Google Gemini To Power iPhone

PLUS: YouTube's New AI Content Labeling, DHS Launches AI Pilot Programs and more.

In partnership with - Smart Browsing with 1-Click AI—where AI meets your browser. Instantly enhance web browsing with article summaries, email crafting, and AI-powered web searches. Ranked #1 of the day and the week on ProductHunt.


  • Google Gemini To Power iPhone

  • AI-Powered Tools Transform Roblox Creation

  • Nvidia Unveils Blackwell B200

  • YouTube's New AI Content Labeling

  • OpenAI's ChatGPT App Store Struggles

  • DHS Launches AI Pilot Programs

Apple Is in Talks to Let Google Gemini Power iPhone AI Features

Apple is in talks with Google to integrate Google's AI engine, Gemini, into the iPhone, potentially revolutionizing the AI industry. The negotiations aim to license Gemini for new iPhone features in 2024. Apple also explored a deal with OpenAI. This move could significantly impact iPhone capabilities, enhancing AI-powered functions. 

The discussions remain confidential, but if successful, this partnership could reshape the AI landscape. The agreement builds upon their existing search pact, where Google pays Apple billions annually to be the default search engine on iPhones. This development underscores the escalating competition and collaboration in the AI domain.

Roblox introduces AI-powered tools to boost 3D content creation

Roblox, at the Game Developers Conference, revealed Avatar Auto Setup and Texture Generator, employing AI to turbocharge 3D content creation. Avatar Auto Setup automates avatar formation from 3D body meshes, slashing setup time. Texture Generator allows quick object texture changes via text prompts, streamlining design. 

Nick Tornow, VP at Roblox, hailed these tools as game-changers, boosting creativity and engagement. Industry experts lauded Roblox's AI strides, foreseeing limitless creative possibilities. Roblox's AI journey also includes Code Assist and chat translations, amplifying efficiency and global connectivity. These innovations reflect Roblox's commitment to staying ahead in the user-generated content landscape.

Nvidia reveals Blackwell B200 GPU, the ‘world’s most powerful chip’ for AI

Nvidia unveils the Blackwell B200 GPU, touted as the "world's most powerful chip" for AI. With 20 petaflops of FP4 horsepower from 208 billion transistors, it promises 30 times the performance for LLM inference workloads and up to 25x cost and energy savings over its predecessor, the H100. Key features include a second-gen transformer engine and a next-gen NVLink networking solution enabling seamless communication between GPUs. 

Nvidia aims to integrate these chips into larger supercomputer-ready designs like the GB200 NVL72, set to be offered by major cloud service providers. The company also offers comprehensive solutions like the DGX Superpod for DGX GB200.

YouTube adds new AI-generated content labeling tool

YouTube Dark Mode 3D icon concept. Write me:, if you need 3D visuals for your products.

YouTube introduces a self-labeling tool for creators to mark videos with AI-generated or synthetic content. Creators must disclose realistic alterations like making someone say or do things they didn't, altering real events, or creating fake scenes. However, beauty filters and obvious animations are exempt. YouTube aims for transparency after previously outlining AI content policies, especially concerning deepfake music. 

While the platform invests in AI detection tools, the labeling system relies on creators' honesty. YouTube may add labels independently for potentially misleading content. The update emphasizes transparency in content creation amid concerns about AI manipulation.

OpenAI’s Chatbot App Store Is Off to a Slow Start

ChatGPT Logo In 3D. Feel free to contact me through email

OpenAI's bid to launch a ChatGPT app store akin to Apple's iPhone success faces a slow start. CEO Sam Altman's vision aimed to empower developers to create and sell AI-powered chatbots. However, four months in, developers report disappointment due to low customer traction and inadequate support from OpenAI.

The store, limited to ChatGPT paying users, also prohibits developers from charging for apps. Despite initial hype, practical challenges hinder the store's success, echoing Apple's subtle push for AI research. As OpenAI navigates these hurdles, its ambitious venture faces the realities of marketplace adoption and developer engagement.

Homeland Security is testing AI to help with immigration, trafficking investigations, and disaster relief

The Department of Homeland Security (DHS) initiates three $5 million AI pilot programs with OpenAI, Anthropic, and Meta. These programs aim to aid various tasks like investigating child exploitation, streamlining disaster relief, and training immigration officials. The pilots involve using generative AI for hazard mitigation planning, large language models for data analysis in investigations, and chatbots for training asylum officers. 

DHS faces challenges such as ensuring reliable data usage and avoiding AI-generated false results. While DHS has prior AI initiatives, these programs mark a significant expansion into AI integration for internal operations, differing from CBP's border-focused AI surveillance efforts.


Uni-SMART, a new model for analyzing scientific literature, tackles the challenge posed by the vast amount of scholarly articles. Unlike traditional text-focused models, Uni-SMART handles multimodal content like molecular structures and charts effectively. It outperforms existing models and shows promise in applications like patent infringement detection, potentially transforming scientific research.

VideoAgent is a new approach to understanding long-form videos, inspired by human cognitive processes. It utilizes a large language model as a central agent to iteratively gather crucial information, aided by vision-language models. Tested on challenging benchmarks, it outperforms current methods in both accuracy and efficiency, showcasing the promise of agent-based strategies in video comprehension.

The paper presents a novel method, Recurrent Drafter, to improve speculative decoding in large language models. It combines the efficiency of single-model approaches like Medusa with a lightweight draft head utilizing recurrent dependency design. Empirical results demonstrate its effectiveness across various language models, offering insights into its trade-offs.

RAFT is a training method enhancing Large Language Models (LLMs) for domain-specific tasks. RAFT enables models to selectively use retrieved documents for answering questions, ignoring distractors, and citing relevant sequences to improve reasoning. Tested on various datasets, RAFT consistently boosts LLMs' performance, providing a valuable post-training recipe for domain-specific tasks.

Alignment Studio is an architecture empowering developers to align large language models with specific contextual regulations, values, and norms. It comprises three components: Framers, Instructors, and Auditors, working together to control model behavior. Illustrated through aligning a chatbot with business conduct guidelines, it offers a framework for navigating contextual requirements.


Spoke - Summarizes missed Slack messages

GovernGPT - Helps asset managers modernize fundraising and investor relations by automating investor updates like DDQs and RFPs.

Sonia - AI-powered cognitive behavioral therapy (CBT) 

Diligent - Automating customer due diligence and compliance tasks with AI

Melon - Personal AI learning companion



Microsoft to hold a special Windows and Surface AI event in May

Microsoft plans a special event on May 20th to unveil its AI strategy for Windows and Surface devices, led by CEO Satya Nadella. The focus is on integrating AI into upcoming Surface hardware and Windows 11, including dedicated AI processors. The event sets the stage for Microsoft's "year of the AI PC." THE VERGE

Nvidia Backs Little-Known Upstart in India’s Biggest AI Bet Yet

In Mumbai, Yotta Data Services eagerly awaits a truck carrying Nvidia semiconductors, crucial for advancing artificial intelligence (AI). While AI investments soar globally, Yotta's acquisition marks India's largest AI commitment yet. The company's move underscores the escalating global AI race, with Nvidia chips at the forefront of innovation. BLOOMBERG

Qualcomm debuts latest Snapdragons 8s AI chip for Android devices

Qualcomm unveils Snapdragon 8s Gen 3 chip for Android smartphones, emphasizing generative AI capabilities for more affordable devices. The chip supports multimodal generative AI on device and can handle large language models up to 10 billion parameters. Shares rise 2% in pre-market trading amid positive market reception. YAHOO!

Adobe Substance 3D’s AI features can turn text into backgrounds and textures

Adobe introduces AI-powered features in Substance 3D Suite, enabling users to create textures and backgrounds from text descriptions. The Text to Texture tool generates textures like scaled skin or woven fabric, while the Generative Background tool creates background images for 3D scenes. Both features use 2D imaging technology, available in beta versions for free. THE VERGE

Productiv launches Sidekick, an AI-powered assistant for smarter SaaS management

Productiv launches Sidekick, an AI-driven chatbot, enhancing SaaS management by providing insights, recommendations, and data analysis through natural language queries. Integrated with Productiv's Spend Management Platform, Sidekick helps optimize software spend, analyze usage, and facilitate license negotiations. The launch addresses the growing need for efficient SaaS management amid increasing cloud-based software adoption. VENTUREBEAT

Nvidia GTC 2024: What to expect from the AI giant’s big conference

Nvidia gears up for its GTC conference with high expectations after a phenomenal year, with CEO Jensen Huang set to unveil potential game-changers like the Blackwell architecture and B100 GPU. Nvidia aims to maintain dominance in AI with advancements in its CUDA software and possibly a new H20 AI chip for the Chinese market. YAHOO!

Data for AI Could be Hacked. This Startup Wants to Protect It.

BigID, a cybersecurity startup, secures $60 million in funding led by Riverwood Capital, reaching a $1 billion valuation. With concerns rising about data security for AI systems, BigID aims to safeguard information crucial for generative AI, reflecting a growing consensus on the need for simultaneous development of AI and data security. THE WALL STREET JOURNAL

SoundHound AI in Focus as Nvidia's GPU Tech Conference Kicks Off—Key Level to Watch

SoundHound AI (SOUN) saw a remarkable 52% surge last week, with Nvidia's GPU Technology Conference on the horizon. Since Nvidia's disclosed stake in SoundHound, its shares have tripled. While not presenting at the conference, SoundHound's explosive performance and Nvidia's backing keep it in focus. Watch for potential volatility around the event, considering the stock's recent rally and short interest. INVESTOPEDIA

Tech Job Seekers Without AI Skills Face a New Reality: Lower Salaries and Fewer Roles

Tech job seekers without AI skills are facing a new reality: lower salaries and fewer roles. The IT job market is expected to shrink by 20,000 to 30,000 jobs in 2024, with AI accelerating automation. Salaries for software developers have declined for the first time since 2019, reflecting changing market demands. Jobs related to generative AI are in high demand, commanding the highest salaries. As the corporate shift towards AI continues, job seekers without AI knowledge may find themselves at a disadvantage. THE WALL STREET JOURNAL

What'd you think of today's edition?

Login or Subscribe to participate in polls.

What are MOST interested in learning about AI?

What stories or resources will be most interesting for you to hear about?

Login or Subscribe to participate in polls.

Join the conversation

or to participate.