ to X’s AI chatbot Grok, after it shared with millions of users false information suggesting that Kamala Harris was not eligible to appear on the 2024 presidential ballot,” according to The Washington Post. The secretaries represent Minnesota, Pennsylvania, Michigan, Washington and New Mexico.
Artists score a win against AI generators, though war continues
Artists who are challenging four AI image generators over using copyrighted images to train AI systems without artist consent won an important battle last week when a federal judge in California said their copyright claims could move forward through the courts.
“U.S. District Judge William Orrick on Monday advanced all copyright infringement and trademark claims in a pivotal win for artists,” wrote The Hollywood Reporter, which posted a copy of Orrick’s decision. “He found that Stable Diffusion, Stability’s AI tool that can create hyperrealistic images in response to a prompt of just a few words, may have been ‘built to a significant extent on copyrighted works’ and created with the intent to ‘facilitate’ infringement. The order could entangle in the litigation any AI company that incorporated the model into its products.”
The January 2023 lawsuit was brought by concept artist Karla Ortiz, who’s worked on projects including Black Panther and who came up with the “main character design for Doctor Strange,” THR added. The suit involves the LAION data set, “which was built using 5 billion images that were allegedly scraped from the internet,” the Reporter noted.
Most AI companies facing copyright challenges, including OpenAI and Microsoft, which are being challenged by The New York Times, have said that scraping the internet for copyrighted material falls under the “fair use” doctrine of copyright law. That doctrine, which Stability AI has also pointed to, allows for some unlicensed use of copyright-protected works in certain circumstances. But it’s up to the courts to decide if the fair use argument, which Google used successfully in a suit involving Google Books in 2013, applies to AI companies slurping up thousands of articles and images for gen AI training purposes.
Former Google CEO kind of encourages using AI to steal content
Former Google CEO Eric Schmidt offered advice to would-be tech entrepreneurs during a recent talk to students at Stanford University that was taken down from the university’s YouTube channel after numerous reporters picked up on what he’d said. His comments, though, are posted in a transcript on GitHub that you can find here, and a video on X that you can find here.
Basically, Schmidt advised would-be startup founders to steal whatever copyrighted content they wanted to get their startup going, using an AI large language model (LLM), and then hire lawyers to “clean the mess up” later.
Schmidt reportedly said, according to the transcript (line 75) and the video (the 5:30 mark), “Here’s what I propose each and every one of you do. Say to your LLM the following: Make me a copy of TikTok, steal all the users, steal all the music, put my preferences in it, produce this program in the next 30 seconds, release it and in one hour, if it’s not viral, do something different along the same lines.”
Schmidt was told his comments were being recorded, and he later reportedly added the following, according to the transcript (line 341) and video (23:42): “I was not arguing that you should illegally steal everybody’s music. What you would do if you’re a Silicon Valley entrepreneur, which hopefully all of you will be, is if it took off, then you’d hire a whole bunch of lawyers to go clean the mess up, right? But if nobody uses your product, it doesn’t matter that you stole all the content. And do not quote me.”
A longtime Silicon Valley observer pointed out to me that Schmidt simply said the quiet part out loud.
In the same interview, Schmidt also said the reason Google is behind OpenAI in the AI race is because Google employees are mostly working remotely and that the company decided that having “work-life balance … was more important than winning.” You can hear those comments in this video excerpt on X.
Schmidt later sent a statement to TechCrunch apologizing for his comments about Google, saying, “I misspoke about Google and their work hours. I regret my error.”
He didn’t respond to my request for further comment on the Google remarks, or to my request for clarification of his reported comments about stealing.
The famous voices most likely to be co-opted by AI
Podcastle, which offers AI-powered audio editing tools, said it asked a representative sample of 1,000 people in the US to identify which celebrity voices they think are most likely to be cloned by an AI.
Topping the list are Arnold Schwarzenegger, Donald Trump, Kim Kardashian, Sylvester Stallone and Christopher Walken, due to their distinctive voices. Rounding out the top 10 are Matthew McConaughey, Morgan Freeman, Taylor Swift, Mike Tyson and Denzel Washington.
Podcastle also asked in the survey what celebrity voices people would most like to hear as their AI assistant. No. 1 on the list is Morgan Freeman. “Almost a quarter (24%) of Americans said yes to Freeman as their phone voice assistant; over one in ten (11%) want to hear him as the next A-list podcast host,” Podcastle found. Those surveyed said he has a voice that’s “pleasant” and “easy to listen to.”
Freeman, who’s narrated films including The Shawshank Redemption, knows that folks like his voice, and he’s not OK with people copying it. In July, he went after a popular TikTok influencer for using an AI version of his distinctive voice without his permission to narrate a fake day in her life.
When it comes to deciphering ancient texts, AI steps up
A new AI project called Fragmentarium is using machine learning to sort through digitized tablet fragments to help Assyriologists — “experts in the study of cuneiform and the cultures that used it” — create a complete version of an ancient poem known as the Epic of Gilgamesh, The New York Times noted.
“Fragments of the epic, which was written more than 3,000 years ago and was based upon still earlier works, have re-emerged as tablets that have been unearthed in archaeological digs, found in museum storerooms or surfaced on the black market,” the paper said. The lack of cuneiform experts, coupled with the fact that there may be as many as 500,000 clay tablets with bits of the poem in collections at museums and universities around the world, means many of the fragments haven’t even been read or published, the NYT added.
The Fragmentarium team is using AI to help researchers fill in missing words and lines faster than might have been even possible by humans, according to the team at the Institute for Assyriology at Ludwig Maximilian University in Germany, Artnet said.
“This isn’t the first time that AI has helped out with deciphering ancient manuscripts,” Artnet added. “A team at Tel Aviv University and Ariel University in Israel unveiled a similar project working with Mesopotamian cuneiform in 2023. And an AI system from University of Kentucky researchers is helping unveil the text on a cache of scrolls burned during the eruption of Mount Vesuvius.”
Cool, right?
