Most people have already concluded that Gemini only gives you a handful of queries per day for free, but, until now, Google wouldn’t say precisely how many.
Now it’s official, and on Google’s website.
Continue reading “Finally: Google reveals how many queries you get on Gemini”
Confidently wrong answers by large language models have been plaguing both users and labs since AIs inception, but a new study from OpenAI seeks to find a solution.
LLMs are trained on finding the next word in huge datasets, they say, focusing solely on finding the correct word in a sequence rather than looking for accuracy.
Continue reading “OpenAI paper: Why hallucinations happen, and what can be done to fix them”
The epic class action lawsuit concerned a library of 7 million pirated books used in training, and had Anthropic looking at $150,000 in penalties per instance of copyright theft, but it was settled last week without disclosing terms.
Continue reading “Anthropic’s copyright settlement to cost $1.5 billion or more”
OpenAI will make custom chips with Broadcom
With Nvidia lurking in the background, more companies are working on their custom AI chips. Now OpenAI has entered the fray, said to produce their own chips with Broadcom next year. It will be for internal use, and won’t be released broadly. They have a long history with this, having first entered talks with TSMC last year. Broadcom said on its earnings call this Thursday that it had secured a $10B order for AI chips without naming from whom, and now the Financial Times is reporting that it is, indeed, OpenAI, who has no comment on this.
More at: Financial Times (Paywalled) and Reuters.
Amazon lens lets you shop for anything you can see
The latest feature in the Amazon Shopping app on iOS lets you simply point your camera on anything you like, and shop for the same or similar items in real-time. It partners with Amazon’s AI shopping assistant, Rufus, to also answer questions about the products in the shop. It should «roll out to more customers in the coming weeks,» meaning there’s likely an Android version in the works.
More at: Amazon’s product page, and The Verge.
Read on for more News!
Continue reading “Friday roundup: Unis hiring AI officers, OpenAI on jobs and Nano Banana”
Rumor is that Apple will use an underlying, custom model from Google for the brains of the assistant, and Apple has been looking at it for quite a while.
Continue reading “Report: Apple planning an AI search engine for Siri as early as March 2026”
The company has just finished a «Series F» fundraising round of $13 billion at a valuation of $183 billion, which will «expand our capacity to meet growing enterprise demand, deepen our safety research, and support international expansion as we continue building reliable, interpretable, and steerable AI systems,» Anthropic writes.
Continue reading “Anthropic is now a 183 billion dollar company”
This includes alerting parents of teens in distress, routing queries to a more powerful reasoning model when appropriate, and giving parents more control over their kid’s usage.
The company has assembled a council of experts in «youth development, mental health, and human-computer interaction,» which will shape how AI will «support people’s well-being,» they say in a blog post.
Overall usage of generative ai has reached a critical point, with almost every student using it for coursework. But what they use it for is equally interesting.
Continue reading “85% of US college students say they use generative AI, 25% use it to cheat”
India is the second largest market for ChatGPT in the world, after the USA, and this marks their first foray into data centers in the country, which usually shuns products not produced locally.
Continue reading “OpenAI set to build India data center of «at least» one gigawatt”
They are anonymized in the filing, but Nvidia clearly states that they account for 23%, 16% and 14% of sales.
Last quarter, two top customers accounted for 14% and 11% of revenue, writes CNBC.
OEM manufacturers?
These customers are not end users of the product, but rather systems manufacturers who buy chips directly from Nvidia and put them on circuit boards and servers — which they call «Direct customers.»
Continue reading “Just three customers accounted for half of Nvidia’s chip sales last quarter”
Most of the bots were created by users, but some high profile ones, garnering more than 10 million views, came from a Meta employee.
Meta admits fault
This is a failure of Meta’s policy enforcement, spokesman Andy Stone tells Reuters — and says there are guidelines in place against intimate celebrity pictures.
Meta has since deleted many of the bots in question, but some still remain, Reuters found.
While there are individual state laws protecting the likeness and appearance of celebrities, Duncan Crabtree-Ireland of SAG-AFTRA — the actors union — says they are pushing for federal laws against AI imitation.
Go read the full story on Reuters. See also: Grok undresses Taylor Swift.
The agentic Realtime model is a native speech-to-speech model that can be used to make customer service agents, phone reps and voice navigation features. It doesn’t go through speech-to-text and text-to-speech loops and generates audio «directly through a single model and API.» OpenAI is marketing this to developers who want more natural flowing speech, and it’s not available as distinct model in ChatGPT – yet. You can hear it and see it in use at places like Zillow, T-mobile, StubHub and Oscar Health, though. With general availability, it will surely show up in a lot more places soon.
More at: OpenAI’s launch page, discussion on r/OpenAI.
Read on for more news!
Continue reading “Friday roundup: A good week for coding, speech models”
This applies to all private users, and if you opt in, your chats will be used not only for training future models, but to improve the safety of the current ones:
— We’re now giving users the choice to allow their data to be used to improve Claude and strengthen our safeguards against harmful usage like scams and abuse, Anthropic writes.
Continue reading “Anthropic starts training models on new Claude chats”
Claude for Chrome is «the next logical step,» writes Anthropic, after connecting Claude to calendars, documents and emails.
— We view browser-using AI as inevitable, they go on, and cite the large portion of work being done in the browser interface.
Continue reading “Anthropic slips out Claude for Chrome, warns against prompt injection hacks”
The center of the suit was Anthropic’s library of 7 million pirated books in their training data, that could carry a penalty of $150,000 per infringement — and as a class action case, a loss would entail damages for every single author.
An unfavorable ruling would therefore be debilitating to Anthropic, and send dark clouds across the industry, likely forcing them toward a settlement.
Continue reading “Anthropic settles «historic» class action copyright case brought by authors”