AI applied sciences, which have turn into more and more widespread in day by day life not too long ago, proceed to develop quickly. Nearly day by day brings a outstanding breakthrough on this discipline, and important milestones are being crossed one after the other. In our “AI Diaries” collection, we proceed to usually report the evolution of this technological revolution, which has the potential to considerably alter life on Earth within the coming interval.
What Occurred within the AI World This Week?
As all the time, new AI instruments have entered our lives this week, alongside main updates to widespread fashions like GPT and Gemini. Listed here are the highlights of the week within the AI world:
Gemini Will get Smarter with 3.1 Professional and Beneficial properties Music Technology by way of Lyria 3
Google launched the brand new Gemini 3.1 Professional, a mannequin that includes advanced problem-solving and superior reasoning capabilities. The mannequin achieved a verified rating of 77.1% on the ARC-AGI-2 benchmark, a rigorous check within the AI world. This result’s greater than double the reasoning efficiency of its predecessor, Gemini 3 Professional. In line with Google, 3.1 Professional doesn’t simply generate right solutions; it could visualize advanced subjects, make giant datasets immediately comprehensible, and supply holistic options for artistic initiatives.
One other main replace for Gemini this week is the addition of a music technology function to the Gemini app. Powered by the Lyria 3 mannequin developed by Google DeepMind, customers can now create 30-second tracks of their desired type utilizing textual content prompts. Moreover, customers can add a photograph or video and ask the AI to compose a track that matches the temper of the content material.
Taalas Unveils Its First Chip, Revolutionizing AI {Hardware}
Taalas, a US-based startup, introduced the HC1 chip, which might revolutionize AI {hardware}. The primary distinction between this chip and platforms like Nvidia Blackwell is that it “embeds” an AI mannequin instantly into the silicon slightly than operating it as software program. The mannequin’s parameters are bodily etched into the chip, reaching sub-millisecond latency and very low-cost inference efficiency.
In line with the corporate, HC1 provides inference speeds two orders of magnitude quicker than GPU-based methods and about 10 instances greater than the Cerebras Techniques platform. The primary model was optimized for Meta’s Llama 3.1 8B mannequin. Taalas goals to enter the market with each {hardware} gross sales and a cloud service mannequin.
Meta Patents “Digital Clone” for Deceased Customers’ Accounts
Meta has secured a patent for an AI mannequin that may proceed posting on social media accounts after a consumer passes away. Authorized final December, the patent goes past textual content and picture sharing; it covers a system able to simulating the deceased individual’s voice and video calls.
The LLM-based system will analyze a consumer’s social media historical past to create a digital copy able to interacting with different customers, replying to direct messages, leaving likes, and writing feedback.
Apple Growing 3 New AI-Powered Wearables
In line with a brand new report by Mark Gurman, Apple is making ready to considerably develop its wearable portfolio. The corporate is engaged on three new AI-focused merchandise that can combine with the iPhone: good glasses, a necklace-type AI accent, and next-generation AirPods geared up with cameras. These units will analyze the consumer’s environment by way of built-in cameras, permitting Siri to take acceptable contextual actions.
OpenAI’s First {Hardware} Gadget Revealed: A Digital camera-Geared up Good Speaker
OpenAI’s collaboration with iconic Apple designer Jony Ive for AI-supported {hardware} has reportedly settled on its first product: a wise speaker with an built-in digital camera. Deliberate for launch in early 2027 (February on the earliest), the machine contains a facial recognition system much like Apple’s Face ID, permitting it to supply contextual recommendations by detecting customers and their environment. OpenAI can be reportedly engaged on good lamps and AI glasses for 2028 and past.
Samsung Introduces the New and Superior Bixby
Samsung introduced a brand new model of Bixby as a part of the One UI 8.5 replace. Positioned as a extra superior “machine agent,” the up to date Bixby contains a redesigned interplay mannequin that understands pure conversational patterns. Customers not must know technical phrases or settings menu names; Bixby analyzes the intent from on a regular basis phrases and seamlessly prompts the suitable settings.
New AI Instruments Launched This Week
Alibaba Qwen 3.5: Alibaba’s most superior LLM but contains a 1M token context window. Whereas skilled on 397 billion parameters, it effectively makes use of solely the required 17 billion energetic parameters per process. It processes each textual content and visible/video inputs.Anthropic Claude Sonnet 4.6: Now the default for Free and Professional customers, that includes a 1M token context window and improved coding and pc use abilities.AnchorWeave: An open-source different to Google’s Genie 3, able to creating 3D interactive worlds.AudioX: A brand new mannequin that analyzes written and visible inputs to generate corresponding audio and music information.
AI World Briefs
Nvidia is making ready to take a position $30 billion in OpenAI, doubtlessly pushing OpenAI’s valuation to $730 billion.Samsung introduced complete Perplexity integration for the Galaxy S26 collection, activated by way of the “Hey Plex” wake phrase.The Pentagon is threatening to penalize Anthropic over restrictions on the army use of its AI, contemplating declaring the corporate a “provide chain threat.”Nvidia CEO Jensen Huang teased {that a} new chip to be launched on March 16 on the GTC 2026 occasion will “shock the world.”BIOPREVENT, an AI software for organ transplants, can predict issues after stem cell and bone marrow transplants months prematurely.It was revealed {that a} faculty shooter in Canada who killed 8 individuals had his ChatGPT conversations flagged by OpenAI workers months prior, however the firm selected to not inform the police.OpenAI CEO Sam Altman denied claims that ChatGPT has excessive water consumption (allegedly 64 liters per question).A brand new report by climate-focused NGOs revealed that tech giants’ claims about AI’s local weather advantages lack verifiable proof.A multi-year settlement between Nvidia and Meta will see tens of millions of Blackwell and Rubin chips provided, alongside Grace and Vera CPUs deployed in knowledge facilities.Claude Code Safety was launched, scanning software program code like a human cybersecurity skilled and recommending acceptable patch packages.Google NotebookLM gained slide enhancing and PPTX export assist, initially accessible to Professional and Extremely customers.
