Meta won't offer future multimodal AI models in EU (3 minute read)
Meta will withhold its next multimodal AI model from customers in the EU because of a lack of clarity from regulators. Apple has also said it won't release its Apple Intelligence features in Europe because of similar concerns. Meta plans to incorporate its new multimodal models in a wide range of products, including smartphones and its Meta Ray-Ban smart glasses. The upcoming larger text-only version of Meta's Llama 3 model will be available in the EU.
'Supermodel granny' drug extends life in animals (6 minute read)
Researchers have discovered a drug that increases the life spans of laboratory animals by nearly 25%. It works by lowering levels of a protein called interleukin-11, which is associated with higher levels of inflammation and other factors that control the pace of aging. Mice with decreased interleukin-11 had lower levels of cancer, improved muscle function, were leaner, had healthier fur, and scored better on many measures of frailty. The drug is already being tested in people, but it is unknown whether it will have the same anti-aging effect.
The Pitfalls of In-App Browsers (9 minute read)
In-app browsers have limited functionality, privacy and security concerns, inconsistent UI/UX, and worse performance than normal browsers. Businesses use in-app browsers to perform man-in-the-middle attacks by injecting arbitrary JavaScript code and intercepting network traffic. This allows them to watch what users are doing within their in-app browsers. Websites can detect when in-app browsers are being used and offer links to users to open them in the default browser instead.
Panic! at the Job Market (72 minute read)
Too much tech speculation tied to speculative funding caused the funding and customers to go away, and now the jobs are gone.
OpenAI debuts mini version of its most powerful model yet (2 minute read)
OpenAI has launched a new AI model called GPT-4o mini. It is available to free users. The 'o' in the name stands for 'omni', hinting at the model's audio, video, and text capabilities. The model will be made available to ChatGPT Enterprise users next week.
A fault with an update issued by CrowdStrike led to worldwide outages on Friday. CrowdStrike is a cybersecurity vendor that develops software to help companies detect and block hacks. It uses cloud technology to apply cyber protections to internet connected devices. The software requires deep access to systems to scan for threats. A fix has been issued, but it could be hard to implement - engineers will have to go into each individual data center running Windows to apply the fix, entering complex encryption keys manually for encrypted machines.
AI paid for by Ads – the gpt-4o mini inflection point (4 minute read)
With OpenAI's recent introduction of GPT-4o mini, it's now possible to build dynamic, AI-generated content entirely supported by ads. Google's AdSense pays approximately $0.0026 per page view. A site could generate results tailored to user search queries and earn a profit after calculating for the cost of the generated content. This post presents a simple example of how this could be done with calculations of how much it would cost/profit.
The rise and fall of software developer jobs (1 minute read)
Demand for software developers has plunged in the last couple of years.
Alexa Is in Millions of Households—and Amazon Is Losing Billions (13 minute read)
While hundreds of millions of customers own Alexa-enabled devices, Amazon's bet that people would spend meaningful amounts of money to buy goods on its platform by talking to the voice assistant hasn't paid off. Customers mostly use Alexa to set alarms and check the weather. Amazon has lost tens of billions of dollars on its devices business. The company will launch a paid tier for Alexa as soon as this month as part of the plan to reverse these losses.
Meta releases the biggest and best open-source AI model yet (7 minute read)
Meta has released Llama 3.1, its biggest open-source AI model yet. The company claims it outperforms GPT-4o and Anthropic's Claude 3.5 Sonnet on several benchmarks. The largest version has 405 billion parameters and was trained with over 16,000 Nvidia H100 GPUs. Meta is making its AI assistant available in more countries and languages and adding a feature that can generate images based on someone's likeness.
Inside the 78 minutes that took down millions of Windows machines (5 minute read)
Businesses around the world experienced outages starting shortly after midnight on Friday. This was caused by a faulty update to CrowdStrike's Falcon security software, which is used widely in Windows machines to prevent cyber threats. The update exposed a massive flaw in the company's product that could have been easily avoided. The issue could have been prevented if CrowdStrike had properly tested its updates with a small group of users and rolled out the update gradually, a fairly standard practice in the industry.
Wiz turns down $23 billion Google deal, says it's aiming for IPO (3 minute read)
The company's next target is to reach $1 billion in annual recurring revenue and to go public.