Apple is expected to announce AI implementations into iOS 18 and its other operating systems at WWDC. The company intends to process data from AI applications inside a virtual black box that will use only Apple's hardware to perform AI processing in the cloud. The approach will prevent both Apple and potential hackers from being able to see app data. Using a cloud approach allows Apple to reduce the hardware requirements of its products.
Neuralink is recruiting another three subjects for its brain implant study. It is seeking subjects aged 22 to 75 with severe quadriplegia due to spinal cord injury or amyotrophic lateral sclerosis for at least one year without improvement. Healthy people, people with morbid obesity, or people with an active device already implanted are disqualified from the study. Participants will be monitored for adverse events for 72 months following the implantation procedure.
BYD has unveiled a new hybrid powertrain that can travel for more than 2,000 kilometers without recharging or refueling. Vehicles with the technology can travel from New York to Miami on a single charge and a full tank of gas. The upgraded tech will be launched in the Qin L and the Seal 06. Both mid-sized sedans, which were unveiled at the Beijing Auto Show in April, cost under $13,800.
Software engineers are destined to wallow in unnecessary complexity due to three fundamental laws: a well-designed system will degrade into a badly designed system over time, complexity is a moat filled by leaky abstractions, and there is no fundamental upper limit on software complexity. Building a new system from scratch without succumbing to these laws is a lot harder than it sounds. Engineers who work on badly designed systems suffer more as badly designed systems have unbound complexity.
While the barrier to entry for building AI products has lowered, creating something effective beyond a demo remains a deceptively difficult endeavor. This series of articles identifies crucial lessons and methodologies for developing products based on large language models gathered by people who have been building real-world applications on top of ML systems over the past year. It is organized into three sections: tactical, operational, and strategic. This first part dives into the tactical nuts and bolts of working with large language models and shares best practices and common pitfalls around prompting, setting up retrieval-augmented generation, applying flow engineering, and evaluation and monitoring.
When Meta released its AI, it made the technology freely available, largely the opposite of what Google, OpenAI, and Microsoft have done. Mark Zuckerberg has become the highest-profile technology executive to support and promote the open-source model for AI. While other companies are using a closed strategy to guard their technology, Zuckerberg believes that the technology is so important and the opportunities are so great that it should be made widely available so that everyone can benefit. The decision to open source Meta's AI tech was well-received and made Zuckerberg a lot more popular in tech circles.
Studies from around the world are finding an increased rate of incidences of pathologically stretched eyeballs among children compared with pre-pandemic levels. The pandemic saw real-life classrooms and playgrounds give way to virtual meetings and digital devices, increasing the time that children spend focusing on screens and other nearby objects significantly. This shift has caused their eyeballs to lengthen to accommodate short-vision tasks. An hour of extra outdoor break time every day can markedly reduce the incidence of short-sightedness.
Starlink launched its first 60 satellites into orbit five years ago on May 23 - it has since launched more than 6,500 satellites, making it the largest satellite constellation in service.
Arm's next-generation Cortex-X925 CPU and Immortalis G925 GPU designs will offer optimized layouts to make it easier for device makers to implement their own system-on-a-chip layouts.
Open-source AI will become a financial drain for model builders, an inferior option for developers and consumers, and a risk to national security - closed-source models will create far more economic and consumer value over the next decade.
Adevinta set out to quantify the impact of ChatGPT and GitHub Copilot [4] on its thousands of developers. They measured adoption, accepted lines of code, and self-reported workflow acceleration.
Mistral is working to find a balance between openness and business success. It has a new license aimed at striking that balance. It will continue releasing other projects under Apache 2.0 as well as the new MNPL license.
A new model from the creators of SSMs, Mamba, and sub quadratic Transformer variants. Importantly, their new company Cartesia has trained a lightning-fast, realistic voice generation model. This points to their desire to move into the assistants' space.
OpenAI has signed licensing deals with The Atlantic and Vox Media, allowing their content to train its AI models and be shared in ChatGPT with proper attribution.
Mistral released a 22B code model with 32k context. It is a powerful and fast model with broad performance across many programming languages. It has open weights and is available via Mistral's platform.
PRISM is a foundation model trained on 1.2 billion small molecule mass spectra to enhance mass spectrometry analysis in drug discovery. It uses self-supervised learning to predict molecular properties from complex mixtures without prior annotations, significantly advancing the field of metabolomics. The model improves the identification and understanding of largely unknown natural molecules, accelerating the discovery of new therapeutics by leveraging massive, previously untapped spectral data.
DMPlug is a novel plug-in method that uses pretrained diffusion models (DMs) to solve inverse problems (IPs). Unlike existing interleaving methods, DMPlug views the reverse diffusion process as a function, addressing both manifold feasibility and measurement feasibility effectively.
Scale has made a private leaderboard for evaluating language models. The ordering isn't too surprising, but one interesting finding is the Llama 3 70B is extremely good at following instructions, often better than Claude Opus.
OpenAI has signed PwC as its largest enterprise customer for ChatGPT, covering 100,000 users. PwC is the first partner to resell OpenAI's enterprise offerings.
Gemini has returned $2.18 billion worth of digital assets to its Earn users, comprising ~97% of the assets owed, marking a 232% recovery since Genesis suspended withdrawals. Gemini has achieved an unprecedented recovery in crypto bankruptcies following its settlement with Genesis and other creditors, ensuring its users will eventually recoup all their assets along with any appreciation.
After eight issuers had their 19B-4 filings approved for a spot Ethereum ETF, BlackRock has updated its S-1 application to begin trading its iShares Ethereum Trust. Bloomberg ETF analyst Eric Balchunas believes there could be one more round of comments from the SEC, then a launch around the end of June or early July.
PYUSD just launched on Solana, extending its presence from Ethereum to increase adoption and reduce gas fees and faster transactions. This strategic expansion aims to transition PYUSD from mere awareness to actual utility by incorporating innovative blockchain features that improve payment security and efficiency, enabling its integration into everyday commercial use.
The Ethereum wallet experience is fragmented and complex due to bridging, gas, and dozens of rollups. MagicSpend++ is a framework that allows users to have a chain abstracted balance instead of isolated token balances across chains. It uses account abstraction to create a single-chain experience with no bridging or latency, mimicking a credit card experience where users have their funds pulled by a paymaster later.
Transporter introduces a new level of security in the crypto space, offering Level-5 security, 24/7 customer support, and a user-friendly experience to ensure safe and confident cross-chain transactions. This innovation aims to address the over $2.8 billion previously lost to hacks in token bridges.
The TON blockchain ecosystem is growing rapidly with significant increases in daily and weekly active wallets, total value locked in DeFi, and trading volumes since the start of "The Open League". Ecosystem highlights include large airdrops, notably $NOT coin, and attractive liquidity provider campaigns with APYs up to 105%.
Since memecoins have zero tangible value, unique metrics must be considered to determine their worth. For example, the size of the audience that a meme can appeal to, meme sustainability, and meme visualization are all important for capturing a wide audience of interested buyers. It should be accessible on a chain with liquidity and a first-mover in its category and have potential catalysts and an interesting story or mystery. The main idea is to find a coin's capacity to capture attention.
Layer 2 solutions on Ethereum not only expand its technical capabilities but also foster a diverse and inclusive culture within the blockchain community, leading to unique technological advancements and cultural influences across different blockchains, ultimately driving innovation within the ecosystem.
Pantera Research Lab studied a sample of crypto users and found a high present bias and low discount factor, indicating impatience and a desire for immediate gratification. Crypto users will only wait a short time frame for a high reward, which should be considered when creating a vesting schedule or airdrop design.
This move could lead to broader adoption and innovation within zk-based applications, affecting everything from rollup interoperability to the simplification of zk app development, ultimately impacting the direction of research and optimization in the crypto space.
WhatsApp is developing an AI feature that lets users create profile photos by providing text prompts to personalize and protect user identities. This joins other AI-powered features in the app, alongside a slew of upcoming updates like a revamped Status UI and offline file transfer.
Google is introducing "Minimized Custom Tabs" to its Chrome browser, enabling users to seamlessly switch between web content and native apps via a picture-in-picture window, which enhances app engagement and user retention. This feature is part of Chrome's latest update and aims to improve the web experience on Android.
Discord is refocusing on gaming by redesigning its app to enhance group messages, voice chats, and video calls, aiming to facilitate connections around games. The company is shifting from a broad community chat app to a gaming-centric platform amid changes like staff layoffs and increased regulatory scrutiny.
Design system experts from Bumble, GitHub, and HP discuss leveraging the new Code Connect feature to integrate design and code, highlighting the importance of a shared language and seamless workflow. They emphasize the need for continuous collaboration and the adoption of best practices to maximize the utility of design systems, ensuring consistency and reducing friction between designers and developers.
Unsung design heroes like Larry Tesler, Susan Kare, and Douglas Engelbart created fundamental UI elements. Corporations often prefer to keep their employees anonymous, leaving great product work largely unrecognized. While the best designs usually go unnoticed, the design narrative tends to spotlight individuals over teams. Still, it's important to seek inspiration from those who have tackled similar challenges and examine their strategies.
2024.05.30·
whydesignishard.substack.com·4 min read·TLDR Design
Use near-black and near-white for comfortable contrast, and saturate neutrals with subtle colors to unify your palette. Prioritize high contrast for essential elements, align optically for a natural look, and adjust letter spacing and line height for readability.
Gradientor is a minimalist tool for generating CSS radial background gradients. It allows the quick creation of vibrant, pastel, or monochromatic gradient backgrounds.