This Week in NLP #321
Keep up with what happened in NLP in the week ending Friday 10th January 2025.
Above the Fold
Overwhelmed? Here’s our pick for five things to know about this week.
Meta blindsided its fact-checking partners by abandoning third-party verification in favor of Community Notes, leaving many organizations concerned about their financial survival. [Wired]
Microsoft announced plans to invest US$80B in AI-enabled data centers through 2025, strengthening its position in AI infrastructure and services. [Yahoo Finance]
Nvidia announced a $3,000 personal AI supercomputer called Digits, capable of running LLMs with up to 200 billion parameters from home or office. [Wired]
OpenAI claimed it knows how to build AGI and is shifting focus toward superintelligence. [TechCrunch]
OpenAI has been losing money on its US$200 monthly ChatGPT Pro subscription due to higher-than-expected usage. [TechCrunch]
Now read on for everything else that happened in NLP this week.
Making News:
What makes this newsletter different?
This Week’s Topics:
If you’re reading this in a desktop web browser, you can access a navigation menu at the left-hand edge of this window.
The Generative AI Wars
Alibaba Cloud slashed prices on its AI offerings by up to 85%, joining other tech giants in an intensifying price war across China and the West. [The Decoder]
AWS announced an US$11B investment in Georgia to expand its cloud and AI infrastructure. [TechCrunch]
Google outlined its ambitious AI strategy for 2025, including plans for Project Astra, while ruling out premium test-time subscriptions similar to ChatGPT Pro. [The Decoder]
Meta removed its widely-ridiculed AI-generated social media profiles after users discovered they couldn’t be blocked and criticized their inappropriate appropriation of marginalized identities. [Engadget]
Microsoft announced AI partnerships across India’s major sectors and pledged US$3B in investments, while planning to train 500,000 people in AI by 2026. [TechCrunch]
Microsoft offers the US a roadmap to win the AI race against China. [GeekWire]
OpenAI's new o3 model achieves record-breaking AI performance scores but at an astronomical cost of over $1,000 per computing task. [Futurism]
Here’s everything Nvidia CEO Jensen Huang announced in his CES 2025 keynote. [Engadget]
Memphis Light, Gas and Water warned it may struggle to meet power demands for xAI's planned supercomputer facility expansion. [Fortune]
Feature Creeps
Perplexity has launched a Tripadvisor integration that enhances hotel search results with detailed information, ratings, and AI-generated recommendations for accommodations. [The Verge]
Google has launched Daily Listen, an experimental feature that transforms users’ Discover feed stories into a personalized five-minute audio digest. [TechRadar]
Google Gemini has been discovered in Android Auto’s code, suggesting an upcoming integration that would enhance driving assistance with AI-powered conversational features. [TechRadar]
Microsoft's Copilot Vision has launched in the Edge browser as a conversational AI tool that analyzes visual and textual content on web pages, offering spoken descriptions and context. [PC Magazine]
And Microsoft has discontinued Windows 11’s ‘Suggested actions’ feature after poor performance and user complaints about inconsistency and irrelevant suggestions. [TechRadar]
Hardware
AMD unveiled a comprehensive lineup of AI and gaming processors at CES 2025, including the Ryzen AI Max, AI 300, and AI 200 series CPUs. [Yahoo Finance]
Apple Intelligence’s storage requirement increased from 4GB to 7GB to accommodate new AI features like ChatGPT integration and Visual Intelligence. [ZDNet]
Asus announced a Copilot Plus-capable mini PC with a dedicated Copilot button, leading a wave of AI-enhanced desktop computers from various manufacturers. [The Verge]
Broadcom secured a large high-bandwidth memory order from SK Hynix for custom AI chips, supporting major tech companies’ efforts to reduce Nvidia dependence. [TechRadar]
Intel launched its Core Ultra Series 2 processors, featuring enhanced AI capabilities, improved efficiency, and performance gains across mobile and desktop computing. [Yahoo Finance]
Lakeside Software has partnered with Intel to develop self-healing AI PCs using NPUs for real-time analytics and predictive IT management. [TechRadar]
Lenovo unveiled its lineup of AI-powered business, consumer and gaming PCs at CES 2025 with a focus on smart AI for all. [VentureBeat]
Nvidia announced its Blackwell GPU series at CES 2025, with the flagship RTX 5090 offering twice the performance of its predecessor for $1,999. [Engadget]
Qualcomm introduced new Snapdragon X Platform chips for affordable AI-capable PCs, with Dell and Lenovo planning to launch compatible laptops in early 2025. [Yahoo Finance]
Virtium Embedded Artists launched the DEEPX iMX8M Mini AI Kit, integrating a 25 TOPS AI Booster with NXP’s processor for efficient edge AI deployment. [Business Wire]
Consumer AI
Google announced plans to integrate Gemini AI into Google TVs with far-field microphones and proximity sensors, enabling more natural voice interactions and ambient features. [Engadget]
Halliday announced AI-enabled smart glasses featuring an invisible display, conversation analysis, and translation capabilities. [Engadget]
LG announced five new AI features for its 2025 TV lineup at CES, including personalized recommendations, AI Concierge, enhanced search, chatbot support, and extended WebOS updates. [ZDNet]
MeetKai, a startup with AI and immersive technology, is launching an AI-based OS for smart glasses. [VentureBeat]
Nuwa Pen has developed a smart ballpoint pen that digitizes handwritten notes through AI and computer vision, launching in February 2025 for $295. [VentureBeat]
Omi launched an $89 wearable device that listens to conversations, provides summaries and information, and features early brain-computer interface capabilities and an app store. [The Verge]
Philips Hue announced an AI assistant for its app that will create personalized lighting scenes based on text or voice prompts. [Engadget]
Samsung's Galaxy S25 is rumored to be launching with numerous unspecified AI features that will reportedly demonstrate leadership over Apple's AI capabilities. [TechRadar]
Samsung and LG announced their 2025 smart TVs will include Microsoft's Copilot AI assistant. [The Verge]
TCL announced its NXTPAPER 4.0 display technology and NXTPAPER 11 Plus tablet at CES 2025, featuring enhanced eye comfort and AI capabilities. [PR Newswire]
It’s Only a Model
Apple has partnered with Nvidia to integrate its ReDrafter technology into TensorRT-LLM, achieving 2.7x faster token generation for LLMs. [TechRadar]
Cerence AI expanded its collaboration with Nvidia to enhance its CaLLM language models using NVIDIA AI Enterprise software and DRIVE AGX Orin hardware. [GlobeNewswire]
Microsoft has developed a Large Action Model AI system that operates Windows programs autonomously, outperforming GPT-4o in speed while achieving a 71% task success rate. [The Decoder]
And Microsoft released its Phi-4 AI model as an open-source project on Hugging Face, making the efficient 14-billion-parameter system available for commercial use. [VentureBeat]
Nvidia launched its Nemotron AI model families at CES 2025, introducing open-source language and vision models designed to advance enterprise-focused AI agents. [VentureBeat]
Whose Data?
Anthropic has agreed to maintain copyright protections after music publishers sued over unauthorized use of song lyrics in training its Claude AI model. [Engadget]
Apple settled a US$95m privacy lawsuit while maintaining that Siri data was never sold to advertisers or used for marketing profiles. [TechRadar]
An investigation by The Atlantic revealed that OpenSubtitles had provided dialogue from over 138,000 films and TV shows to train AI models without permission. [NiemanLab]
The LLM Ecosystem
Anthropic’s Model Context Protocol is planning first-half 2025 developments across remote connectivity, reference implementations, distribution tools, agent capabilities, and ecosystem standardization. [Model Context Protocol]
AWS launched AI-powered features in Amazon Q Developer that transform .NET Framework applications to cross-platform .NET, promising 4x faster modernization and 40% lower licensing costs. [InfoQ]
Bubba AI launched an open-source platform automating SOC 2, ISO 27001, and GDPR compliance. [AiThority]
Cohere has released North, a secure AI workspace platform that directly challenges Microsoft Copilot and Google Vertex AI in the enterprise market. [VentureBeat]
Meta released BLT, an open-source LLM architecture that processes byte patches dynamically instead of using tokenization, achieving Llama 3-level performance with half the FLOPS. [InfoQ]
Nvidia announced foundation models for RTX AI PCs, introducing NIM microservices and AI Blueprints for local AI processing and development capabilities. [VentureBeat]
Agentic AI
Accenture launched its AI Refinery for Industry, introducing 12 industry-specific AI agent solutions to help organizations deploy specialized workforce-enhancing networks faster. [Business Wire]
EdgeRunner AI has partnered with Intel to integrate its air-gapped, on-device AI agent technology Athena into Intel AI PCs, enabling offline AI capabilities. [Business Wire]
Nvidia launched AI Blueprints, a collection of partner-developed and in-house tools designed to help customers build autonomous AI applications with reasoning capabilities. [TechRadar]
SupplyWhy has launched out of stealth mode with Jenae, an agentic AI assistant designed to improve automotive supply chain planning through predictive insights and workflow automation. [Business Wire]
Swimlane has launched Hero, a private AI security operations companion that leverages contextual awareness and proprietary LLM technology to accelerate threat response times. [Business Wire]
Other LLM Sightings
AI Unlimited Group has unveiled its 2025 vision for an integrated ecosystem of AI platforms spanning financial planning, debt management, travel, and operational efficiency. [GlobeNewswire]
Amazon Ads launched an AI-powered SQL generator for Amazon Marketing Cloud that helps advertisers create audience-targeting queries using natural language commands. [Yahoo Finance]
Articul8 AI launched A8 Essential, a self-service generative AI product helping enterprises extract insights from complex data without requiring technical expertise. [AiThority]
Cirium has launched a generative AI assistant for analyzing airline and airport on-time performance data and trends. [Business Wire]
DataXstream launched OMS+ 4, an AI-enhanced platform that accelerates order processing and automation within SAP environments through intelligent document handling and material matching. [Business Wire]
DigitalChalk has partnered with AWS to enhance its Learning Management System with AI-powered features for content development, testing, and personalized learning experiences. [Business Wire]
EliseAI launched AI-Guided Tours, an AI-powered self-guided real-estate touring system that enables prospective renters to view properties independently while receiving virtual assistance. [Business Wire]
Google unveiled a Gemini AI-powered News Brief feature for its TV operating system that summarizes daily news from internet sources and YouTube headlines. [TechCrunch]
Inworld AI has partnered with Streamlabs and Nvidia to develop an AI streaming assistant that provides real-time technical support and co-hosting capabilities for content creators. [The Verge]
JargonAi has launched an AI-powered tool that automatically summarizes SEC filings and delivers real-time alerts to users tracking public companies. [Business Wire]
KRAFTON unveiled an AI companion framework, developed with Nvidia, that enables game characters to interact naturally with players using small language models. [Business Wire]
Product Genius launched an AI-powered customer engagement platform with a $15,000 ROI guarantee, offering real-time issue resolution and automated feedback collection. [Business Wire]
Razer unveiled Ava, a controversial AI gaming assistant that analyzes gameplay footage to provide real-time coaching and tips, raising concerns about guide attribution and gameplay disruption. [The Verge]
Taxfyle announced the upcoming launch of Luis, an AI customer assistant trained on CPA and IRS publications to streamline tax filing processes. [PR Newswire]
Veritone has enhanced its Contact solution with Contact Analytics, enabling law enforcement agencies to analyze stop data in near real-time rather than waiting for year-end reports. [Business Wire]
XAgent AI launched an intelligent bidding platform combining AI agents, vertical LLMs, and big data analytics to revolutionize global procurement markets. [EIN Presswire]
Risks and Responses
Apple acknowledged issues with its AI-generated news summaries and promised warning labels, but critics argue stronger measures are needed to address accuracy problems. [Six Colors]
Geoffrey Hinton doubled his prediction of AI-driven human extinction to 20% within 30 years. [Inventiva]
McAfee launched an AI-powered Scam Detector to protect users against text, email, and deepfake video scams. [Business Wire]
Meta announced plans to relocate its content moderation teams from California to Texas, alongside other major changes to its content oversight approach. [The Verge]
OpenAI has shut down an engineer’s ChatGPT-powered robotic rifle project that could autonomously aim and fire in response to voice commands. [Futurism]
OpenAI has developed a three-stage training process for its o-series AI models to actively reason through safety guidelines, though hackers have demonstrated the system remains vulnerable to manipulation. [The Decoder]
OpenAI released two papers detailing its advanced red teaming approach, combining external expertise with automated testing to identify AI model vulnerabilities. [VentureBeat]
TrueMedia.org, the Seattle-based nonprofit that offered AI tools to identify and combat deepfakes in the lead-up to the 2024 elections, will shut down its online service and open-source its technology. [GeekWire]
xAI has revealed plans for Grok’s Unhinged Mode’, which will deliver deliberately offensive responses. [TechCrunch]
Regulation
The Texas Responsible AI Governance Act (TRAIGA) aims to establish broad AI regulations requiring compliance documentation and imposing liability for algorithmic discrimination. [Hyperdimensional]
New York Assemblymember Alex Bores has drafted the RAISE Act to regulate advanced AI systems, incorporating safety plans and whistleblower protections while addressing concerns from California’s failed SB 1047. [MIT Technology Review]
Conversational AI
Botzbrain has launched Fiona AI Agent, an AI-powered virtual assistant that handles customer service tasks including calls, scheduling, and sales outreach. [EIN Presswire]
Talkdesk has launched AI Agents for Retail, introducing autonomous AI capabilities that handle complex retail customer service operations around the clock. [Business Wire]
Vonage has enhanced its Contact Center platform with AI-powered tools and unified communications capabilities to streamline customer service operations and boost agent productivity. [TechRadar]
Voice News
Cerence's stock soared 65% after announcing an expanded partnership with Nvidia to enhance its automotive language models using Nvidia’s AI Enterprise platform. [Yahoo Finance]
MIT researchers have developed an AI system that produces human-like vocal imitations without training, using a model of the human vocal tract and context-aware algorithms. [MIT News]
SoundHound AI has partnered with Lucid to launch the Lucid Assistant, a multilingual voice-control system for their electric vehicles. [Business Wire]
And SoundHound AI unveiled what it says is the first in-vehicle voice commerce platform, enabling drivers to order takeout hands-free through their car’s infotainment system. [Business Wire]
Translation
EarFun announced the Air Pro 4+ AI-translation earbuds with dual drivers and a companion Auracast dongle, launching in May 2025 for $99.99. [TechRadar]
ELEHEAR unveiled its Beyond Pro smart hearing wearable, featuring AI translation, tinnitus relief, and VOCCLEAR technology for enhanced sound clarity. [PR Newswire]
Timekettle has launched Babel OS, an AI-driven operating system for its translation earbuds that enables real-time interpretation across 40 languages with human-like emotion. [VentureBeat]
Traini launched an iOS app featuring AI-powered pet translation technology, enabling dog owners to better understand their pets’ emotions through voice and behavior analysis. [Business Wire]
Ulatus has introduced its JavaScript Proxy Solution, which simplifies software localization by making a product multilingual with just one line of code. [MultiLingual]
Vasco launched its Translator E1 earbuds, offering real-time translation between 51 languages through a tap-to-talk system requiring no subscription. [Wired]
VideoLAN demonstrated VLC’s new offline AI-powered real-time subtitle generation and translation feature at CES 2025, supporting over 100 languages without cloud services. [The Verge]
Search
Alibaba International's AI-powered B2B search engine Accio has attracted 500,000 SME users since its November 2024 launch, offering NLP across multiple languages. [PR Newswire]
Google transformed internet search by introducing AI Overviews in 2023, using language models to generate conversational answers instead of just providing links. [MIT Technology Review]
Microsoft has been accused of deceiving users by making Bing’s search results for ‘Google’ mimic Google’s homepage, prompting criticism from a senior Google executive. [TechRadar]
WaveCX, a provider of personalized, digital product engagement solutions for financial institutions, launched an AI-powered documentation site and platform update featuring its Curator search tool. [Business Wire]
Writing Assistance
Grammarly launched Authorship, a beta feature that tracks and verifies text sources within Google Docs, helping users demonstrate the origins of their content. [ZDNet]
TCL unveiled its 60 XE NXTPAPER 5G smartphone, featuring a Text Assistant that facilitates translation, summarization, and rewriting. [PR Newswire]
AI in Journalism
Baekdal‘s custom ChatGPT bot analyzes news articles’ audience fit and suggest content improvements, offering free limited access to journalists. [journalism.co.uk]
Better Leaders Lab founder Anita Zielina shared multiple AI-powered management hacks to help newsroom leaders improve efficiency and reduce workplace stress. [journalism.co.uk]
Otherweb launched Press Hub, an AI-powered suite of tools designed to help journalists handle routine tasks and produce content more efficiently while maintaining editorial control. [Business Wire]
Health Tech
CentralReach has achieved significant success with its AI-powered clinical note solutions, helping over 100 organizations protect US$100m in revenue since launching in July. [GlobeNewswire]
eClinicalWorks has partnered with Clinica Medica Primaria de Rising Sun to implement Sunoh.ai, reducing documentation time by one-third while improving Spanish-language patient care. [Business Wire]
Hippocratic AI has launched an AI Agent App Store enabling clinicians to create and monetize healthcare AI agents for patient care tasks. [Business Wire]
MD Revolution has launched RevConnect Workflow, a low-code integration platform that connects AI models and applications to its RevCare care management system. [PRWeb]
Movano Health has launched EvieAI, a virtual wellness assistant trained on medical journals that provides Evie Ring users with scientifically-verified health information. [ZDNet]
Sensory has partnered with Intuition Robotics to integrate TrulyHandsfree technology into ElliQ, enabling voice-activated interactions for their AI companion robot for older adults. [EIN Presswire]
Sonde Health has integrated its voice-based health monitoring technology into Qualcomm's Snapdragon S7+ Gen 1 Sound Platform, enabling health tracking in everyday audio devices. [Business Wire]
Verana Health's CEO discusses how unstructured data sources like physician notes and imaging will enhance clinical research efficiency in 2025. [Applied Clinical Trials]
Legal Tech
Alexi has launched Advanced Legal Reasoning, an AI platform enabling lawyers to analyze thousands of documents and uncover strategic insights within minutes. [Business Wire]
Housing Court Answers has launched Roxanne, an AI-powered assistant helping New York City tenants navigate repair issues and understand their legal rights. [LawSites]
Spellbook has launched Playbooks, an AI-powered contract review feature that helps in-house legal teams codify and automate their negotiation processes for faster document review. [LawSites]
Funding
Anthropic is in advanced talks to raise US$2B in funding, valuing the AI startup at US$60B and making it America’s fifth-most valuable startup. [Yahoo Finance]
Bluenote has raised US$10m in Lux Capital-led funding to expand its AI platform that streamlines regulatory workflows for life sciences companies. [FinSMEs]
Carecode has raised US$4.3m in pre-seed funding to develop AI agents that handle pre- and post-appointment tasks for healthcare providers in Brazil. [TechCrunch]
Consist.AI has raised US$580k in seed funding to expand its AI-driven healthcare insurance fraud management operations across India, Middle East, and Africa. [FinSMEs]
Deep Vector has raised US$1.5m in seed funding to expand its AI platform that converts insurance underwriting documents into actionable data. [FinSMEs]
Fazeshift has raised US$4m in seed funding to develop its AI-powered accounts receivable automation platform for enterprises. [Business Wire]
Hamming AI raised US$3.8m in seed funding to develop automated testing and monitoring solutions for AI voice agents. [Hamming AI]
Hippocratic AI has raised US$141m in Series B funding, reaching a US$1.64 billion valuation while expanding its healthcare-focused LLM services. [Business Wire]
HuLoop Automation has secured Series A funding to enhance its unified automation platform serving financial services clients. [FinSMEs]
Innovaccer has secured US$275m in Series F funding to expand its healthcare AI platform, enhance cloud capabilities, and scale its developer ecosystem. [Business Wire]
Jentic, an Irish startup building the integration layer for AI, has raised €4min in pre-seed funding. [Tech Funding News]
Mili has raised US$2m in seed funding to expand its AI-powered meeting documentation platform for wealth management firms. [FinSMEs]
Nvidia has invested US$1B in 50 AI startups during 2024. [Yahoo Finance]
Prudentia Sciences has raised US$7m in funding to enhance its AI-powered platform for optimizing life sciences investments and drug development decisions. [FinSMEs]
Qualified Health has launched with US$30m in seed funding to develop infrastructure enabling safe implementation of generative AI solutions across healthcare organizations. [AiThority]
SoSoValue has raised US$15m in Series A funding to expand its AI-driven investment research platform and develop new products for retail investors. [FinSMEs]
Stellar has secured US$2.7m in seed funding to expand its AI solutions business. [Business Wire]
AI cloud infrastructure startup Vultr raised US$333m in growth financing at a US$3.5B valuation. [Crunchbase News]
Acquisitions
CUBE has acquired Thomson Reuters Regulatory Intelligence and Oden, doubling its workforce and expanding its global RegTech services to approximately 1,000 customers. [FinTech Global]
HubSpot has completed its acquisition of AI-powered conversation intelligence platform Frame AI. [MarTech Series]
Lusha has acquired Novacy, integrating AI-powered conversation intelligence into its B2B sales platform to provide comprehensive sales solutions from prospecting through deal closure. [FinSMEs]
Thomson Reuters has acquired cloud-based tax automation company SafeSend for US$600m, expanding its tax preparation software portfolio and strengthening its last-mile tax return capabilities. [TechCrunch]
There’s More
Here are some reflections from Sam Altman. [Sam Altman]
Consumers don’t want AI to seem human. [Harvard Business Review]
Shares in the quantum computing sector plunged after Nvidia's CEO said useful quantum computers were decades away. [Axios]
OpenAI's new o3 reasoning model has triggered a wave of anxiety among computer science majors who fear AI will edge them out of the job market. [Axios]
This piece argues that writers will regain recognition and value in 2025 as AI-generated content’s limitations drive demand for authentic human creativity and quality journalism. [Wired]
The Back Cover
Here’s Passenger with a cover of David Bowie’s Space Oddity.
If you know of a great cover version of a back-catalogue classic, drop an email to us at backcover@language-technology.com and we’ll consider it for inclusion here.
Got this from someone else?
Did you find this newsletter useful? if so, you might forward it to a friend; or you could email us at news@language-technology.com to tell us what you want more of.
Did you hate this newsletter? if so, you could forward it to an enemy; or you could email us at news@language-technology.com to tell us why – make it better!