TLDR 2024-03-27

Apple WWDC šŸ“±, Microsoft AI PC rules šŸ’», why Google didn't make GPT-3 šŸ¤–

šŸ“±
Big Tech & Startups

Microsoftā€™s new era of AI PCs will need a Copilot key, says Intel (3 minute read)

Microsoft's new specifications for OEMs building AI PCs includes a requirement that they must have a Copilot key. To be recognized as an AI PC, OEM partners must also provide a combination of hardware and software including a Neural Processing Unit (NPU), the latest CPUs and GPUs, and access to Copilot. It is unclear what OEMs will get in return for adhering to Microsoft's AI PC definition. Intel recognizes some laptops without the Copilot key as AI PCs due to their integrated NPUs.

Appleā€™s WWDC 2024 is set for June 10th (2 minute read)

Apple's Worldwide Developers Conference will start with a keynote on June 10 and run through until June 14. This year's event should be similar to previous years, with announcements likely focusing on the company's fall software updates and new hardware. Apple has reportedly spent millions per day training its own AI models and it has been rumored to be courting news outlets for training content partnerships. The company may be planning a deal with third parties to supply cloud-based AI features. It may also be planning to open up an ecosystem for AI developers to create deep integrations into its devices.
šŸš€
Science & Futuristic Technology

Startup Plans Largest Ocean Geoengineering Plant (5 minute read)

Californian startup Ebb Carbon plans to use seawater to remove tons of carbon dioxide from the air. Its pilot carbon dioxide removal plant, codenamed Project Macoma, could start operations as soon as this year. It will pump hundreds of thousands of liters of seawater every day, splitting the water into acidic and alkaline streams using an electrochemical process. The alkaline outflow mixes with carbon dioxide in ambient seawater to create bicarbonate, a stable way to store carbon. While the impact of the plant will be small, the process, if scaled up, could help mitigate the effects of climate change.

It's hearty, it's meaty, it's mold (5 minute read)

One of the most promising sources of alternative foods is fungi, which contain nutritious proteins, fats, antioxidants, and flavor molecules. Scientists are exploring the many possibilities for new flavors and textures that can be made by modifying genes already present in fungi, as opposed to introducing genes from wildly different species. Researchers have modified a fungus called koji mold, usually used in East Asia to ferment starches, to be more meat-like, offering an alternative route for creating meat substitutes. The approach can create meat substitutes without the complex ingredients lists, costs, and technical difficulties of cultured meat.
šŸ’»
Programming, Design & Data Science

Why Google failed to make GPT-3 + why Multimodal Agents are the path to AGI ā€” with David Luan of Adept (64 minute read)

This article contains an interview with David Luan, one of OpenAI's early hires, a past leader of Google's LLM efforts and co-leader of Google Brain, and founder of Adept, one of the leading companies in the AI agents space, where he discusses his time with early OpenAI and how Adept is building agents that can do anything humans can do on a computer. Google had a huge lead with AI in 2017, but it was OpenAI that ended up making GPT 1/2/3. While Google's team created Transformers, the company's internal processes made it difficult for its researchers to get work done. OpenAI was able to beat Google because it took big swings and focused.

What Computers Cannot Do: The Consequences of Turing-Completeness (32 minute read)

This article helps readers understand the limits of computers and what computers cannot do. Most programmers seem to not understand these hard limits or what they mean. Understanding Turing-completeness and its consequences for our finite machines, which are technically not Turing-complete, is important for programming.
šŸŽ
Miscellaneous

Hereā€™s why AI search engines really canā€™t kill Google (11 minute read)

Google Search seems to be getting worse as AI tools get better. Google is complex and people use it for many different things. Large language model-based bots may be genuinely more useful than a page of Google results in some cases, but it will be difficult for AI to replace Google as the center of the web for most use cases. This article looks at how search works and explains how Google is still much better than LLM-based systems for the majority of queries.

How to trade software for small money? (4 minute read)

It's easy for system engineers to make something that is valuable and widely used, but quite a lot harder to get paid for it. The competition is often free software maintained by well-funded teams and there is a community norm where it is seen to be mildly immoral to charge money for software. Pricing models for systems software is also tricky. This post looks at various options for pricing models. SaaS software is likely more prevalent as VCs don't think that non-SaaS companies make money.
āš”
Quick Links

Robinhood launches 3% cash back credit card (2 minute read)

The Robinhood Gold Card, which offers 3% cash back on all categories, doesn't have any annual fees but will require a Robinhood Gold membership, which costs $5 per month or $50 annually.

Facebook snooped on usersā€™ Snapchat traffic in secret project, documents reveal (4 minute read)

Facebook had a secret project that intercepted and decrypted Snapchat network traffic to understand user behavior and help Facebook compete with Snapchat.

Thoughts on Apple Vision Pro (10 minute read)

A review of the Apple Vision Pro from the developer who built the largest open-source projects for both of Apple's AR frameworks: ARKit and RealityKit.

Ask HN: What non-AI products are you working on? (Hacker News Thread)

Projects discussed in this thread include a synthetic DNA assembly company, a Christian app and tool discovery and distribution platform, and a trading terminal tailored to scalping traders, prop firms, and brokerages.

A native version of Chrome arrives for Arm-based Windows PCs (2 minute read)

Chrome users with Windows machines powered by Arm-based processors will have access to a much faster native browser starting this week.
Get the most important tech, science, & coding news in a free daily email. Read by +1,250,000 software engineers and tech workers.
Join 1,250,000 readers for