While SpaceX's Starlink system for cell phones is still in its testing phase, the technology is already good enough to power a video call. SpaceX recently posted a video that showed a video call being completed through Starlink Direct to Cell satellites from unmodified mobile phones. The company has increased the number of cellular Starlink satellites from six to 38. Starlink tech can deliver download speeds as high as 17Mbps to unmodified Android phones.
Google will soon start testing search and shopping ads within AI Overviews for users in the US. The company says it will display ads in AI Overviews when they're relevant to both the query and the information in the AI Overview. Advertisers already running certain campaigns through Google will automatically become eligible to appear in AI Overviews. It is unclear when exactly the change will roll out.
The focus for the next Starship mission will be to control the reentry of the rocket's Super Heavy booster and Starship upper stage - both vehicles broke apart during their descent on the last flight. The company aims to make the Super Heavy booster make a controlled pinpoint splash down in the Gulf of Mexico as the Starship upper stage attempts to survive its reentry back into Earth's atmosphere. The test will be a repeat of the previous test without the propellant transfer, without the Pez door open, and without the other items that SpaceX was working to demonstrate. NASA's main interest in the test will be seeing that SpaceX replicates the good engine performance it demonstrated on the last two flights. SpaceX is not planning to attempt a Raptor engine restart during the test.
Data from a small clinical trial shows that a zap of electricity from well-placed electrodes on the back of the tech can help patients with tetraplegia regain modest yet potentially life-changing functioning off their hands and arms. The non-invasive procedure offers an accessible and more affordable means for those living with paralysis to regain some meaningful function. The therapy's potential remains limited as it is not yet fully understood why it works. A startup working to make the therapy commercially available has submitted a device to the Food and Drug Administration for review.
Stripe's codebase is mostly written in Ruby and lives in a single large monorepo. It supports a number of services that share code extensively. The company's tooling was built and is maintained by a succession of teams and individuals, with the needs of the environment changing as the team and codebase grew and evolved. This article discusses the features of Stripe's developer environment and explains the reasoning behind the choices made during its development. Maintaining developer productivity as an engineering organization grows is difficult - it is almost inevitable that per-engineer productivity drops as companies and codebases grow.
GitHub Copilot Extensions enables developers to build and deploy to the cloud in their natural language with their preferred tools and services without leaving the IDE or GitHub.com. It helps developers stay in the flow longer, uplevel their skills, and innovate faster. Extensions can be found on the GitHub Marketplace. Organizations can create private Copilot Extensions for their homegrown developer tooling.
Braindump allows developers to build top-down/2.5D games or interactive worlds by simply typing prompts. It generates 3D models, game data, and scripts to make prompts come to life. This article shares learnings from Braindump's development. Challenges in its development include designing a UX for prompting and a game API for large language models. The game API has opened up interesting UX flows like generated macros. While Braindump is pretty good at executing commands, its team wants to expand it to handle more vague and bigger tasks. Braindump will open up for alpha testing in the coming weeks.
Anthropic recently published a public research paper explaining why its AI chatbot chooses to generate content about certain subjects over others. Its researchers deciphered what parts of the chatbot's neural network mapped to specific concepts using a process known as 'dictionary learning'. The research showed how neurons associated with a topic fired together when the model was thinking about something associated with the topic - similar sets of neurons firing can evoke adjacent subjects. A link to the paper is available at the end of the article.
Have you ever wanted to see your own brand in TLDR? Our team already helps companies like AWS, Google Cloud, and Vercel reach an audience of smart tech-savvy professionals like yourself, find out more here [15]!
The StreamSaver bundle, which cuts more than 30 percent off the separate price of certain tiers of each service, will be rolled out on May 29 in the US.
Shoichiro Irimajiri, who was chairman and CEO of Sega, led Sega to invest $5 million into Nvidia, which the GPU manufacturer desperately needed - Sega later sold its Nvidia stock for about $15 million.
The monstrous series F comes from Accel and previous investors. Demand for the services provided is enormous and Scale is uniquely positioned to continue powering the modern AI data wave.
Elon Musk's AI company, xAI, is advancing its Grok chatbot to support multimodal inputs, allowing users to upload photos and receive text-based answers.
This article rounds up all of the major announcements Microsoft just made, including a new lineup of Copilot Plus PCs and generative AI-powered features like Recall, which helps users find apps, files, and other content they have viewed in the past.
Anthropic has made a breakthrough in mechanistic interpretability by mapping millions of concepts inside of Sonnet. It even found that you could push on the internal concepts and change Sonnet's sense of self (e.g., researchers made it believe it was the Golden Gate Bridge).
The Smart Expert System is a new approach using Large Language Models (LLMs) for text classification. This system streamlines the process by reducing the need for extensive preprocessing and domain expertise.
Open source LLMs like Llama 3 and Mixtral are game-changing. But getting them to run efficiently on vLLM requires a ton of manual optimizations. OctoStack [12] from OctoAI achieves better throughput, lower latency, and efficient GPU utilization by baking the latest optimization techniques into an inference stack you control. Read the
This project introduces a novel CNN-based SpatioTemporal Attention (CSTA) method to improve video summarization. Unlike traditional attention mechanisms, CSTA effectively captures the visual significance of frames using a 2D CNN to understand relationships and key attributes in videos.
This project identifies a critical bias in Large Vision-Language Models (LVLMs), where outputs lean more towards language model priors than actual visual inputs. The project effectively reduces this bias by introducing "calibration" and "debias sampling" techniques, leading to more accurate and vision-focused responses across various tasks.
DeepSeek-VL is a new open-source Vision-Language Model tailored for real-world uses with an emphasis on diverse data from web screenshots to charts and OCR.
The departures of Ilya Sutskever and Jan Leike have brought to light OpenAI's restrictive NDA, which prevents former employees from criticizing OpenAI under threat of losing their vested equity. CEO Sam Altman responded to the story by promising a fix.
Apple has announced upcoming accessibility features leveraging AI and machine learning, such as Eye Tracking for iPad, iPhone navigation using the front-facing camera, and Vocal Shortcuts that enable Siri to execute tasks via custom sounds. Other upcoming features include Music Haptics for tactile music feedback and advanced speech recognition for atypical speech patterns.
This Hacker News thread compares GPT-4-Turbo, the default model in ChatGPT Plus, to Claude Opus, Anthropic's competing model. Most developers seem to prefer Claude as it is apparently better at coding and engineering work. Its writing style also seems to be preferred, but part of that is due to the recognizability of GPT-4-Turbo's style. OpenAI's app functionalities, like its code interpreter and ability to search the internet, are still big reasons to continue using ChatGPT Plus.
Automatically & easily clean up your old posts from services like Twitter, Reddit, Facebook, Discord, and more. Use smart filtering, schedule a deletion cadence, or just nuke everything! Try Redact free
Donald Trump, aiming for a second term as U.S. President, has announced that his campaign will now accept cryptocurrency donations, marking a significant shift from traditional campaign financing. Supported by Coinbase, his platform now accepts a wide array of cryptocurrencies, including Bitcoin, Ethereum, and various meme coins, positioning him as the first major party Presidential nominee to embrace such a diverse range of digital assets for campaign contributions.
Uniswap Labs argues that the protocol fully complies with US law and that the decentralized exchange does not fall under the “securities exchange” or “broker” definition. It notes that over 90% of trading volume is on commodities pairs, stablecoins, or out of the country, expanding beyond the SEC's jurisdiction. Lawyers for the Ripple and Grayscale cases are providing outside counsel, and Uniswap is confident it will win if forced to litigate.
Rocketpool, the second-biggest Ethereum liquid staking protocol, is reinventing its tokenomics for better value capture and Ethereum alignment. It has introduced technical improvements to increase security and capital efficiency and will distribute surplus fees from ETH staking to RPL holders. The upgrades are planned to occur around the Pectra hard fork, currently slated for late 2024 or early 2025.
Coinbase is adding retail-sized gold and oil futures to its derivatives platform on June 3. It aims to increase trader retention amidst growing crypto exchange competition. Coinbase Derivatives launched in June 2023 and has grown to host fifteen perpetual futures products.
Researchers have identified key factors in airdrop design through a comprehensive analysis of over 2 million airdrop events across 40 protocols, such as the quantity of tokens distributed and the target recipients. Their findings suggest that airdrops targeted at core users with smaller token quantities tend to show higher price stability and engagement post-distribution, challenging the effectiveness of widespread, larger airdrops.
The quality of token launches has declined significantly due to new stringent exchange requirements demanding high user metrics prior to listing, which has led projects to focus on generating inflated user activities through point systems rather than organic growth. This shift has resulted in an airdrop farming culture where users engage not out of genuine interest but to earn rewards, leading to rapid sell-offs post-launch and damaging the overall market dynamics and the perceived value of new tokens.
The U.S. DOJ has arrested Rui-Siang Lin, the head of "Incognito Market", an illegal dark web marketplace that facilitated the sale of illegal drugs using mainly crypto (including bitcoin and Monero). The marketplace processed ~$80M in crypto transactions over three years and charged a 5% fee on drug purchases.
This article discusses the complex financial maneuvers used by major global economies, focusing on how the Federal Reserve, U.S. Treasury, and Bank of Japan manipulate the dollar-yen exchange rate to manage their economic and political challenges. It explains that by conducting dollar-yen swaps, these entities can artificially strengthen or weaken currencies to impact their economies, forestall fiscal issues, and influence global trade dynamics, particularly affecting the competitive balance between China and Japan in global markets.
Adobe is bringing even more Firefly-powered artificial intelligence tools to Adobe Lightroom, including Generative Remove and AI-powered Lens Blur. Generative Remove is particularly impressive as it lets users remove unwanted objects from any photo entirely non-destructively with just a single click.
Microsoft Paint's Cocreator is an AI-powered tool that generates images from text prompts and doodles using diffusion-based algorithms. The feature, demonstrated at the Surface event, includes a creativity slider for balancing AI and user input, offering quicker edits and precise renderings.