News about Android is being relegated to a side show at Google’s annual developer conference, Google I/O, next month. On Monday, the company announced it would share information about the latest updates coming to the Android ecosystem at an upcoming virtual event called “The Android Show: I/O Edition,” airing on May 13. The decision to […]
Categoria: AI
Auto Added by WPeMatico
-
Final 7 days: Save $210 + 50% off a second ticket to TechCrunch Sessions: AI
The hub of AI innovation is here — don’t miss out on the last 7 days of Early Bird ticket savings! AI is reshaping the world, and TechCrunch Sessions: AI is where you’ll witness the future unfold. Join 1,000+ founders, investors, researchers, and innovators on June 5 at UC Berkeley’s Zellerbach Hall for a full […] -
Apple’s complicated plan to improve its AI while protecting privacy
Apple says it’s found a way to make its AI models better without training on its users’ data or even copying it from their iPhones and Macs. In a blog post first reported on by Bloomberg, the company outlined its plans to have devices compare a synthetic dataset to samples of recent emails or messages from users who have opted into its Device Analytics program.
Apple devices will be able to determine which synthetic inputs are closest to real samples, which they will relay to the company by sending “only a signal indicating which of the variants is closest to the sampled data.” That way, according to Apple, it doesn’t access user data, and the data never leaves the device. Apple will then use the most frequently picked fake samples to improve its AI text outputs, such as email summaries.
Currently, Apple trains its AI models on synthetic data only, potentially resulting in less helpful responses, according to Bloomberg’s Mark Gurman. Apple has struggled with the launch of its flagship Apple Intelligence features, as it pushed back the launch of some capabilities and replaced the head of its Siri team.
But now, Apple is trying to turn things around by introducing its new AI training system in a beta version of iOS and iPadOS 18.5 and macOS 15.5, according to Gurman.
Apple has been talking up its use of a method called differential privacy to keep user data private since at least 2016 with the launch of iOS 10 and has already used it to improve the AI-powered Genmoji feature. This also applies to the company’s new AI training plans as well, as Apple says that introducing randomized information into a broader dataset will help prevent it from linking data to any one person.
-
RLWRLD raises $14.8M to build a foundational model for robotics
As robotics has advanced, industry has steadily adopted more robots to automate away many kinds of grunt work. More than 540,000 new industrial robots were installed worldwide in 2023, taking the number of total industrial robots active to above 4 million, per IFR. Industrial robots typically excel at repetitive tasks, but they find it challenging […] -
Debates over AI benchmarking have reached Pokémon
Not even Pokémon is safe from AI benchmarking controversy. Last week, a post on X went viral, claiming that Google’s latest Gemini model surpassed Anthropic’s flagship Claude model in the original Pokémon video game trilogy. Reportedly, Gemini had reached Lavender Town in a developer’s Twitch stream; Claude was stuck at Mount Moon as of late […] -
Nvidia says it plans to manufacture some AI chips in the U.S.
Nvidia says it has commissioned more than a million square feet of manufacturing space to build and test AI chips in Arizona and Texas as part of an effort to move a portion of production to the U.S. The chipmaker says that Nvidia Blackwell chips have started production at TSMC’s chip plants in Phoenix, Arizona, […] -
ChatGPT will now remember your old conversations
OpenAI is giving ChatGPT a memory upgrade that allows it to recall old conversations that you didn’t ask it to save. OpenAI CEO Sam Altman said on X that the chatbot can “now reference all your past conversations,” and that the update aligns with the company’s goal to develop “AI systems that get to know you over your life.”
This builds on the “Memory” feature that was added to ChatGPT last year, which allowed limited information like queries, prompts, and customizations to be retained and used for future responses. With the long-term memory update, ChatGPT will now recall information in two ways — using the “saved memories” that users have manually asked it to remember, and “reference chat history,” which are “insights ChatGPT gathers from past chats to improve future ones,” according to OpenAI.
The update will be available everywhere except in the EU, UK, Switzerland, Norway, Iceland, and Liechtenstein, likely due to these regions having tight AI regulations that Altman has objected to in the past. It’s currently being rolled out to users paying for ChatGPT’s $200 monthly Pro subscription and will be available “soon” for $20 Plus subscribers, according to Altman. OpenAI also says it will be available to Team, Enterprise, and Edu users “in a few weeks,” but there’s no word on when — or if — it will roll out to free users.
Memory is an optional feature for ChatGPT. Users who don’t want the chatbot to save any conversations can toggle off saved memories under the ChatGPT personalization settings, or use the temporary chat function to ask it inquiries that won’t use or affect memory. ChatGPT’s memory upgrade follows a similar update that Google made to Gemini AI in February that allows it to recall older conversations to provide more personalized or relevant responses.
-
Judge berates AI entrepreneur for using a generated ‘lawyer’ in court
A man’s recent attempt to use an AI-generated avatar in his legal appeal made an immediate impression on a New York courtroom, but probably not the one he was hoping for.
Jerome Dewald — a 74-year-old that The Register notes is behind a startup that says it’s “revolutionizing legal self-representation with AI” — was chewed out during an employment dispute hearing on March 26th for failing to inform judges that he had artificially generated the man presenting his oral argument. While the court had approved Dewald to submit a video for his case, Justice Sallie Manzanet-Daniels became confused when the unknown speaker, who clearly wasn’t Dewald, appeared on the screen.
“Hold on,” Manzanet-Daniels said, interrupting the video after the avatar had barely finished its first sentence. “Is that counsel for the case?”
“I generated that,” Dewald responded. “It’s not a real person.”
Dewald told The Register that the avatar — a “big, beautiful hunk of a guy” called Jim — was one of the stock options provided by an AI avatar company called Tavus. Dewald says the video was submitted due to difficulties he experiences with extended speaking, but the courtroom was unaware that the video contents were artificially generated.
“It would have been nice to know that when you made your application. You did not tell me that, sir, I don’t appreciate being misled.” said Manzanet-Daniels, responding to Dewald’s admission. “You are not going to use this courtroom as a launch for your business.”
This is the latest of several snafus that have occurred when people try to mix legal processes with AI technology. Two attorneys and a law firm were penalized in 2023 for submitting fictitious legal research that had been made up by ChatGPT. DoNotPay, a “robot lawyer” company, was also ordered to pay the FTC a $193,000 settlement in February for advertising, without evidence, that its AI legal representation is as good as a real human lawyer.
-
OpenAI countersues Elon Musk to stop his attacks and ‘fake takeover bid’
OpenAI filed a countersuit against Elon Musk on Wednesday, saying on X that “Elon’s nonstop actions against us are just bad-faith tactics to slow down OpenAI and seize control of the leading AI innovations for his personal benefit.”
In the lawsuit, OpenAI’s lawyers argue that “Musk’s continued attacks on OpenAI, culminating most recently in the fake takeover bid designed to disrupt OpenAI’s future, must cease. Musk should be enjoined from further unlawful and unfair action, and held responsible for the damage he has already caused.”
Musk, who was part of the initial founding team at OpenAI, initially sued last spring, saying he wanted to force the company to “return to its mission to develop AGI for the benefit of humanity” instead of pursuing profits. (The Verge’s editor-in-chief, Nilay Patel, found Musk’s legal case against OpenAI “hilariously bad.”)
Musk dropped that lawsuit in June but sued OpenAI again in August. In December, OpenAI published a blog post with the headline “Elon Musk wanted an OpenAI for-profit,” with receipts. The case is scheduled to go to trial in the spring of 2026.
Earlier this year, Musk also offered $97.4 billion to buy OpenAI, saying in a statement that “it’s time for OpenAI to return to the open-source, safety-focused force for good it once was.” OpenAI’s board of directors unanimously rejected the offer, which today’s filing called a “sham bid.”
Disclosure: The Verge’s parent company, Vox Media, partners with OpenAI.
-
AI insurtech Ominimo bags its first investment at a $220M valuation
How do you get talented engineers to work for a startup in a mundane field at a time when more exciting companies are paying well and hiring aggressively? Here’s an answer from one insurance startup out of Poland called Ominimo: make pay competitive, but more importantly, give those engineers the license to apply their talent […]