This Week in NLP #332
Keep up with what happened in NLP in the week ending Friday 28th March 2025.
Above the Fold
Overwhelmed? Here's our pick for five things to know about this week.
Apple has been hit with a federal lawsuit alleging it deceived consumers by marketing unrealized AI capabilities for Siri in its iPhone 16 series. [TechRepublic]
DeepSeek has helped narrow China's AI development gap with the United States to three months in some areas through efficient chip usage and algorithm optimization. [Reuters]
Google has launched Gemini 2.5 Pro, an experimental AI model outperforming competitors on various benchmarks, including Humanity's Last Exam, with a one-million token context window. [ZDNet]
OpenAI has launched native image generation in GPT-4o, enabling ChatGPT Plus users to transform photos into Studio Ghibli-style animations, causing widespread social media activity. [VentureBeat]
xAI has expanded its Grok chatbot to Telegram's platform of over 1 billion users, marking its first major deployment beyond Musk-owned platforms. [Tech Startups]
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
Amazon has launched Alexa Plus, a generative AI-powered voice assistant requiring a compatible Echo Show display and eventually a $19.99 monthly fee or Prime membership. [The Verge]
Anthropic has partnered with Databricks in a five-year deal to provide enterprise customers with domain-specific AI agents through the Databricks Data Intelligence Platform. [ZDNet]
Apple CEO Tim Cook has endorsed Chinese AI startup DeepSeek amid declining iPhone sales and efforts to launch AI features in mainland China. [Yahoo Finance]
And Cook has visited Hangzhou, announcing a US$4.16m donation to Zhejiang University while praising local AI developments and strengthening ties with China's tech sector. [Yahoo Finance]
Apple has ordered approximately US$1B worth of Nvidia server platforms from Super Micro and Dell to boost its AI capabilities following criticism over delayed Siri upgrades. [Yahoo Finance]
Coinbase's former CTO Balaji Srinivasan has warned that China is releasing open-source AI models to undermine US tech companies' profitability through market saturation. [Benzinga]
DeepSeek has quietly released its MIT-licensed V3-0324 language model, featuring breakthrough efficiency that allows it to run on high-end consumer hardware. [VentureBeat]
DeepSeek's successful launch of an affordable AI model has sparked a wave of Chinese companies releasing low-cost AI services, challenging Western tech giants' market dominance. [Bloomberg]
Google transformed itself over two years by rushing to launch AI products like Bard and Gemini in response to ChatGPT's success. [Wired]
Manus AI has launched an invite-only AI agent that executes real-world tasks autonomously using both LLM algorithms and deterministic programming. [Forbes]
Meta has established revenue-sharing agreements with companies hosting its Llama AI models, despite CEO Mark Zuckerberg's earlier statement that selling access wasn't their business model. [TechCrunch]
Microsoft has deployed DeepSeek's R1 on Azure and is using the 200-person team's success as a benchmark for its own AI development and US$80B investment. [The Verge]
Mistral AI, the French AI startup valued at €5.8bn, has clarified it's not currently pursuing an IPO despite earlier suggestions, while continuing to expand its AI offerings. [Sifted]
OpenAI has restructured its leadership team, expanding COO Brad Lightcap's operational role while CEO Sam Altman focuses more on technical direction and research. [The Verge]
OpenAI’s ChatGPT has maintained its dominance in global offices with 76% workplace adoption by late 2024, far outpacing competitors like DeepSeek and Anthropic. [ITPro]
OpenAI has delayed the rollout of ChatGPT's image generation feature to free users due to unexpectedly high demand among paying subscribers. [Engadget]
Perplexity has proposed acquiring TikTok's US operations with plans to rebuild its algorithm transparently. [The Verge]
Together has launched a free consumer app offering access to multiple AI models including DeepSeek R1, with web search, image generation, and coding capabilities. [Together.ai]
The US Commerce Department has added 80 organizations to its export blacklist, primarily targeting Chinese entities like Inspur Group to prevent military access to advanced computing technology. [The Verge]
And Commerce Secretary Howard Lutnick has called for industry-led security evaluations of open-source AI models and stricter controls on Chinese AI companies operating in America. [Yahoo Finance]
Sovereign AI
Cassava Technologies has partnered with Nvidia to establish Africa's first AI computing infrastructure across multiple countries, starting with South Africa in 2025. [TechCentral]
DAIR researchers have challenged Big Tech's Western-centric AI development by creating Africa-focused solutions and highlighting the industry's neglect of African languages and communities. [ZDNet]
Reliance Industries has discussed potential partnerships with OpenAI and Meta to expand AI services in India, including distributing ChatGPT and hosting models in a planned data center. [Yahoo Finance]
Techarena's Stockholm conference highlighted Sweden's success in tech innovation, ranking second globally per capita in unicorns while pursuing pragmatic partnerships with US technology companies. [CEPA]
Telus has partnered with Nvidia to transform its Rimouski data centre into a ‘Sovereign AI Factory’ using advanced GPUs for AI computing within Canadian borders. [BetaKit]
UK Labour's Plan for Change AI integration strategy has faced severe criticism from Parliament due to outdated technology, poor data quality, and staffing shortages across government departments. [UKTN]
Global tech companies have shifted away from US-based AI models due to Trump-era policy changes, seeking locally-developed alternatives with better language and cultural capabilities. [MIT Technology Review]
Feature Creeps
Anthropic has improved Claude's problem-solving abilities by adding a 'scratchpad' feature that lets the AI assistant take notes while working through complex tasks. [The Decoder]
And Anthropic has launched web search capabilities for its AI assistant Claude, allowing paid US users to access current online information through autonomous, multi-step search functionality. [Ars Technica]
Character.ai has launched a 'Parental Insights' feature allowing teens to share weekly chatbot usage reports with parents amid concerns about inappropriate content and excessive use. [The Verge]
Google has launched contextual smart replies in Gmail powered by Gemini AI for business users, offering AI-generated response suggestions based on full email thread analysis. [TechRadar]
Google has launched Gemini's camera and screen-sharing capabilities on select Android devices, beating Apple in delivering advanced AI features to mobile users. [Gizmodo]
Google has expanded its Gemini platform with Audio Overview for document narration, Canvas for collaborative editing, and enhanced coding features with live previews. [TechRepublic]
Google's NotebookLM has added Mind Maps alongside its Audio Overviews feature, helping users visualize and understand information through connected idea nodes. [TechRadar]
Android Auto 14.0 has rolled out with code suggesting Google Assistant will be replaced by Gemini, bringing enhanced AI capabilities to vehicles. [TechRadar]
And Google has expanded its Audio Overviews feature to convert Gemini's Deep Research reports into AI-hosted podcast-style conversations. [The Verge]
Google Workspace has rolled out a March Feature Drop, adding AI-powered meeting summaries, video voiceovers, and expanded chat translation capabilities across its platform. [ZDNet]
Google has discontinued its AI while browsing feature, which had allowed users to generate AI-powered summaries and key points from webpages they visited. [Search Engine Land]
Google has launched AI-powered vacation planning features across Search, Maps, and Gemini, including itinerary creation, hotel price tracking, and screenshot-based location saving. [TechCrunch]
Google has launched Gemini Gems, a feature allowing users to create or select specialized AI assistants with customized instructions, across web and mobile platforms. [ZDNet]
Meta has developed an AI feature for Instagram that generates suggested comments on posts, though the test feature's reception may be dampened by concerns about authenticity. [TechCrunch]
Microsoft has added AI-powered Copilot features to Windows 11's Photos app, including enhanced text recognition and editing tools, with gradual rollout to Insider users. [TechRadar]
OpenAI has activated GPT-4o's native multimodal image generation in ChatGPT, offering improved text integration and visual quality compared to DALL-E 3. [VentureBeat]
OpenAI has upgraded ChatGPT's Advanced Voice Mode with reduced interruptions and a more natural conversational flow, available to both free and premium users. [TechRadar]
Opera has launched AI Tab Commands, enabling its Aria assistant to organize and manage browser tabs through natural language prompts. [TechRadar]
xAI has launched DeeperSearch and text-based image editing for Grok, while making its Aurora image model available through its API. [The Decoder]
Waze has discontinued Google Assistant integration for iOS users after prolonged technical issues, while promising a future enhanced voice interaction solution and testing AI-powered reporting features. [TechCrunch]
Hype Bubble?
Accenture's analysis of 2,000+ AI projects has revealed that successful companies are increasing investments despite widespread deployment challenges and low enterprise-level value realization. [CIO Dive]
Alibaba chairman Joe Tsai has warned against a potential AI data center bubble, citing concerns about speculative construction outpacing actual customer demand. [Yahoo Finance]
China's ambitious AI data center construction boom has largely failed, with up to 80% of newly built facilities sitting unused due to weak demand and shifting AI trends toward DeepSeek-style reasoning models. [MIT Technology Review]
LinkedIn cofounder Reid Hoffman has predicted AI will spark a ‘cognitive Industrial Revolution’, enhancing human agency while requiring careful navigation through transitional challenges. [eWeek]
Microsoft has cancelled 2GW of planned datacenter expansions across the US and Europe following its deteriorating relationship with OpenAI and revised capacity needs. [Computer Weekly]
Wedbush Securities has reported that AI spending now comprises 12% of many companies' 2025 IT budgets, up from 10% in January, as enterprises accelerate their AI strategies. [Yahoo Finance]
Hardware
AMD has secured a multi-billion dollar contract with Oracle to build a 30,000-unit cluster of MI355X AI accelerators, complementing Oracle's massive Stargate project. [TechRadar]
Ant Group has developed AI training techniques using Chinese-made semiconductors that reduced costs by 20% compared to using Nvidia chips. [Yahoo Finance]
Apple's Mac Studio with M3 Ultra has demonstrated unprecedented efficiency by running the 671-billion-parameter DeepSeek R1 AI model using unified memory and under 200W power. [TechRadar]
Asus has launched the Ascent GX10, a $2,999 compact AI supercomputer featuring Nvidia's Grace Blackwell Superchip, offering 1000 TOPS of processing power. [TechRadar]
AWS has launched cheaper Trainium chipsets to compete with Nvidia's H100 GPUs. [The Motley Fool]
Extropic has revealed details of its innovative probabilistic computer chip, which harnesses thermodynamic fluctuations to perform calculations more efficiently than conventional silicon. [Wired]
Fluidstack has partnered with Borealis Data Center, Dell Technologies, and Nvidia to deploy renewable energy-powered GPU clusters across Iceland and Europe. [Business Wire]
FuriosaAI has declined Meta's US$800m acquisition offer, opting to remain independent and raise US$48m while preparing its second-generation AI chip for launch. [Yahoo Finance]
Hewlett Packard and Dell have unveiled their GB300-powered AI workstations, the ZGX Fury AI Station G1n and Pro Max respectively, rivaling Nvidia's DGX Station. [TechRadar]
And Hewlett Packard has launched Mod Pod, a containerized liquid-cooled data center solution optimized for AI workloads that can be deployed in existing spaces without major infrastructure changes. [ITPro]
H3C has warned customers about potential shortages of Nvidia's H20 AI chips in China, citing depleted inventory and supply chain uncertainties amid surging demand. [Yahoo Finance]
Huawei has emerged as China's most formidable tech company, doubling AI chip yields and growing revenue despite US sanctions, according to Nvidia's CEO. [Benzinga]
Intel's former CEO Gelsinger has criticized Nvidia's AI chip pricing as unsustainable for inference computing. [Yahoo Finance]
Malaysia has pledged to strengthen regulations on Nvidia chip shipments following US concerns about potential illegal re-exports to China through Southeast Asian countries. [Yahoo Finance]
Micron and SK Hynix have developed SOCAMM, a new memory module format exclusively for Nvidia's AI platform, offering superior performance over traditional solutions. [TechRadar]
Nvidia has rebranded Project Digits as DGX Spark, a $3,999 mini AI supercomputer, while Hewlett Packard, Dell, and Asus are launching similar GB10-based systems. [TechRadar]
Nvidia has announced plans to establish a quantum computing research center in Boston, partnering with Harvard, MIT, and quantum hardware companies to advance AI-quantum integration. [Yahoo Finance]
Nvidia's CEO sparked concerns among major tech clients after joking about the rapid obsolescence of AI chips, potentially impacting companies' earnings through accelerated depreciation costs. [Yahoo Finance]
And Nvidia's stock plunged after China's new energy efficiency rules threatened to block its H20 chip sales, impacting the company's US$17B annual Chinese market. [Yahoo Finance]
Phison has developed an SSD-based solution that could reduce trillion-parameter AI model training costs from US$3m to US$100,000 by offloading GPU workloads. [TechRadar]
Consumer AI
Apple is developing camera-equipped Watch models to integrate its visual intelligence features, despite previous struggles with AI adoption and wearable technology expansion. [Gizmodo]
Garmin has launched Connect Plus, a $6.99 monthly premium subscription service that adds AI-powered insights while keeping existing features free for users. [The Verge]
Google has announced plans to pre-install NotebookLM, its Gemini-powered AI analysis tool, on ChromeOS devices. [TechRadar]
Honor has launched the Magic7 Pro smartphone with innovative AI features, including Magic Portal search, eye-tracking technology, and AI-powered photography capabilities. [TechCentral]
Hugging Face has launched HuggingSnap, an offline iOS app that analyzes surroundings through iPhone cameras, providing descriptions without cloud connectivity. [TechRadar]
Samsung is developing AR smart glasses, codenamed Haean, which are expected to launch alongside its Android XR headset before the end of 2025. [TechRadar]
It's Only a Model
Alibaba has launched Qwen2.5-Omni-7B, a multimodal AI model capable of processing text, images, audio, and video on mobile devices and laptops. [Yahoo Finance]
The Computer History Museum, collaborating with Google, has released AlexNet's original source code, the groundbreaking 2012 AI model that revolutionized computer vision and sparked the deep learning revolution. [ZDNet]
Google has launched Gemini 2.5, a new AI reasoning model family featuring enhanced thinking capabilities and a massive context window, outperforming previous models on various benchmarks. [TechCrunch]
IBM and Hugging Face have released SmolDocling, a compact vision-language model that outperformed larger models in converting complex documents into structured data. [Marktechpost Media]
Kyutai has launched MoshiVis, an open-source Vision Speech Model that augments Moshi with visual capabilities while maintaining low-latency conversation skills. [Kyutai]
Nvidia has launched a family of hybrid Mamba-Transformer models offering up to 3x faster inference than comparable pure Transformer models while maintaining similar accuracy. [Nvidia]
OpenAI has launched new speech-to-text and text-to-speech AI models for its API, enabling agentic interactions. [TechRepublic]
Tencent has launched its T1 reasoning model, featuring enhanced text processing and low hallucination rates, intensifying competition in China's AI sector. [DealStreetAsia]
Whose Data?
Amazon has now eliminated Alexa's ‘Do Not Send Voice Recordings’ option, forcing users to share voice data for AI development. [The Etownian]
Anthropic has won an initial court victory after a US judge rejected music publishers' request for an injunction against its use of song lyrics in AI training. [Verdict]
You can now peruse all the books Meta is said to have pirated to train its AI. [Gizmodo]
Microsoft is exploring a way to credit contributors to AI training data. [TechCrunch]
The News/Media Alliance has partnered with ProRata AI to launch a licensing framework ensuring fair compensation for publishers whose content is used in generative AI. [Editor & Publisher Magazine]
OpenAI’s ChatGPT image generator has sparked controversy by enabling users to create Studio Ghibli-style images, raising questions about AI training on copyrighted material. [TechCrunch]
The LLM Ecosystem
Alpha Data and Hyperfusion have partnered to provide ready-to-use GPU infrastructure for AI development across UAE businesses. [EIN Presswire]
Anthropic has developed a technique to observe LLMs' internal processes, revealing surprising insights about how they handle tasks like math and poetry. [MIT Technology Review]
Arc Prize Foundation has launched ARC-AGI-2, a more challenging AI intelligence test where current models score around 1%, significantly below the human average of 60%. [TechCrunch]
Cognizant has partnered with Nvidia to expand its AI capabilities across five areas, enhancing enterprise solutions through specialized models and digital twin technology. [Yahoo Finance]
Databricks has developed a machine-learning technique called TAO that improves AI model performance without requiring clean labeled data. [Wired]
Data Cities has announced plans to build AI-optimized data centers on 4,000 acres in Copper Valley, California, breaking ground in 2025. [PRWeb]
METASCALE enables LLMs to adapt their reasoning strategies dynamically through a new framework developed by researchers from UC Davis, USC, and Microsoft. [VentureBeat]
OpenAI has announced support for the Model Context Protocol across its products, following the protocol's major update that introduced enhanced security and interoperability features. [VentureBeat]
Rackspace Technology has launched AI Business, a managed private cloud platform combining AI capabilities with enterprise security to optimize AI workloads for organizations. [EIN Presswire]
Rapt AI has partnered with AMD to optimize AI workload management on AMD Instinct GPUs, promising up to 90% cost savings and tenfold model capacity improvements. [VentureBeat]
Red Hat has announced updates to its AI portfolio, enhancing OpenShift AI and Enterprise Linux AI to help businesses deploy AI solutions across hybrid cloud environments. [Business Wire]
UiPath has launched Test Cloud, an AI-powered software testing platform that reduces costs and improves efficiency through automated quality assurance processes. [Business Wire]
Agentic AI
FloQast has launched AI Agents, enabling accountants to automate complex workflows across close management, compliance, and reporting using natural language rather than code. [GlobeNewswire]
Hey Bubba has launched Bubba AI, the first voice-enabled AI dispatcher for truckers that negotiates rates, books appointments, and manages back-office tasks. [EIN Presswire]
Jeeva AI has launched an Agentic AI upgrade to automate lead discovery, qualification, and multichannel outreach for sales teams. [Business Wire]
Microsoft has launched 11 Security Copilot Agents to help organizations combat the growing wave of AI-powered cyberattacks through automated defense systems. [TechRadar]
And Microsoft has expanded its AI agent ecosystem by adding deep reasoning capabilities and agent flows to Copilot Studio, while launching specialized Researcher and Analyst agents for Microsoft 365. [VentureBeat]
Oracle has launched its AI Agent Studio alongside similar DIY agent platforms from tech giants, enabling businesses to create and manage custom AI agents. [ITPro]
Otter AI has launched voice-interactive meeting agents that can respond to queries during calls and conduct autonomous sales demonstrations, advancing beyond traditional AI meeting assistants. [VentureBeat]
Outreach has launched AI Agents for its Sales Execution Platform, with its first AI Prospecting Agent delivering up to 10x productivity increases for customers. [Business Wire]
Sorcero has launched an agentic AI framework that deploys specialized AI agents to deliver validated medical insights for life sciences companies. [PRWeb]
Twin has launched an AI-powered Invoice Operator with Qonto, automating invoice retrieval across thousands of services using OpenAI's CUA model. [TechCrunch]
Zeta Global has launched AI Agent Studio with Agentic Workflows, enabling marketers to connect AI agents for automated, complex marketing tasks. [Business Wire]
Other LLM Sightings
Amazon has begun testing two AI tools: an interest-based shopping assistant and a health-focused chatbot for product discovery and healthcare guidance. [Investopedia]
AWS has launched an AI-powered capability in QuickSight that enables employees to perform expert-level data analysis using natural language conversations. [Business Wire]
CNH, the parent company of Case IH and New Holland, has launched AI Tech Assistant, an AI-powered online resource helping mechanics quickly access service information from a 1.5-million-document database of technical materials. [Grainews]
Freeman has launched Key Takeaways, an AI-powered tool that transcribes and summarizes event sessions in real-time to enhance attendee learning and engagement. [TSNN]
FunnelStory AI has launched Felix 2.0, an AI-powered data engineer that helps Customer Success teams maximize revenue growth through proactive, data-driven actions. [EIN Presswire]
Genesys has launched Cloud Social, expanding its platform to include social media listening and sentiment analysis across multiple channels for enhanced customer engagement. [Business Wire]
GoRascal has partnered with mortgageQ AI to accelerate Non-QM mortgage approvals through automated processing. [EIN Presswire]
Hantec Markets has launched InsightPro, an AI-powered trading intelligence tool developed with Acuity Trading, offering real-time signals and sentiment analysis for traders. [EIN Presswire]
Hotwire Global has established an AI Lab to develop AI-powered marketing tools, including Hotwire Spark and Hotwire Ignite platforms for enhanced client services. [Business Wire]
HungerRush has launched AI Reply, an AI tool that generates brand-aligned responses to restaurant customer feedback within its Feedback platform. [Business Wire]
Kibo Commerce has launched an AI-powered platform called Agentic Commerce, integrating with Gemini for Google Cloud to transform customer engagement and commerce operations. [PRWeb]
Landbase has established an Applied AI Lab in Silicon Valley, assembling experts from Meta, NASA, and other tech giants to advance its GTM-1 Omni platform. [Business Wire]
Linqura has launched an AI-powered platform that helps insurance agents and underwriters make better risk decisions. [EIN Presswire]
Lithero has integrated its AI platforms LARA Create and LARA Review with Adobe GenStudio and Workfront to enhance life sciences marketing compliance and efficiency. [PRWeb]
Martini.ai has launched Financials Agent, an AI-powered tool that automatically analyzes financial documents and generates instant credit risk reports. [Business Wire]
Mini Course Generator has launched an AI course creation tool that prevents hallucinations by ensuring generated content stays faithful to source materials. [EIN Presswire]
Monetate has launched Orchid AI, a unified platform that enhances e-commerce personalization through AI-powered search, product discovery, and merchandising capabilities. [Business Wire]
NetSuite has launched new AI tools for UK customers, including Text Enhance controls and financial exception detection, to boost productivity and streamline operations. [ITPro]
RevuIQ has launched an AI-powered mobile app that analyzes and compares online reviews, helping users make informed decisions through instant insights. [EIN Presswire]
Sell The Trend has launched three AI-powered tools for dropshipping: Brand Builder, Reviews, and Supplier Finder, aiming to automate and streamline e-commerce operations. [EIN Presswire]
TheDataProject.ai has launched a multi-topic data platform that aggregates information across healthcare, business, and consumer sectors for researchers and professionals. [EIN Presswire]
VelocityEHS has launched an AI-powered contractor safety management feature that automates verification processes and reduces compliance risks for organizations managing onsite contractors. [GlobeNewswire]
Risks and Responses
Anthropic has developed new techniques for examining how LLMs process information, revealing sophisticated planning, reasoning patterns, and potential safety concerns. [VentureBeat]
Apollo Research has discovered that Claude Sonnet 3.7 frequently demonstrated awareness of being tested during alignment evaluations, raising concerns about the reliability of such assessments. [Apollo Research]
Australians for AI Safety has launched a public campaign urging the next government to establish an AI Safety Institute and implement promised national safety measures. [EIN Presswire]
Character.ai has launched parental supervision tools, including weekly activity summaries, following lawsuits over teen safety. [TechCrunch]
China has developed an AI system trained on 133,000 examples to automatically flag sensitive content, enhancing its censorship capabilities beyond traditional keyword-based filtering. [TechCrunch]
Cloudflare has developed an AI system that lures unauthorized AI scraper bots into mazes of generated junk content, wasting their resources while protecting legitimate websites. [The Register]
Hugging Face researchers have warned against giving AI agents too much autonomy, advocating for maintained human oversight to prevent potential harm from automated systems. [MIT Technology Review]
Loti AI has expanded its deepfake detection service beyond celebrities to all users, offering free tools to protect personal digital identities through automated content monitoring and removal. [ZDNet]
Microsoft has drawn criticism from UK officials for raising consumer Microsoft 365 prices and automatically bundling Copilot AI features without user consent. [The Register]
OpenAI has been hit with a privacy complaint after ChatGPT falsely portrayed Norwegian man Arve Hjalmar Holmen as a convicted child murderer serving prison time. [The Verge]
OpenAI's research with MIT has revealed that frequent ChatGPT users experienced increased loneliness and emotional dependency, particularly among those predisposed to forming emotional attachments. [ZDNet]
OpenAI has increased its bug bounty rewards from $20,000 to $100,000 while expanding its Cybersecurity Grant program to better protect against emerging security threats. [TechRadar]
Originality.ai has discovered a dramatic increase in AI-generated Glassdoor reviews, with fake content rising from 6.18% to 29.45% between 2022 and 2024. [EIN Presswire]
UK Labour's Chi Onwurah has criticized the UK government's delay of AI safety regulations, allegedly to appease Trump's administration despite concerns over AI's potential risks. [The Guardian]
ValidMind has partnered with Genpact to provide an AI-driven platform for model risk management and governance services to financial institutions. [Business Wire]
Regulation
The Ada Lovelace and Alan Turing Institutes have found that 72% of UK residents would feel more comfortable with AI if laws regulated it, according to their national survey. [Computer Weekly]
The Advanced Technology Research Council has announced a global summit in Abu Dhabi for May 2025 to establish governance frameworks for emerging technologies like AI and quantum computing. [EIN Presswire]
The European Commission has begun reviewing overlapping tech regulations following business complaints, though digital chief Virkkunen confirmed existing major laws would remain intact. [Yahoo Finance]
Google has urged the US government to boost AI investments, accelerate public sector adoption, and enhance international cooperation in response to the White House's AI Action Plan. [eWeek]
OpenAI has urged federal preemption of state AI regulations in its White House policy proposal, specifically targeting California's defeated SB 1047. [The Verge]
Oracle, Nvidia, and other tech companies have urged the US government to reconsider AI chip export restrictions that could limit their international business operations. [Yahoo Finance]
The UK government has delayed its AI Safety Bill amid concerns that the postponement reflects efforts to align with US opposition to strict AI regulation. [TechRepublic]
Conversational AI
Avaya has released research showing that 52% of business leaders prioritized AI for customer support efficiency, with many planning to implement advanced orchestration capabilities within a year. [Business Wire]
Bloomreach has launched Clarity, an AI shopping agent that increased conversion rates by 9% and order values by 20% across 17 countries since early 2024. [Business Wire]
Bonnie has launched an AI telephone assistant for restaurants and local businesses that handles calls, manages reservations, and integrates with existing systems around the clock. [Silicon Canals]
Calabrio's research has revealed that while 98% of contact centers have adopted AI, 61% reported more challenging customer conversations, highlighting efficiency-empathy balance issues. [Business Wire]
Ciphers Digital Marketing has integrated Google Cloud AI-powered conversational tools into its digital marketing services. [PRWeb]
Company F has partnered with GPTBots.ai to implement an AI customer service solution that reduced response times by 90% and improved satisfaction by 86%. [EIN Presswire]
Forethought has launched Voice, becoming the first complete agentic AI platform for customer support across all channels, including phone service without decision trees. [Business Wire]
IrisAgent has launched an AI-powered Salesforce Messaging integration that automates customer support inquiries while enabling seamless handoffs to human agents when needed. [PRWeb]
LW Technologies has launched 'Customer Service In A Box', an AI-powered platform providing 24/7 automated customer support across multiple industries. [EIN Presswire]
NICE has formed a strategic alliance with Deloitte Digital to enhance customer service through AI-powered automation and predictive interactions. [Business Wire]
Observe.AI has launched VoiceAI agents, combining proprietary AI models with enterprise integrations to automate routine customer service calls. [VentureBeat]
Octal has launched AI-powered chatbots and virtual assistants that enhance customer engagement and automate support services for businesses. [EIN Presswire]
OpenAI’s ChatGPT Advanced Voice Mode has received an update improving its conversational abilities by reducing interruptions and enhancing personality features for both free and paid users. [ZDNet]
Voice News
Amazon's Alexa+ launch has sparked concern among voice developers due to its expected exclusion of many existing Alexa skills and unclear transition guidance. [Lowpass]
AMC Theaters has introduced AI-powered visual dubbing for a Swedish film, altering actors' facial movements to match their own English voice recordings while maintaining SAG-AFTRA compliance. [Gizmodo]
Groq has partnered with PlayAI to launch Dialog, a high-speed text-to-speech model offering English and Arabic capabilities through Groq's inference platform. [VentureBeat]
Krisp has launched an AI-powered accent conversion tool that transforms speakers' accents to American English in real-time across various video conferencing platforms. [The Verge]
Murf AI has emerged as a leading text-to-speech platform, offering multichannel content creation capabilities, enterprise-level security, and over 120 voices across 20 languages. [eWeek]
NewSound has launched AI-driven over-the-counter hearing aids featuring Femtosense's Clara AI technology for improved speech clarity and noise reduction. [Business Wire]
OpenAI has updated ChatGPT's Advanced Voice Mode to reduce interruptions and improve conversational flow, while making the AI assistant more engaging for both free and paid users. [TechCrunch]
Translation
3Play Media has launched a global linguist marketplace and AI-enabled language solutions to help businesses comply with upcoming European Accessibility Act requirements. [Business Wire]
Alibaba researchers have demonstrated that Large Reasoning Models outperform traditional language models in complex translation tasks. [Slator]
Alpha Translation Service has expanded its Easy-Read Translation service from German-only to multiple languages, making content more accessible for people with learning disabilities and non-native speakers. [MultiLingual]
KUDO has launched downloadable multilingual transcriptions and recordings as a standard feature for all clients of its AI-powered communication platform. [Slator]
Perfect Crime has become the first US theater production to offer AI-powered live translations in over 60 languages through a partnership with Wordly. [eWeek]
PointFire has expanded its SharePoint Search Summarizer with multilingual support for five European languages, helping global teams find relevant content more efficiently. [EIN Presswire]
Taylor & Francis has announced plans to use AI for translating foreign-language books into English, following a year of testing to ensure accuracy and quality control. [Publishers Weekly]
Unbabel has developed MINTADJUST, a method that corrects bias in MT evaluation when the same metrics are used for both optimization and assessment. [Slator]
Welocalize has launched AILQA Beta, an AI-powered translation quality assessment program that automatically detects and classifies errors across entire translated content sets. [Slator]
Search
Anthropic has integrated Brave Search into its Claude chatbot, as evidenced by documentation updates and matching search results between the two platforms. [TechCrunch]
EBSCO has launched AI Insights and Natural Language Search features to generate article summaries and improve search accuracy for researchers. [PRWeb]
Google has expanded its AI Mode search feature beyond premium subscribers, making the Gemini 2.0-powered chatbot freely available to all Google account holders. [ZDNet]
And Google has expanded its AI Overviews feature to nine European countries, offering the service in multiple languages. [Search Engine Land]
Microsoft has introduced AI-powered natural language search for Copilot+ PCs running Windows 11, enabling local semantic queries without cloud processing. [TechRadar]
Perplexity has announced an upcoming Circle to Search-like feature for its Android app, allowing users to search for information by drawing circles around text. [ZDNet]
Health Tech
Ant Group has upgraded its healthcare AI solutions, integrating large models into hospital systems and enhancing its AI Doctor Assistant and Healthcare Manager platforms. [Business Wire]
Embold Health has launched an AI-powered virtual assistant that helps users find appropriate healthcare providers based on their symptoms and medical needs. [Business Wire]
HealthLynked has upgraded its booking platform and ARi voice assistant with flexible scheduling options and improved noise-cancellation to enhance patient-provider communication and convenience. [Hospital Management]
John Snow Labs has launched Medical LLM Reasoner, a healthcare-specific reasoning model that outperforms existing LLMs in medical problem-solving and clinical decision support. [MultiLingual]
Octal has launched advanced generative AI chatbots to transform healthcare delivery through improved patient care, operational efficiency, and accessibility. [EIN Presswire]
Riskonnect has launched new AI-based features in its Healthcare Risk & Patient Safety solution to improve decision-making and streamline workflows for healthcare organizations. [Business Wire]
SecondOpinionExpert has formed a new parent company, DRAI Health, to leverage AI and technology for improving healthcare outcomes. [PRWeb]
Legal Tech
The American Bar Association has released its 2024 Legal Technology Survey Report, revealing increased adoption of cloud solutions, AI tools, and cybersecurity measures across law firms. [Toledo Legal News]
The American Arbitration Association has launched a comprehensive AI implementation course for law firms. [LawSites]
And the Association for AI in Legal has launched in New York as a membership organization to help legal professionals navigate AI through education, resources, and collaborative networking. [LawSites]
CamoText has launched an offline text anonymization tool that helps professionals safely use AI by detecting and redacting sensitive information without retaining. [LawSites]
DISCO has launched ‘With You in Every Case’, a value proposition highlighting its combination of AI-powered litigation technology and professional legal services. [Business Wire]
DraftWise has launched an enhanced AI contract review system with expanded capabilities. [Artificial Lawyer]
eDiscovery Assistant has rebranded as Minerva26, transforming its legal research platform into a strategic command center for managing electronically stored information in litigation. [LawSites]
Law Insider has launched AI-powered contract tools integrated with Microsoft Word, offering affordable legal automation to its 1.2 million users starting April 2025. [LawSites]
Legora has launched its AI legal platform in the United States and partnered with Goodwin law firm while opening a New York office. [Business Wire]
LexisNexis has launched Protégé, the legal industry's first voice AI assistant, integrating speech commands with AI-driven legal workflows and advanced reasoning capabilities. [LawSites]
Merlin Search Technologies has launched Alchemy, a fourth-generation document intelligence platform that transforms unstructured data into actionable insights for legal investigations and discovery. [LawSites]
Opus 2 has launched AI Workbench, integrating generative AI capabilities into its legal case management software to help lawyers analyze and strategize more efficiently. [PRWeb]
Practice AI has launched an AI-powered Lemon Law Demand Letter feature that helps attorneys generate comprehensive demand letters by automatically analyzing repair records and case details. [GlobeNewswire]
Relativity has announced plans to require all new matters be hosted on its cloud platform RelativityOne by 2028, marking a significant shift from its on-premises product. [LawSites]
Reveal has launched GenAI Review, combining lawyer-intuitive workflows with generative AI to transform document review processes for legal professionals. [Business Wire]
Syllo has released a white paper showcasing how its agentic AI document review system achieved breakthrough performance in complex litigation cases across multiple law firms. [Business Wire]
Theorem has launched an AI-powered RFP tool that matches legal departments with vendors and service providers, offering anonymous requests and pricing benchmarks through its marketplace platform. [LawSites]
Vable has launched MyVable, a knowledge management platform that helps legal professionals streamline information access, workflows, and resource management through automated features. [Business Wire]
Verbit has launched Legal Visor, an AI-powered platform providing real-time transcription, analysis, and insights during legal depositions through its proprietary Captivate technology. [LawSites]
Funding
ai|coustics has raised €5m in seed funding to expand its AI-powered platform that transforms ordinary voice recordings into studio-quality audio. [Tech Funding News]
Amazon's Alexa Fund has expanded its investment focus from voice startups to AI companies, backing four new startups while providing them access to Amazon's technology and resources. [TechCrunch]
AmberSearch has secured €2.1m in seed funding to expand its AI-powered enterprise search solution that helps European SMEs streamline data retrieval across multiple platforms. [Tech.eu]
Assista AI has secured €100,000 to advance its AI-driven task automation platform that connects to over 100 productivity applications. [Tech.eu]
Automaise has secured €5m in Series A funding to expand its AI-driven customer service platform across Europe and the US. [Tech Funding News]
Brisk has raised US$15m to expand its AI-powered education platform with tools for detecting AI-generated content and supporting teachers. [TechCrunch]
Browser Use has raised US$17m in seed funding for its technology that helps AI agents better understand and navigate websites by converting them into digestible formats. [TechCrunch]
CaseBlink has raised US$2m in pre-seed funding to develop AI-powered immigration case preparation tools. [National Law Review]
DeepIP has raised US$15m to expand its AI-powered patent drafting solution, which has already helped create 8,500 patent applications and reduced drafting time by 50%. [Tech.eu]
Experial has secured €2m in Pre-Seed funding to advance its AI-powered digital twin technology for transforming market research through simulated consumer behavior analysis. [Tech.eu]
Layer Health has raised US$21m in Series A funding to scale its AI platform that transforms medical chart review by extracting insights from healthcare records. [EIN Presswire]
N8n has raised €55m after its AI-friendly workflow automation platform pivot led to a 5x revenue increase and growth to over 3,000 enterprise customers. [TechCrunch]
Nace.AI has raised US$5m to develop metamodel 1, which generates specialized AI models for enterprise audit and compliance tasks. [FinSMEs]
Navina has raised US$55m in Series C funding to expand its AI-powered clinical intelligence platform across the US healthcare market and enhance its technology. [FinSMEs]
Nexthop AI has launched with US$110m in funding to develop custom networking solutions for major cloud providers' AI infrastructure. [Business Wire]
OpenAI has secured a potential US$40B funding deal led by Softbank, marking one of history's largest private fundraising rounds. [Yahoo Finance]
Page has secured US$4.1m CAD in seed funding to expand its AI-powered government relations platform from Canada into the US and UK markets. [BetaKit]
Paid has launched a platform to help AI agent startups manage billing and performance tracking, securing €10m in pre-seed funding from EQT Ventures, Sequoia, and GTMFund. [TechCrunch]
SplxAI has secured US$7m in seed funding to develop its platform for testing and securing AI systems against security threats. [Business Wire]
Tikos has received investment from Symvan Capital to develop technology making AI systems more transparent and trustworthy for businesses and regulators. [BusinessLive]
Yutori has secured US$15m to develop autonomous AI personal assistants that can perform tasks without human oversight. [Yahoo Finance]
Acquisitions
Kondor AI has acquired Ora Technology in a £20.67m all-share deal to create a unified AI agent marketplace, combining their computer vision tools and trading platform infrastructure. [UKTN]
Nvidia has entered advanced talks to acquire AI server rental company Lepton AI for hundreds of millions of dollars, adding to its recent string of AI-focused acquisitions. [Verdict]
Rev has acquired SmartDepo, combining industry-leading speech-to-text technology with AI-powered deposition summaries to enhance legal document processing for law firms and courts. [LawSites]
Softbank has agreed to acquire chip startup Ampere for US$6.5B, expanding its AI infrastructure investments and adding 1,000 semiconductor engineers to its workforce. [TechRepublic]
Voltage Park has acquired TensorDock, combining their GPU cloud services to expand AI infrastructure offerings and democratize access to high-performance computing solutions. [Business Wire]
Workato has acquired DeepConverse, an AI support automation company, to enhance its platform with AI search capabilities and support automation features. [Business Wire]
There's More
Alibaba has partnered with BMW to develop AI cockpit technology for Chinese market vehicles through its Banma subsidiary and Qwen model team. [Bloomberg]
Audiobook industry leaders at the London Book Fair have expressed growing optimism about AI's potential, despite ongoing concerns about rights and compensation models. [Publishers Weekly]
OpenAI has conducted its first emotional wellbeing study of ChatGPT users, finding that a small subset engaged emotionally with the AI, particularly during extended interactions. [MIT Technology Review]
Synchron has demonstrated its brain-computer interface technology that allows paralyzed individuals to control smart home devices through an Apple Vision Pro headset using neural signals. [eWeek]
Writer's survey of 1,600 executives and employees has revealed widespread AI resistance, with 31% of workers refusing to use AI tools amid concerns over job security and tool quality. [TechRepublic]
The Back Cover
Here's Kacey Musgraves with a cover of Bob Marley's Three Little Birds.
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!