French AI Startup ZML Launches Free Software to Cut AI Inference Costs

·4 views

French AI startup ZML has released a new free product called ZML/LLMD, designed to speed up AI inference across a wide range of chips. The software aims to make running AI models less expensive and more efficient. ZML has gained notable attention in the AI industry, having received an endorsement from Yann LeCun, a Turing Award-winning AI pioneer. The release positions ZML as a competitive player in the growing market for AI infrastructure optimization tools.

Read the full story at TechCrunch

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Samsung Galaxy Unpacked Set for July 22 With Wider Foldable Expected

Samsung has officially announced its next Galaxy Unpacked event for July 22, teasing a new form factor with the tagline 'A new shape unfolds.' The company is widely expected to unveil a shorter, wider book-style foldable phone, expanding its foldable lineup to three formats. The move appears aimed at competing with Huawei's Pura X Max and Apple's anticipated foldable iPhone. Samsung is also expected to reveal updated versions of its existing Galaxy Flip and Fold devices at the same event, with the Fold potentially receiving a rebrand.

0 comments Read more at The Verge

TechnologyTechCrunch ·

SambaNova Raises $1B at $11B Valuation Amid Earlier Intel Acquisition Rumors

AI chip maker SambaNova has secured a $1 billion funding round, valuing the company at $11 billion. The raise comes just five months after the startup completed its previous large-scale funding round. The latest investment marks a dramatic increase in the company's perceived value. This development is notable given that Intel had reportedly been in talks to acquire SambaNova for approximately $1.6 billion not long ago.

0 comments Read more at TechCrunch

TechnologyArs Technica ·

HalluSquatting Flaw in 9 Major AI Tools Could Let Hackers Build Botnets

A technique called HalluSquatting exploits a fundamental weakness in large language models — their tendency to fabricate answers rather than admit uncertainty. Researchers found that nine widely used AI tools are vulnerable to this method, which attackers could leverage to assemble large-scale botnets. The flaw stems from LLMs confidently generating fictitious package names or URLs, which hackers can register in advance to trap unsuspecting users or systems. This form of attack turns AI hallucination from an inconvenience into an active cybersecurity threat.

0 comments Read more at Ars Technica

TechnologyThe Verge ·

Meta to disable smart glasses camera if privacy LED is tampered with

Meta has announced a software update for its smart glasses that will automatically disable the camera if the device detects tampering with its privacy LED light. The move comes in response to public backlash and reports of users physically drilling out or otherwise destroying the indicator light to record covertly. Earlier measures, introduced with the second-generation glasses, only prompted users to uncover a blocked LED but did not prevent recording. Modders had reportedly found workarounds to those softer restrictions, prompting Meta to take a more forceful technical approach. The new feature is intended to close loopholes that allowed covert recording without any visible indicator to bystanders.

0 comments Read more at The Verge