If you’ve been dreaming about spending your summer whispering sweet nothings into the digital ears of one of the seductive ChatGPT voice assistants that OpenAI showed off last month, you’ll have to dream a little longer. On Tuesday, the company announced that its “advanced Voice Mode” feature needs more time in the oven “to reach our bar to launch.” The feature will be available to a small group of users to gather feedback, and then launch to all paying ChatGPT customers in the fall.
“We’re improving the model’s ability to detect and refuse certain content,” OpenAI posted on X. “We’re also working on improving the user experience and preparing our infrastructure to scale to millions while maintaining real-time responses.”
We're sharing an update on the advanced Voice Mode we demoed during our Spring Update, which we remain very excited about:
We had planned to start rolling this out in alpha to a small group of ChatGPT Plus users in late June, but need one more month to reach our bar to launch.…
Voices have been a part of ChatGPT since 2023. But last month, OpenAI demoed an upgraded version that sounded so humanlike it drew comparisons with Samantha, the seductive voice assistant in the 2013 movie Her, played by Scarlett Johansson. Weeks after the presentation, the actress accused OpenAI of copying her voice despite denying permission.
OpenAI said it's still figuring out when the new voices (minus the Johansson soundalike) will roll out to paying users this fall. Another feature that lets the voice assistants use your phone’s camera to understand the world around you has also been delayed until that time. “Exact timelines depend on meeting our high safety and reliability bar,” the company said.
This article originally appeared on Engadget at https://www.engadget.com/openai-has-delayed-its-seductive-chatgpt-voice-assistants-233027946.html?src=rss
OpenAI plans to block people from using ChatGPT in China, a country where its services aren’t officially available, but where users and developers access it via the company’s API anyway. Securities Times, a Chinese state-owned newspaper reported on Tuesday that OpenAI had started sending emails to users in China outlining its plans to block access starting July 9, according to Reuters.
“We are taking additional taps to block API traffic from regions where we do not support access to OpenAI’s services,” an OpenAI spokesperson told the publication. The move could impact several Chinese startups which have built applications using OpenAI’s large language models.
Although OpenAI’s services are available in more than 160 countries, China isn’t one of them. According to the company’s guidelines, users trying to access the company’s products in unsupported countries could be blocked or suspended — although the company hasn’t explicitly done so until now.
It’s not clear what prompted OpenAI’s move. Last month, the company revealed that it stopped covert influence operations — including one that originated from China — that used its AI models to spread disinformation across the internet. Bloombergpointed out that OpenAI’s move coincides with Washington's pressure on American tech companies to limit China’s access to cutting-edge technologies developed in the US.
This article originally appeared on Engadget at https://www.engadget.com/openai-will-block-people-in-china-from-using-its-services-200801957.html?src=rss
Google is reportedly building new AI-powered chatbots based on celebrities and YouTube influencers. The idea isn’t groundbreaking — startups like Character.ai and companies like Meta have already launched products like this — but neither is Google’s AI strategy so far.
Google’s celebrity chatbots will be powered by the company’s Gemini family of large language models according to The Information, which broke the story on Monday. The company is trying to strike partnerships with influencers as well as celebrities, and is also working on a feature that will let people create their own chatbots simply by describing their personalities and appearance — something that Character.ai already lets you do. A fun fact: Noam Shazeer, one of the co-founders of Character.ai, is a former Google engineer and one of the creators of “transformers,” the fundamental tech that made today’s generative AI possible.
It isn’t yet clear which celebrities or influencers Google might partner with. Meta’s chatbots, for instance, are based on personalities like TikTok star Charli D’Amelio, YouTube phenomenon Mr. Beast, Snoop Dogg, Tom Brady and Paris Hilton among others, while Character.ai’s characters include politicians, philosophers, fictional characters, and even objects like a block of cheese that talks. Google’s project is reportedly being led by a longtime executive called Ryan Germick who works on Google Doodles, and a team of ten.
It also sounds like Google’s bots could be just an experiment — according to the report, the bots might only show up on Google Labs, the company’s website for experimental products, instead of being available more broadly.
It isn’t clear why Google’s doing this. Meta’s AI chatbots based on celebrities never really took off despite the company stuffing them in every product it makes. As The Information pointed out, the company’s chatbot based on Snoop Dogg has only 15,000 followers on Instagram compared with 87.5 million followers who follow the human rapper.
This article originally appeared on Engadget at https://www.engadget.com/google-is-reportedly-building-ai-chatbots-based-on-celebrities-and-influencers-235731655.html?src=rss
The US government has sued Adobe and two senior company executives for allegedly deceiving consumers by hiding early termination fees and making them jump through hoops to cancel subscriptions to Adobe products.
The complaint filed by the Department of Justice on Monday accuses the Adobe of pushing consumers towards its “annual paid monthly” subscription plan without adequately disclosing that canceling the plan within the first year could result in an early termination fee. The complaint also alleges that Adobe’s early termination fee disclosures were buried in fine print or required consumers to hover over tiny icons to find them.
“Americans are tired of companies hiding the ball during subscription signup and then putting up roadblocks when they try to cancel,” said Samuel Levine, director of the FTC’s Bureau of Consumer Protection, in a statement. “The FTC will continue working to protect Americans from these illegal business practices.”
Dana Rao, Adobe's general counsel and chief trust officer said that the company would fight the FTC in court. In a statement published on the company's website, Rao said: "Subscription services are convenient, flexible and cost effective to allow users to choose the plan that best fits their needs, timeline and budget. Our priority is to always ensure our customers have a positive experience. We are transparent with the terms and conditions of our subscription agreements and have a simple cancellation process. We will refute the FTC’s claims in court.”
The FTC said that it took action against Adobe after receiving complaints from consumers around the country who said that they were not aware of Adobe’s early termination fee. It noted that Adobe continued the practice despite being aware of consumers’ confusion. Any consumers who reached out to Adobe’s customer service to cancel their subscription encountered other obstacles like dropped calls and chats and being transferred to multiple representatives, the FTC’s statement adds.
The FTC’s action follows a wave of customer outrage over Adobe’s latest terms of service. Users were concerned that Adobe’s vague language suggested that the company could freely use their work to train its generative AI modes. In response to the backlash, Adobe announced updates to its terms of service to provide more detail around areas like AI and content ownership.
Update, June 17 2024, 1:39 PM ET: This story has been updated with a statement from Adobe.
This article originally appeared on Engadget at https://www.engadget.com/the-us-has-sued-adobe-for-early-termination-fees-and-making-subscriptions-hard-to-cancel-165808358.html?src=rss
My task for our AI overlords was simple: help me crack the New York Times Spelling Bee.
I had spent a large chunk of a Saturday evening trying to shape the letters G, Y, A, L, P, O and N into as many words as possible. But three hours, 141 points and 37 words — including “nonapology”, “lagoon” and “analogy” — later, I had hit a wall. A few more words was all I needed to propel myself into Spelling Bee’s "genius" echelon, the title reserved for those who unscramble 70 percent of all possible words using the given letters, and the point at which the puzzle considers itself, effectively, solved. (100 percent solvers are bestowed the rank of "queen bee," but even I have limits.)
My human mind was clearly struggling, but this task seemed like child’s play for AI, so I fired up ChatGPT, told it I was trying to win the Spelling Bee, gave it my letters and laid out the rules. “Give me a list of every word that can be formed with them,” I typed. “You’re allowed to repeat letters, but each word must contain 'N'.”
When I hit enter, this is what ChatGPT spat back:
Here is a list of words formed with the letters "G Y A L P O N" that must contain the letter "N": GLNPAYO YPNL PGNOYL ONAPYLG LGNAPYO ANLY NAYGOPL GYPON AOGPLNY PNAOGLY This is just a sample of the list due to the extensive number of permutations possible.
Not a great start.
“The words need to be in the dictionary,” I typed back. “Sure,” came the response. “I’ll provide a list of valid English words based on the given letters.” ANGLY, GALON, LANG.
Clearly, this wasn’t working, so I turned to Microsoft’s Copilot (YANG, PLAYING, PLANNING, ANNOYINGLY), Google’s Gemini (GAPON, GON, GIAN), and Anthropic’s Claude (MANGO, ONGOING, LAWN17.LAY). Meta AI helpfully told me that it made sure to only include words that are recognized by dictionaries in a list that contained NALYP and NAGY, while Perplexity — a chatbot with ambitions of killing Google Search — simply wrote GAL hundreds of times before freezing abruptly.
Perplexity, a chatbot with ambitions of killing Google Search, went to pieces when asked to form words from a set of letters.
Screenshot by Pranav Dixit / Engadget
AI can now create images, video and audio as fast as you can type in descriptions of what you want. It can write poetry, essays and term papers. It can also be a pale imitation of your girlfriend, your therapist and your personal assistant. And lots of people think it’s poised to automate humans out of jobs and transform the world in ways we can scarcely begin to imagine. So why does it suck so hard at solving a simple word puzzle?
The answer lies in how large language models, the underlying technology that powers our modern AI craze, function. Computer programming is traditionally logical and rules-based; you type out commands that a computer follows according to a set of instructions, and it provides a valid output. But machine learning, of which generative AI is a subset, is different.
“It’s purely statistical,” Noah Giansiracusa, a professor of mathematical and data science at Bentley University told me. “It’s really about extracting patterns from data and then pushing out new data that largely fits those patterns.”
OpenAI did not respond on record but a company spokesperson told me that this type of “feedback” helped OpenAI improve the model’s comprehension and responses to problems. "Things like word structures and anagrams aren't a common use case for Perplexity, so our model isn't optimized for it," company spokesperson Sara Platnick told me. "As a daily Wordle/Connections/Mini Crossword player, I'm excited to see how we do!" Microsoft and Meta declined to comment. Google and Anthropic did not respond by publication time.
At the heart of large language models are “transformers,” a technical breakthrough made by researchers at Google in 2017. Once you type in a prompt, a large language model breaks down words or fractions of those words into mathematical units called “tokens.” Transformers are capable of analyzing each token in the context of the larger dataset that a model is trained on to see how they’re connected to each other. Once a transformer understands these relationships, it is able to respond to your prompt by guessing the next likely token in a sequence. The Financial Times has a terrific animated explainer that breaks this all down if you’re interested.
I mistyped "sure", but Meta AI thought I was suggesting it as a word and told me I was right.
Screenshot by Pranav Dixit / Engadget
I thought I was giving the chatbots precise instructions to generate my Spelling Bee words, all they were doing was converting my words to tokens, and using transformers to spit back plausible responses. “It’s not the same as computer programming or typing a command into a DOS prompt,” said Giansiracusa. “Your words got translated to numbers and they were then processed statistically.” It seems like a purely logic-based query was the exact worst application for AI’s skills – akin to trying to turn a screw with a resource-intensive hammer.
The success of an AI model also depends on the data it’s trained on. This is why AI companies are feverishly striking deals with news publishers right now — the fresher the training data, the better the responses. Generative AI, for instance, sucks at suggesting chess moves, but is at least marginally better at the task than solving word puzzles. Giansiracusa points out that the glut of chess games available on the internet almost certainly are included in the training data for existing AI models. “I would suspect that there just are not enough annotated Spelling Bee games online for AI to train on as there are chess games,” he said.
“If your chatbot seems more confused by a word game than a cat with a Rubik’s cube, that’s because it wasn’t especially trained to play complex word games,” said Sandi Besen, an artificial intelligence researcher at Neudesic, an AI company owned by IBM. “Word games have specific rules and constraints that a model would struggle to abide by unless specifically instructed to during training, fine tuning or prompting.”
“If your chatbot seems more confused by a word game than a cat with a Rubik’s cube, that’s because it wasn’t especially trained to play complex word games."
None of this has stopped the world’s leading AI companies from marketing the technology as a panacea, often grossly exaggerating claims about its capabilities. In April, both OpenAI and Meta boasted that their new AI models would be capable of “reasoning” and “planning.” In an interview, OpenAI’s chief operating officer Brad Lightcap told the Financial Times that the next generation of GPT, the AI model that powers ChatGPT, would show progress on solving “hard problems” such as reasoning. Joelle Pineau, Meta’s vice president of AI research, told the publication that the company was “hard at work in figuring out how to get these models not just to talk, but actually to reason, to plan…to have memory.”
My repeated attempts to get GPT-4o and Llama 3 to crack the Spelling Bee failed spectacularly. When I told ChatGPT that GALON, LANG and ANGLY weren’t in the dictionary, the chatbot said that it agreed with me and suggested GALVANOPY instead. When I mistyped the world “sure” as “sur” in my response to Meta AI’s offer to come up with more words, the chatbot told me that “sur” was, indeed, another word that can be formed with the letters G, Y, A, L, P, O and N.
Clearly, we’re still a long way away from Artificial General Intelligence, the nebulous concept describing the moment when machines are capable of doing most tasks as well as or better than human beings. Some experts, like Yann LeCun, Meta’s chief AI scientist, have been outspoken about the limitations of large language models, claiming that they will never reach human-level intelligence since they don’t really use logic. At an event in London last year, LeCun said that the current generation of AI models “just do not understand how the world works. They’re not capable of planning. They’re not capable of real reasoning," he said. "We do not have completely autonomous, self-driving cars that can train themselves to drive in about 20 hours of practice, something a 17-year-old can do.”
Giansiracusa, however, strikes a more cautious tone. “We don’t really know how humans reason, right? We don’t know what intelligence actually is. I don’t know if my brain is just a big statistical calculator, kind of like a more efficient version of a large language model.”
Perhaps the key to living with generative AI without succumbing to either hype or anxiety is to simply understand its inherent limitations. “These tools are not actually designed for a lot of things that people are using them for,” said Chirag Shah, a professor of AI and machine learning at the University of Washington. He co-wrote a high-profile research paper in 2022 critiquing the use of large language models in search engines. Tech companies, thinks Shah, could do a much better job of being transparent about what AI can and can’t do before foisting it on us. That ship may have already sailed, however. Over the last few months, the world’s largest tech companies – Microsoft, Meta, Samsung, Apple, and Google – have made declarations to tightly weave AI into their products, services and operating systems.
"The bots suck because they weren’t designed for this,” Shah said of my word game conundrum. Whether they suck at all the other problems tech companies are throwing at them remains to be seen.
Update, June 13 2024, 4:19 PM ET: This story has been updated to include a statement from Perplexity.
This article originally appeared on Engadget at https://www.engadget.com/if-ai-is-going-to-take-over-the-world-why-cant-it-solve-the-spelling-bee-170034469.html?src=rss
We don’t know if OpenAI, the creator of ChatGPT, is actually making any money so far. But thanks to a Wednesday report in The Information, what we do know is that the company doubled its annualized revenue — a measure of the previous month’s revenue multiplied by 12, as the publication helpfully explained — in the last six months.
OpenAI’s annualized revenue was $3.4 billion, CEO Sam Altman reportedly told staff. That’s up from $1.6 billion around the end of last year, and $1 billion a year ago. Most of this revenue came from a subscription version of ChatGPT, which offers higher messaging limits to people who pay at least $20 a month, as well as from developers who pay the company to use the company’s large language models in their own apps and services. About $200 million on an annualized basis comes from Microsoft, which gives OpenAI a cut of sales of OpenAI’s large language models to customers using Azure, Microsoft’s cloud computing platform aimed at businesses.
Notably, an OpenAI spokesperson told The Information that the financials were "inaccurate" but did not explain which details it disputed. OpenAI did not immediately respond to Engadget's request for comment.
Earlier this week, Apple announced a partnership with OpenAI. The company plans to hook ChatGPT right into its operating systems for iPhones, iPads, and Macs, letting Siri reach out to ChatGPT to answer questions. The financial terms of that deal, however, are still unknown.
This article originally appeared on Engadget at https://www.engadget.com/openais-revenue-is-reportedly-booming-230324957.html?src=rss
Elon Musk dropped a lawsuit against OpenAI one day before a judge in California state court was set to hear OpenAI’s request for dismissal. Musk’s suit, which was filed in February, had accused OpenAI co-founders Sam Altman and Greg Brockman of violating the company’s non-profit status and instead prioritizing profits over using AI to help humanity.
In the 35-page suit, Musk had alleged that OpenAI had become a “closed-source de facto subsidiary” of Microsoft, which invested $13 billion in the company and owns a 49 percent stake. Microsoft uses OpenAI’s technology to power Copilot, the company’s generative AI tools that are deeply integrated in products like Windows and Office.
OpenAI had reportedly requested for the lawsuit to be dismissed, arguing that Musk would use any information that emerged as a result to get access to the company’s “proprietary records and technology.” The company had also said that there was no founding agreement for it to breach.
OpenAI and Musk’s lawyer, Alex Spiro, did not respond to a request for comment from Engadget.
Musk, who was one of the founders of OpenAI in 2015, left the company three years later after disagreements over the direction of the organization. He runs xAI an AI startup that makes Grok, a ChatGPT rival that is built into X and is available for paid users. xAI recently raised a $6 billion funding round from top investors including Andreessen Horowitz and Sequoia Capital.
On Monday, Musk said that he would ban Apple devices from his company’s after Apple integrated ChatGPT in its operating systems through a partnership with OpenAI.
This article originally appeared on Engadget at https://www.engadget.com/musk-withdraws-his-breach-of-contract-lawsuit-against-openai-221316519.html?src=rss
Apple is integrating GPT-4o, the large language model that powers ChatGPT into iOS 18, iPadOS 18 and MacOS Sequioa thanks to a partnership with OpenAI announced at WWDC, the company’s annual developer conference, on Monday. But shortly after the keynote ended, Craig Federighi, Apple’s senior vice president of software engineering said that the company might also bake in Gemini, Google’s family of large language model, into its operating systems.
“We want to enable users ultimately to choose the models they want, maybe Google Gemini in the future,” Federighi said in a conversation with YouTuber iJustine after the keynote. “Nothing to announce right now.”
The news is notable because even though Apple did mention plans to add more AI models into its operating system in the keynote, it didn’t mention Gemini specifically. Letting people choose the AI model they want on their devices instead of simply foisting one on them would give Apple devices a level of customization that none of its competitors like Google or Samsung have.
This article originally appeared on Engadget at https://www.engadget.com/apple-may-integrate-googles-gemini-ai-into-ios-in-the-future-220240081.html?src=rss
Apple is going all in on AI in the most Apple way possible. At WWDC, Apple's annual conference for developers, the company revealed Apple Intelligence, an Apple-branded version of AI that is more focused on infusing its software with the technology and upgrading existing apps to make them more useful. Apple Intelligence will be powered both by Apple’s homegrown tech as well as a partnership with OpenAI, the maker of ChatGPT, Apple announced.
One of Apple’s biggest AI upgrades is coming to Siri. The company’s built-in voice assistant will now be powered by large language models, the tech that underlies all modern-day generative AI. Siri, which has languished over the years, may become more useful now that it can interact more closely with Apple’s operating systems and apps. You can, for instance, ask Siri to give you a summary of a news article, delete an email or edit a photo. The assistant will also be able to take more than 100 actions, such as finding photos based on a general description of their contents, or extracting personal information from a picture of your ID to automatically fill in forms online. Finally, you can type your question into Siri instead of using your voice.
Apple Intelligence will be highlight relevant content in Safari as you browse. You’ll also be able to use it to quickly catch up on priority notifications. And just like Gmail and Outlook, your devices will be able create fleshed out responses to emails and text messages on your behalf. Apple also announced a suite of new features called Writing Tools that uses AI to write, rewrite, proofread and summarize text across the system, useful to draft emails and blog posts, for instance.
Apple
Apple Intelligence will use AI to record, transcribe and summarize your phone calls, rivaling third-party transcription services like Otter. All participants are automatically notified when you start recording, and a transcript of the conversation's main points is automatically generated at the end. You can also use AI to generate images, stickers and custom emoji (which Apple calls Genmoji) in any app.
Apple
Thanks to its partnership with OpenAI, Apple also is baking the base version of GPT-4o — OpenAI's newest large language model — into Siri as well as Writing Tools. Siri can act as an intermediary for user queries to GTP-4o, and Writing Tools can use the LLM to help compose text. Apple claims unless you connect your paid ChatGPT account to your Apple device, the company won't store your requests or other identifying information like your IP address.
Apple Intelligence, which the company says will be in beta at launch, will be restricted to the iPhone 15 Pro and Pro Max and iPads and Macs with M1 (or higher) chipsets. Your device will also need to be set to US English.
Apple's AI features are a long time coming. Generative AI has shaken up Silicon Valley ever since OpenAI launched ChatGPT around the end of 2022. Since then, Apple’s rivals like Google, Samsung and Microsoft, as well as companies like Meta have raced to integrate AI features in all their primary products. Last month, Google announced that AI would be a cornerstone of the next version of Android and made major AI-powered changes to its search engine. Samsung, Apple’s primary smartphone competitor, added AI features to its phones earlier this year that can translate calls in real time and edit photos. Microsoft, too, unveiled AI-powered Copilot PCs, aimed at infusing Windows with AI features that include live captioning, image editing, and beefing up systemwide search.
Apple plans to build a password management app right into the next versions of iPhone and Mac operating systems, reportedBloomberg’s Mark Gurman on Thursday. The new app, simply called Passwords, will compete against existing password managers like 1Password and LastPass, which typically charge people a monthly fee for generating and storing unique passwords. Apple plans to reveal the app at the company’s annual Worldwide Developers Conference on June 10.
Apple already generates and stores unique passwords through iCloud Keychain, a feature that syncs passwords across all Apple devices you own as well as Windows PCs through a browser extension. But passwords stored in iCloud Keychain live — weirdly — in the Settings app, often making them cumbersome to find or change. Having a dedicated app for passwords built into Apple devices would not only make this easier but also give people one more reason to stay in the Apple ecosystem.
Just like its rivals, Apple’s Passwords app will reportedly split passwords into different categories like accounts, WiFi networks, and Passkeys (here’s our deep dive explaining how they work). It will also allow you to import passwords from rival apps and will fill them in automatically when your device detects you’re logging into a website or an app. Passwords will also work on Apple’s $4,000 Vision Pro headset, and, just like Google Authenticator and Authy, will support two-factor verification codes. What is still unclear is whether the Passwords app will let you securely store files and images in addition to passwords, something that both 1Password and LastPass offer.
In addition to Passwords, Apple is expected to reveal the next versions of iOS, iPadOS, MacOS, WatchOS and VisionOS on Monday. The new versions of the software will reportedly be infused with brand new AI features.
This article originally appeared on Engadget at https://www.engadget.com/apple-will-reportedly-build-a-dedicated-passwords-app-for-the-iphone-and-mac-211812245.html?src=rss