Javier Ruiz Pérez’s Post

Why did Meta invest so much in Llama3(and previous versions) and open source it? Llama3 compares well with an earlier version of Gpt4 on some benchmarks and outperforms Gpt3.5 on many. A conversation on the bird app, pointed me to a good resource which might help explain why. HT to Niran Babalola for sharing this resource https://lnkd.in/gZucZUaV It is a good read, and here is my take on common perspectives on why they are doing this Since Meta makes money from advertising, and advertisers pay for people engaging/clicking on their ads --> Meta needs more people, or more engagement/clicks from the existing user base. People come to meta to fullfill social and informational needs. More content(from genAI tools) certainly helps, though the jury is out if platforms will get swamped with generic AI generated content. Its possible some of this multi media content that users generate might help with their foray into VR. Incidentally, they also just open sourced their OS for the headsets 🤔 By releasing a good enough model for free they are effectively 'commoditizing their complement'. A complement being some product or service which is usually consumed along with the main product or service and adds value. Hence, if a firm can drive the price of that complement to zero, it can increase consumption of its product/service. In this case, the complement would be commercial LLM's such as OpenAI, which are attracting user attention and being used for content generation. Similarly, one can compare this move to what Google did when open sourcing Android to counteract Apple's IOS platform. Another interesting perspective from Emad mustaqe of Stability AI fame, discusses the value in having the community build on Llama3 and improve it. By his estimates, Meta is using about 10% of its compute to train Llama models, and if the community improves the model by 10%, Meta got a good return on investment. Its not clear if the cost of 10% compute and 10% improvement in the models is equivalent, though meta is also using llama to power AI functionality across interfaces under the Meta.ai banner (https://lnkd.in/gUCM-Pg5) Finally, it could also help Meta attract more developers who see this openness as an attractive option. This openness also means that users/developers will innovate on top of the base models and release many more variants and Meta gets more visibility. It is has been six days since launch and there are over 3500 variants (as on date) based on the two models (two versions each) released. See for yourself at https://lnkd.in/gBVBBCR6

Exciting news from Microsoft as they restructure Windows and Surface divisions, prioritizing AI integration and consumer-focused products. #Microsoft #AI #Innovation • Microsoft has reorganized its Windows and Surface divisions, reuniting them under a single leader, Pavan Davuluri, who will oversee both hardware and software development. • The move comes after a previous shake-up that separated Windows and Surface teams, with the intention of enhancing focus on web and AI features. However, this approach faced challenges and frustrations within the organization. • Microsoft has hired Mustafa Suleyman, the co-founder of DeepMind, as the new CEO of Microsoft AI, indicating a shift in strategy and a renewed emphasis on consumer AI products like Copilot, Bing, and Edge. • The company aims to provide a clearer direction for AI integration in Windows and explain the significance of "AI PCs" to consumers, moving beyond marketing terms and highlighting practical benefits. • Microsoft is expected to unveil Arm-powered versions of the Surface Pro 10 and Surface Laptop 6, potentially signaling a significant shift in its relationship with Intel and a transition towards Arm-based processors for Surface devices. check out this link from The Verge: https://lnkd.in/gNH_G7kR

What caught my eye this week… $3 trillion club Nvidia just hit a market value of $3 trillion, surpassing Apple. Now, it's the second most valuable company in the world, joining the exclusive $3 trillion club with Microsoft and Apple. What's really impressive is how fast Nvidia got there. It took just over three months since they hit $2 trillion, while Apple took nearly three years and Microsoft took about 2 ½ years to make the same leap from $2 trillion to $3 trillion. Moravec’s landscape of human competence Hans Moravec uses a great metaphor, comparing human skills to a landscape. He explains that simpler tasks like arithmetic and memorization are in the lowlands, while more complex abilities like playing chess and proving theorems are in the foothills. The highest peaks are things like walking, hand-eye coordination, and social interaction. Moravec describes the progress of computer capabilities as a rising flood that first covered the lowlands, then moved to the foothills, and is now getting close to the peaks. Early on, when AI started handling simple tasks, people were amazed, but as computers began taking on more complex skills, we started seeing the real challenge. Moravec suggests that we should prepare for a future where machines might match humans in nearly every area, and we need to be ready to adapt to these changes. Wining Gen Z Facebook is trying to shake off its "for old people" reputation and attract more Gen Z users by focusing on practical features like Marketplace, Groups, and Dating. It's not about being "cool" but about being useful for young adults. The marketplace is already popular with young bargain hunters, and Facebook is boosting its short video content to compete with TikTok and Instagram Reels. While there has been some progress (40 million daily active users aged 18 to 29 in the US and Canada) Facebook still has a hard time with teen engagement. Only 33% of teens use it, compared to much higher numbers for Instagram and TikTok.

5G + Edge + GenAI = New Business Model opportunities Meta has released Llama 3.2 lightweight models in 1B and 3B sizes available on #watsonx --> advantages: > Small size is perfect for mobile and edge devices or CPU only inferencing. > Optimized for ARM based processors, support Qualcomm and MediaTek hardware from day one (covering majority of mobile processors out in use). > Enhanced Privacy - No need to send sensitive data such as messages and calendar invite to the cloud, maintaining user privacy. > Supports up to 128K tokens and optimised for summarisation, rewriting, instruction following, language reasoning, and tool use.  > Tool Calling: The 1B and 3B models excel in tool calling, enabling developers to create personalized, on-device applications. > Instantaneous Responses: Local processing makes prompts and responses feel instantaneous. > Tiny size achieved through network pruning and knowledge distillation using larger Llama 3.1 > Post-training step uses supervised fine-tuning (SFT), rejection sampling (RS), and direct preference optimization (DPO) for alignment. Example Use Case: Integrated Applications that could help summarize the last 10 messages received, extract action items, and leverage tool/function calling to directly send calendar invites for follow-up meetings. RIGHT MODEL --> RIGHT USE CASE --> RIGHT GOVERNANCE --> Success !   

Microsoft Enhances Copilot AI: A Glimpse into the Future of Computing Microsoft has unveiled major updates to its Copilot AI for Windows, promising a more intelligent and interactive experience. The new features include direct communication with the AI, a Click to Do option that provides context-aware tools, and enhanced photo editing capabilities. Imagine erasing or adding objects to your images or upscaling your old photos to 8K resolution effortlessly. One standout feature is Copilot Vision, an AI that understands and responds to questions about your Edge browser screen. During a demo, a user asked for TV show recommendations while browsing, and Copilot provided tailored suggestions based on what was visible on the screen. Alongside Copilot Vision, the updated Windows Search allows users to find files using descriptive queries instead of exact names. Need a file about "new Microsoft updates"? Simply type it in, and Search will fetch the relevant document. Ahead of the holiday season, Microsoft's AI advancements aim to boost Copilot+ PC sales. This move comes as Apple prepares to introduce Apple Intelligence on its Mac computers, setting the stage for an AI rivalry. With these innovations, Microsoft is not just pushing technological boundaries; it's reimagining the way we interact with our devices. #MicrosoftAI #TechInnovation #DigitalTransformation

I have a dream... 😉 🤓 Yes - I have been dreaming for years of the OS on my device to be a mind-map-like fabric that is woven for tying everything with anything. It would allow me to search, discover, connect existing content, and then call in #AI to assist with transformation and synthesis of new content. Allowing to invite seamlessly for sneak peak, for collaboration my close ones and teams on the stuff in secure ways. Is it coming true at last? Will LLMs/edge LMs be the solution to this task that has been elusive so far? Eager to see what this will look like in reality: https://lnkd.in/gq8C9eCh Related: I am also very eager to see what Apple is brewing. My belief was that they might be late to market, but their catch-up act will be overwhelming with user friendliness. It has been their strong suite over Microsoft and others in many areas for many years. Let's see what comes out here: https://lnkd.in/gjc5qxkS And never discount Google: Gemini integrations in Google Sheets are far cooler and easier to use for free user than Microsoft alternatives. Nicholas Thompson - do you think this might be another "most interesting thing in tech"? #mostinterestingthingintech Planning to buy a new laptop this fall. What's it gonna be...? Will it sport the #mindfulAI I'm so much in need for? 🤓 🖖

I personally am extremely bullish on Apple because I believe their unique ecosystem provides them a *major* advantage in this new era of ‘personalized Ai.’ Personalized Ai thrives on personalized data, and if you want personalized Ai, then you need personalized data—the new oil of the digital age. Apple’s integrated ecosystem—spanning hardware like iPhones, Apple Watches, and their own custom processors like the M1 and A-series chips, along with services like iCloud and Siri—enables seamless and secure collection of user data. This comprehensive ecosystem is unmatched by competitors like Microsoft, Meta, and others who face challenges because of their limited access to personalized data. Specifically because they're not in the smartphone market. Hence the era of strategic (and unusual) partnerships. In today’s mobile-first world, much of our daily activity is conducted through our phones. While desktops with Windows can support work-related Ai applications, they are outmatched in a mobile-centric era where smartphones are the primary devices for personal and professional tasks. This mobile dominance allows Apple to collect and utilize data more effectively, offering personalized experiences that competitors like Microsoft, Meta, and others cannot replicate. I personally think this is why Microsoft introduced their Recall feature as part of the Copilot+ initiative; to get a chance at capturing any type of personalized data through their existing apps just to stay in the personalized Ai race. Microsoft's Recall works by scraping data and analyzing screenshot images across various Microsoft applications, like Word and Excel, to infer user activity. This approach attempts to compensate for Microsoft's lack of a cohesive mobile ecosystem. However, this method raises major privacy and security concerns since it involves constantly capturing screen content, which could include sensitive information. Additionally, Microsoft has no way to trigger actions in your apps and programs, whereas Apple’s AI uses app intents in a seamlessly integrated manner to trigger actions across various applications without these privacy risks. Apple’s cohesive and secure approach ensures a smooth user experience, giving them a significant edge in the personalized AI space – do you agree?

Microsoft CEO Satya Nadella unveiled a new line of #laptops designed to leverage #AI with the goal of building, "computers that understand us versus us having to understand computers.” The new PCs further Microsoft's partnership with OpenAI as the industry accelerates its moves toward integrating AI #technology such as #ChatGPT and #GPT4o. https://lnkd.in/erb7kXaA

Under Satya Nadella's leadership, Microsoft continues to excel, demonstrating a commanding execution of their AI strategy and its market integration. The unveiling of Copilot+ PCs has sparked curiosity about Apple's upcoming announcements at WWDC this summer. Copilot+ PC, an ARM-based SoC, encompasses a CPU, NPU, and GPU, akin to Apple's M series chips. Microsoft asserts a 58% performance increase over the M3 MacBook Air and boasts a 22-hour battery life. This innovation aligns with their vision of "AI at the edge," emphasizing local, on-device processing for daily tasks. Despite Apple's four-year lead with potent neural processing units, compelling AI features remain scarce, limited to niche third-party applications or subtle machine learning tasks, such as Siri's email-based calendar suggestions—contrasting sharply with Microsoft's Copilot+ PC offerings. With Recall, semantic searching becomes intuitive; your PC recalls past interactions, discerns intent, and aids in future searches. Despite potential privacy concerns, Microsoft assures that all processes are local, encrypted, and secure. The Cocreator feature transforms rudimentary sketches into complete concepts, functioning similarly to DALL-E but utilizing the sketch as the prompt instead of text. Additionally, the system can instantly translate up to 40 languages into English during video calls, using AI. Microsoft has set a high standard, and anticipation builds for Apple's response with iOS 18, macOS, and iPadOS updates. The iPad Pro serves as a reminder: despite its unparalleled hardware, the software has yet to match its potential. #ai #pc #hardware #apple #microsoft https://lnkd.in/gPrGh8-i

Here is a glimpse into the future of computing, Microsoft's Copilot+ PC announcement heralds a new era of performance and innovation. 🚀 Here are some key highlights: ❇ Revolutionary Performance: The Copilot+ PCs deliver an astonishing 40+ TOPS, providing unparalleled speed and intelligence for any task. ❇ All-Day Battery Life: Enjoy up to 22 hours of video playback or 15 hours of web browsing, setting a new standard for efficiency. ❇ Advanced AI Models: Equipped with the latest AI models, these PCs offer localized, real-time capabilities that enhance productivity and creativity. ❇ Recall Feature: Instantly retrieve information with a digital photographic memory, making data access intuitive and immediate. ❇ Cocreator: Integrated into Windows, the near real-time AI image generation and editing redefine the limits of digital artistry. ❇ Live Captions: Overcome language barriers with live translations from 40+ languages into English, available even when offline. ❇ Windows Studio Effects: Automatically adjust your image and sound during virtual interactions for a professional online presence. ❇ Personal AI Companion: A simple tap of the Copilot key activates a powerful AI assistant, simplifying complex tasks with ease. 🎥 Watch the video here: https://lnkd.in/e_kEt5UW 📥 Read the entire blog here: https://lnkd.in/erANncyS Can't wait to try all this magic! ✨