- AIdeations
- Posts
- Meta’s Bold 2024 Play: AR, VR, AI & The Big Shake-Up at OpenAI
Meta’s Bold 2024 Play: AR, VR, AI & The Big Shake-Up at OpenAI
From Meta’s ambitious AR and VR updates to a wave of high-profile departures at OpenAI, today’s newsletter explores the latest moves in tech, AI-driven business growth for small companies, and how AI is transforming both cybercrime and cybersecurity.
Brent Moreno
September 26, 2024
Meta’s Big 2024 Play: New AR, VR, and a Flood of AI Across Everything
Quick Byte:
Meta Connect 2024 just dropped, and while Zuck isn’t throwing around the word “metaverse” like he used to, Meta’s big tech ambitions are still alive and kicking. From a budget VR headset to prototype AR glasses and a whole lot of AI being shoved into their apps, Meta’s message is clear: they’re doubling down on the future—whether we’re ready or not.
Key Takeaways:
Meta’s Quest 3S—Cheaper VR for the Rest of Us
If the $500 price tag of the Quest 3 gave you sticker shock, Meta’s got you. The Quest 3S starts at $300 but with some trade-offs: worse displays and no depth sensor for hand tracking. It still packs the same chip and RAM as the Quest 3, so it’s no slouch—and it’ll run the entire Quest game and app library. As a bonus, if you buy now, you’ll get the upcoming Batman: Arkham Shadow game for free.“Orion” AR Glasses—Meta’s First Real Play at Augmented Reality
Meta finally revealed its prototype AR glasses, codenamed “Orion”. They’re bulky, but the tech is impressive: a single waveguide screen, hand-eye tracking through seven sensors, and a wristband for gestures. It’s still early, but these glasses are Zuckerberg’s long-awaited metaverse dream starting to take shape in real life.Windows PCs to Quest Casting—Mac Mirroring for PC Users
Meta and Microsoft teamed up to make it possible for any Windows 11 PC to cast directly to Quest headsets. This will give you a mirrored screen to interact with inside VR. Think Apple Vision Pro’s Mac mirroring but for your PC. No release date yet, but it’s on the way.Ray-Ban Smart Glasses Get Smarter
Meta’s Ray-Ban Smart Glasses were a sleeper hit, and now they’re leveling up. New features include voice-activated reminders, QR scanning, and integrations with apps like Spotify and Audible. They’re also rolling out live translation in real-time across multiple languages. And for anyone who needs prescription lenses? You’re covered too.AI Everywhere—From Facebook to Instagram
Meta’s got AI on the brain. The new Meta AI is hitting Facebook, Instagram, Messenger, and WhatsApp with voice features and vision capabilities. You can use it to identify birds in photos, edit images on command, or generate content tailored to your interests. Expect even more AI-generated content in your feed soon.Llama 3.2 Models—Mobile-First AI
Meta is pushing the boundaries of AI with new Llama models. They’re rolling out multimodal models that can understand text and images—perfect for mobile devices. The smaller Llama models will run directly on your phone instead of relying on the cloud, thanks to partnerships with Qualcomm and MediaTek.
Bigger Picture:
Meta isn’t slowing down. The metaverse may have faded from the headlines, but AI and augmented reality are now at the heart of their vision. Whether it’s AR glasses, VR headsets, or AI-driven everything, Meta’s still betting on a future where we live, work, and play in digital spaces.
Big Names Are Leaving OpenAI—Here’s What It Means for the AI Giant
Image Source: OpenAI
Quick Byte:
Another high-profile exec is leaving OpenAI—Mira Murati, the company’s CTO, just announced her departure. And she’s not the only one. Within hours, Barret Zoph and Bob McGrew—both top-tier research leaders—followed suit. It’s the latest shake-up in a company that’s been making waves for its rapid AI development and sky-high ambitions.
Key Takeaways:
Mira Murati, CTO and key player in the creation of ChatGPT and DALL-E, is stepping down after 6.5 years at OpenAI.
Barret Zoph and Bob McGrew—two critical leaders in AI research—also announced their exits shortly after.
OpenAI has seen a wave of departures in recent months, including other co-founders and execs who played vital roles in the company’s meteoric rise.
OpenAI continues to push forward with new projects like o1, a model designed to tackle advanced reasoning tasks, even as it faces leadership turnover.
Why It Matters:
The tech world is watching closely—OpenAI is not just any AI company; it’s the company that has been leading the charge in defining what’s possible in AI. When key leaders like Murati leave, it sparks questions about what’s happening inside and how it could shape the future of AI development.
OpenAI’s Shake-Up: What’s Happening?
Mira Murati has been one of the major forces behind OpenAI’s success, leading the teams that gave the world ChatGPT and DALL-E. Now, she’s stepping down, saying she wants time for “her own exploration.” Within hours of Murati’s announcement, Barret Zoph and Bob McGrew—both key AI researchers—also announced their exits.
This is part of a broader trend at OpenAI. In recent months, multiple high-profile leaders have left, including co-founder Ilya Sutskever and VP of consumer product Peter Deng.
So what’s going on? While Sam Altman has said he’s excited for what’s next for Murati and her colleagues, this level of turnover in a company pushing the cutting edge of AI raises questions. Does this signal internal shifts at OpenAI? And what impact could it have on its ambitious roadmap, especially as it seeks a $150 billion valuation?
Bigger Picture:
OpenAI’s rapid evolution has been fueled by brilliant minds pushing the limits of what AI can do. But leadership matters. With Murati and other key players gone, OpenAI faces a critical moment: how to keep innovating at the same breakneck pace while navigating internal transitions. The good news? AI innovation doesn’t stop, and the company still has a deep bench of talent to lean on.
98% of Small Businesses Are Winning Big with AI
Quick Byte:
AI is no longer just a tool for big companies with deep pockets. According to a recent survey by the U.S. Chamber of Commerce, 98% of small businesses are now using at least one AI-enabled tech platform, and nearly 40% have embraced generative AI. Small businesses are proving they can punch way above their weight class by leveraging AI tools to increase efficiency, cut costs, and boost growth.
Key Takeaways:
98% of small businesses are using AI in some capacity, with 40% adopting generative AI like ChatGPT.
AI is giving smaller firms a competitive edge, allowing them to run leaner and smarter.
Businesses using AI are significantly more likely to see growth in sales, profits, and workforce expansion.
81% of small businesses plan to ramp up their use of AI and technology platforms in the near future.
The AI Edge in Business Growth
The numbers don’t lie. Businesses embracing AI are seeing real growth in sales, profits, and workforce expansion compared to their tech-lagging peers. AI adoption is helping small firms streamline operations, handle customer queries more efficiently, and manage backend systems like accounting and payments with ease. This isn’t just about productivity—it’s about leveling the playing field.
Growth and Innovation Through AI
AI doesn’t replace human workers—it empowers them. As Amanda Reineke, CEO of Notice Ninja, put it, “AI won’t replace human work but will augment and lift it.” The data backs her up. 91% of small businesses actively using AI believe it will help them grow, and 81% plan to lean even further into tech platforms going forward.
Bigger Picture:
AI is no longer a luxury—it’s becoming the norm for small businesses that want to survive and thrive. From improving customer service to increasing efficiency, small firms using AI are driving growth and scaling faster than ever before. In an increasingly competitive landscape, AI isn’t just a tool—it’s a game changer.
The Cybercrime Game Just Got Upped – Thanks to AI
Quick Byte:
Cybercrime isn’t just growing—it’s evolving, and AI is at the heart of this revolution. Laura Kankaala, a cybersecurity expert from F-Secure, shares her frontline experiences and paints a picture of just how fast online scams are advancing. Spoiler: AI tools are making these attacks slicker, more realistic, and harder to detect.
Key Takeaways:
Cybercrime is largely about money, and it's only becoming easier for hackers to access your data.
AI is now being used to create sophisticated scams, including deepfake videos and voice clones that trick victims into thinking they're talking to real people.
Phishing toolkits and even malware-for-sale packages have lowered the barrier for entry into cybercrime, making it easier than ever for criminals to launch attacks.
Cybersecurity experts are racing to stay ahead of AI-driven attacks, but the scams are getting better every day.
The Dark Side of AI: Smarter Scams, Bigger Problems
Imagine being tricked by a scammer using a deepfake video filter to pose as your favorite celebrity. Sounds outlandish? Not anymore. AI is making it scarily easy to create sophisticated, realistic scams. Laura Kankaala, head of threat intelligence at F-Secure, says that scams have evolved from crude, low-effort attempts to high-level operations powered by AI tools. These days, scammers don’t even need to know how to code—phishing toolkits and ready-made malware are available to buy online, complete with “customer support.”
That’s the new frontier of cybercrime.
It’s Easier Than Ever to Fall for Scams
Scammers have become disturbingly good at using AI to mimic trusted individuals. Want to pretend to be a CEO, a celebrity, or an authority figure? Deepfake tools make it happen. Laura’s team at F-Secure has seen AI used to clone voices and send fake voice notes over WhatsApp, convincing people to hand over money. She’s also seen romance scams where AI-generated deepfakes fool people into believing they’re video chatting with a celebrity. AI-powered scams are no longer just theoretical—they’re happening right now.
Bigger Picture:
The rise of AI in cybercrime is a reminder that the tech we develop to make our lives easier can also be turned against us. As AI tools become more sophisticated, scammers will continue to push the boundaries of what’s possible, meaning businesses and individuals will need to be more vigilant than ever. Fortunately, cybersecurity experts like Laura Kankaala are on the front lines, using the same technology to fight back. But one thing’s clear: AI is changing the game—for both sides.
How to Install and Run the Latest & Best Open Source Multimodal LLM Available - Llama 3.2 with Vision
Authors:
Fan Zhou, Zengzhi Wang, Qian Liu, Junlong Li, Pengfei Liu
Affiliated with: Shanghai Jiao Tong University, Shanghai Artificial Intelligence Laboratory, and Sea AI Lab.
Summary:
The paper introduces PROX (Programming Every Example), a framework that allows small language models (even with just 0.3 billion parameters) to clean and refine large datasets as well as human experts. PROX turns data refinement into a programming task where the model generates operations like string normalization and noisy content removal for every example. This process improves the quality of data for pre-training large language models, leading to better performance with less computational cost. The results are significant, with improvements of over 2% in downstream tasks and much more in domain-specific applications like mathematical datasets.
Why This Research Matters:
Pre-training data for large language models is often noisy and requires manual cleaning, which can be labor-intensive and inconsistent. PROX automates this process, offering a scalable solution that improves the quality of training data without the need for large human oversight. This results in better-performing language models with fewer computational resources. For companies or organizations training language models, this translates to cost savings and efficiency. PROX shows significant improvements even with smaller models, making it accessible for those who don’t have massive computational resources.
Use Cases:
Pre-training large language models: PROX can be applied across various domains, refining the data used to train models like GPT or LLaMA for improved performance.
Domain-specific improvements: It performs exceptionally well in specialized domains, such as mathematics, where it showed a 20.3% improvement for CODELLAMA-7B on mathematical tasks.
Data-heavy industries: Industries dealing with large, noisy datasets (e.g., healthcare, finance, and education) can use PROX to refine data for more accurate AI models.
Efficient LLM training: PROX makes it possible to train models with fewer tokens, which reduces costs and time spent on pre-training.
Immediate Impact:
Today, PROX allows businesses and researchers to reduce the overhead associated with large-scale AI training. By refining data efficiently, even smaller companies can train high-performing models without needing massive computational infrastructure. This is especially important as AI continues to expand into smaller sectors where computational power is limited.
Future Impact:
In the future, PROX could revolutionize how data is prepared for AI training, making high-quality models more accessible. As the demand for domain-specific AI models grows, PROX’s ability to tailor data refinement to specific fields will be a game changer. Additionally, its efficiency in saving FLOPs (floating point operations per second) points towards a future where AI development becomes more cost-effective and environmentally sustainable.
This research also lays the groundwork for more automated and refined data processing tools that can adapt to various domains, boosting the precision and adaptability of AI systems in industries worldwide.
Open Music - Next-gen diffusion model designed to generate high-quality music audio from text descriptions!
M1 Project - AI-powered marketing assistant that helps businesses create, execute, and optimize marketing strategies. The platform offers tools like Ideal Customer Profile generation, ad creation, and custom marketing tasks.
NotCommon - Helps users discover legitimate social media accounts and brands by providing verification and popularity data, while also offering a browser extension to ensure users avoid scams and fake profiles online.
Dubble - Watches how you work and translates your actions into written step-by-step guides, videos and screenshots - so you don’t have to.
Llama Coder - AI-powered coding assistant by Meta.
Heatbot - Data-driven generative UI builder. Upload user analysis heatmaps and let AI generate code for an improved website in seconds.
How To Gamify Your Offer:
I need help generating gamification ideas for my [PRODUCT/SERVICE/COMMUNITY].
My target audience is [DESCRIBE YOUR IDEAL USER].
The main goals of my offering are to help users [LIST 2-3 KEY OBJECTIVES].
The key actions I want users to take are [LIST 3-5 DESIRED ACTIONS].
Please generate 5 unique gamification concepts that could be applied to my offering. For each concept, include:
A catchy name for the gamification system
A brief overview of how it works
The specific user behaviors it aims to encourage
The rewards or incentives users can earn
How it ties into my business objectives
Any potential challenges or considerations for implementation
The ideas should be creative, engaging, and tailored to my specific audience and goals. Avoid generic concepts like simple point systems or basic leaderboards unless they have a unique twist.
Consider incorporating elements like:
Narrative or storytelling components
Social interaction or competition
Personalized challenges or quests
Virtual economies or currencies
Unlockable content or features
Progress visualization
Surprise and delight mechanics
Aim for a mix of both extrinsic motivators (points, badges, rewards) and intrinsic motivators (mastery, autonomy, purpose).
The gamifications should be simple and relatively easy to implement.
This new ChatGPT Advanced Voice mode 🤯🤯
— Jason Staats⚡ (@JStaatsCPA)
4:51 PM • Sep 25, 2024