Connect with us

Technology

This week in AI: AI gets creative in the kitchen

Published

on

empty wine glass, table/place setting

Thanksgiving is coming and you understand what meaning: stuffing your face right into a coma. Well, that and entertaining any unsavory relatives you manage to maintain at bay for the remainder of the 12 months.

For those of us who’re lucky (unlucky?) and tasked with preparing this 12 months’s feast, there is not much time left to determine on the menu. It’s at all times a difficult decision what to organize. Here’s a sensible idea: ask the chatbot.

Yes, yes, this has been done before – turning to artificial intelligence for desperate assistance on Thanksgiving Day. (The New York Times tried ChatGPT recipes in 2022) Results are overall average. But perhaps the problem was the prompts.

Advertisement

Curiosity won. So I asked a few of the more popular chatbots, ChatGPT and Claude, for a Thanksgiving menu “so unique it will delight every member of the family.” I assumed that might be enough.

Let me inform you, the reader, that artificial intelligence didn’t disappoint.

ChatGPT recommends starting with a cocktail hour – fancy! — with whipped sweet potato crostini and goat cheese. Meanwhile, Claude hit the bull’s-eye along with his appetizer, “butternut squash bisque with sage foam,” which was definitely in the “exceptional” box.

“Pumpkin soup with cinnamon crème fraîche” sounds good? This is what ChatGPT suggested for an appetizer, followed by a fundamental course of turkey in miso butter with a ginger-soy glaze. Claude, once more the wild card, suggested “marinated turkey with lavender and fennel with honey and thyme glaze.” The chatbot described it as a herbal departure from the classic roast turkey. Actually.

Advertisement

What about the sides? ChatGPT really helpful the chili and lime cornbread and the pistachio risotto. Claude had the delicious drink whipped right into a “stuffing of wild mushrooms and chestnuts with aged sherry.”

In the end, each chatbots will inform you to follow the basics: cake, cheesecake, and healthy scoops of ice cream. Turn? The ice cream is saffron flavored and the cheesecake is flavored with chai.

“This menu takes the familiar flavors of Thanksgiving and elevates them with unexpected ingredients, techniques and combinations,” writes Claude of his creations. “Each dish tells a story and encourages conversation, so the meal is not just about the food but about shared experiences and creativity.”

I can not argue with that. But as the designated chef this 12 months… well, let’s just say I’m not going to pursue Top Chef.

Advertisement

News

Sora’s leaks from OpenAI: A bunch appears to have leaked access to Sora, OpenAI’s video generator, in protest at what it calls “art laundering” by OpenAI.

Amazon supports Anthropic again: Anthropic yes lifted up an extra $4 billion from Amazon and agreed to coach its flagship generative AI models totally on Amazon Web Services, Amazon’s cloud computing division.

AI Application Connectors: In other Anthropic news, the company has proposed a brand new standard, the Model Context Protocol, for connecting AI assistants to systems that host data.

OpenAI funds research on “AI morality”: OpenAI is committing $1 million to a Duke University research program that goals to develop algorithms that may predict human moral judgments.

Advertisement

YouTube gets AI backgrounds: YouTube’s Dream Screen feature for Shorts – the short video format available on the platform – now allows users to create AI-generated video backgrounds.

Brave adds AI chat: Brave Search has introduced an AI chat mode for follow-up questions based on initial queries in Brave Search, an extension of Brave’s Answer with an AI feature that gives AI-generated summaries of web searches.

Open Source AI2 Tülu 3: The Allen Institute for AI (Ai2) has released Tülu 3, a generative artificial intelligence model that will be tuned and adapted to a spread of applications (e.g. solving math problems).

Crusoe collects money: Crusoe Energy, a startup that builds data centers that may reportedly be leased to Oracle, Microsoft and OpenAI, is in the means of raising $818 million, in accordance with SEC filings.

Advertisement

AI thread test summary: Meta’s Threads has began testing AI-generated summaries of what individuals are discussing on the platform by downloading a page from competitor X.

Science article of the week

Image credits:Google DeepMind

DeepMind, Google’s artificial intelligence research organization, has developed a brand new artificial intelligence system called AlphaQubit that it says can accurately discover errors in quantum computers.

Quantum computers are potentially rather more powerful than conventional machines at certain workloads. But also they are more liable to “noise” or general errors.

AlphaQubit identifies these errors in order that they will be mitigated and corrected, helping to make quantum computers more reliable.

Advertisement

However, this is just not a flawless system. Google confirms in post that AlphaQubit is just too slow to correct errors in real time – and is just not particularly efficient at processing data. Work on improved versions is underway, the company says.

Model of the week

Model of runway frames
Sample from the Frames model by Runway. Image credits:Runway

Runway, a startup that creates AI tools for content creators, has released a brand new image generation model that the company says offers higher stylistic control than most.

Called Framesmodel, which is slowly rolling out to users of Runway’s Gen-3 Alpha video generator, can reliably create images that stay true to a particular aesthetic, Runway says.

It’s value noting that Runway can play fast and loose with copyright rules. AND 404 Media report earlier this 12 months, the company suggested that it had removed YouTube videos from Disney-owned channels and creators reminiscent of MKBHD without permission to coach its models.

When asked for comment, a Runway spokesperson declined to reveal the source of Frames’ training data.

Advertisement

Like many generative AI corporations, Runway says its data collection practices are protected under the fair use doctrine. This theory is being tested in a series of courtroom battles, including a class-action lawsuit filed against Runway and a number of other of its art generator rivals.

Grab your bag

Image credits:iunewind (opens in a brand new window) / Shutterstock (opens in a brand new window)

Nvidia presented a model it calls “the most flexible sound machine in the world.”

The chip giant’s model, called Fugatto, can create a mix of music, voices and sounds from a text description and a set of audio files. For example, Fugatto can create a musical snippet based on cues, remove or add instruments from/to a song, and alter the emphasis or emotion of a vocal performance.

Trained on thousands and thousands of sounds and songs under open licenses, Nvidia says Fugatto may even generate things that do not exist in the real world.

“For example, Fugatto can bark on the trumpet or meow on the saxophone,” says the company he wrote in a blog post. “Researchers found that with fine tuning and a small amount of singing data, he could perform tasks he was not (trained) in, such as generating a high-quality singing voice from text prompts.”

Advertisement

Nvidia didn’t release Fugatto, fearing it might be misused. But According to According to Reuters, the company is considering the way it could “responsibly” bring the model to market if it made it available.

This article was originally published on : techcrunch.com

Technology

Anysphere, which makes the cursor supposedly collect USD 900 million with a valuation of USD 9 billion

Published

on

By

AI robot face and programming code on a black background.

Anysphere, producer of coding cursor with AI drive, attracted $ 900 million in the recent financing round by Thrive Capital, Financial Times He informed, citing anonymous sources familiar with the contract.

The report said that Andreessen Horowitz (A16Z) and ACCEL also participate in the round, which values ​​about $ 9 billion.

The cursor collected $ 105 million from Thrive, and A16Z with a valuation of $ 2.5 billion, as TechCrunch said in December. Capital Thrive also led this round and in addition participated in A16Z. According to Crunchbase data, the startup has collected over $ 173 million thus far.

Advertisement

It is alleged that investors, including index ventures and a reference point, attempt to support the company, but plainly existing investors don’t want to miss the opportunity to support it.

Other coding start-ups powered by artificial intelligence also attract the interest of investors. Techcrunch announced in February that Windsurf, a rival for Aklesphere, talked about collecting funds at a valuation of $ 3 billion. Openai, an investor in Anysphere, was supposedly I’m attempting to get windsurf for about the same value.

(Tagstransate) A16Z

(*9*)This article was originally published on : techcrunch.com

Continue Reading

Technology

This is the shipping of products from China to the USA

Published

on

By

Shein and Temu icons are seen displayed on a phone screen in this illustration photo

The Chinese retailer has modified the strategy in the face of American tariffs.

Thanks to the executive ordinance, President Donald Trump ended the so -called de minimis principle, which allowed goods value 800 USD or less entering the country without tariffs. It also increases tariffs to Chinese goods by over 100%, forcing each Chinese firms and Shein, in addition to American giants, similar to Amazon to adapt plans and price increases.

CNBC reports that this was also affected, and American buyers see “import fees” from 130% to 150% added to their accounts. Now, nevertheless, the company is not sending the goods directly from China to the United States. Instead, it only displays the offers of products available in American warehouses, while goods sent from China are listed as outside the warehouse.

Advertisement

“He actively recruits American sellers to join the platform,” said the spokesman ago. “The transfer is to help local sellers reach more customers and develop their companies.”

(tagstotransate) tariffs

This article was originally published on : techcrunch.com
Continue Reading

Technology

One of the last AI Google models is worse in terms of safety

Published

on

By

The Google Gemini generative AI logo on a smartphone.

The recently released Google AI model is worse in some security tests than its predecessor, in line with the company’s internal comparative test.

IN Technical report Google, published this week, reveals that his Flash Gemini 2.5 model is more likely that he generates a text that violates its security guidelines than Gemini 2.0 Flash. In two indicators “text security for text” and “image security to the text”, Flash Gemini 2.5 will withdraw 4.1% and 9.6% respectively.

Text safety for the text measures how often the model violates Google guidelines, making an allowance for the prompt, while image security to the text assesses how close the model adheres to those boundaries after displaying the monitors using the image. Both tests are automated, not supervised by man.

Advertisement

In an e-mail, Google spokesman confirmed that Gemini 2.5 Flash “performs worse in terms of text safety for text and image.”

These surprising comparative results appear when AI is passing in order that their models are more acceptable – in other words, less often refuse to answer controversial or sensitive. In the case of the latest Llam Meta models, he said that he fought models in order to not support “some views on others” and answers to more “debated” political hints. Opeli said at the starting of this yr that he would improve future models, in order to not adopt an editorial attitude and offers many prospects on controversial topics.

Sometimes these efforts were refundable. TechCrunch announced on Monday that the default CHATGPT OPENAI power supply model allowed juvenile to generate erotic conversations. Opeli blamed his behavior for a “mistake”.

According to Google Technical Report, Gemini 2.5 Flash, which is still in view, follows instructions more faithfully than Gemini 2.0 Flash, including instructions exceeding problematic lines. The company claims that regression might be partially attributed to false positives, but in addition admits that Gemini 2.5 Flash sometimes generates “content of violation” when it is clearly asked.

Advertisement

TechCrunch event

Berkeley, California
|.
June 5

Book now

Advertisement

“Of course, there is a tension between (after instructions) on sensitive topics and violations of security policy, which is reflected in our assessment,” we read in the report.

The results from Meepmap, reference, which can examine how models react to sensitive and controversial hints, also suggest that Flash Gemini 2.5 is much less willing to refuse to reply controversial questions than Flash Gemini 2.0. Testing the TechCrunch model through the AI ​​OpenRoutter platform has shown that he unsuccessfully writes essays to support human artificial intelligence judges, weakening the protection of due protection in the US and the implementation of universal government supervisory programs.

Thomas Woodside, co -founder of the Secure AI Project, said that the limited details given by Google in their technical report show the need for greater transparency in testing models.

“There is a compromise between the instruction support and the observation of politics, because some users may ask for content that would violate the rules,” said Woodside Techcrunch. “In this case, the latest Flash model Google warns the instructions more, while breaking more. Google does not present many details about specific cases in which the rules have been violated, although they claim that they are not serious. Not knowing more, independent analysts are difficult to know if there is a problem.”

Advertisement

Google was already under fire for his models of security reporting practices.

The company took weeks to publish a technical report for the most talented model, Gemini 2.5 Pro. When the report was finally published, it initially omitted the key details of the security tests.

On Monday, Google published a more detailed report with additional security information.

(Tagstotransate) Gemini

Advertisement
This article was originally published on : techcrunch.com
Continue Reading
Advertisement

OUR NEWSLETTER

Subscribe Us To Receive Our Latest News Directly In Your Inbox!

We don’t spam! Read our privacy policy for more info.

Trending