PA Media’s Post

The new term, "AI PC". The tech and investment news channels were abuzz with it yesterday as multiple big tech companies made announcements. What is it? Essentially, any PC that has either a SoC with AI capabilities (think ARM-based designs like Apple's M series), a modern GPU, and/or a specialised AI processor (you might hear terms like NPU, LPU, and TPU). "Hold on a minute, I have one or more of those in my current PC/laptop!" Yep, you already have an "AI PC" then! Well, kinda, you will also need... * Any number of local "smaller" models (<= ~35b parameters based on today's device hardware), such as Llama 3, Mistral, Gemma, Stable Diffusion and Phi-3. * Any number of specialised apps. MS are pushing Copilot+, and you can do something similar using open source; check out "Ollama" and associated UIs like "OpenWeb UI". What is this about then? In real terms, this is about edge compute and enabling AI workloads on devices, on-prem servers, and in the cloud. This approach addresses various constraints, including NFRs, costs, privacy, bias, cultural differences, regulatory uncertainties and trust challenges... the list goes on. It’s also about enabling the vast wave of products, services, and associated professional services big tech and partners want to sell to businesses and consumers, by giving it a "snappy" marketing name and making it "A NEW THING!" How did we get to this point? "Open" AI models (varying licences and levels of openness apply). The evolution of these models is moving and iterating fast, very fast! This time last year (~May '23), a Google internal memo leaked, stating, ‘The uncomfortable truth is, we aren’t positioned to win this arms race and neither is OpenAI. While we’ve been squabbling, a third faction has been quietly eating our lunch.’ They weren't wrong, and they were right to join in... * July '23 Meta's Llama 2: Good, but not quite state-of-the-art (SOTA) compared to closed models from OpenAI and Google, but similar to GPT 3.5, which was SOTA as recently as February '23. * April '24 Meta's Llama 3: The smallest version (8b) runs very well on a PC or Apple Laptop (my inference speed is ~64 words/s on an Nvidia GPU and ~41 words/s on Apple M series silicon), and it matches the capability of Llama 2 70b, which requires a dedicated on-prem server or data centre. Llama 3 70b is on par with SOTA closed models like GPT-4, Gemini 1.5, and Claude 3. * What next? Last week, Google announced Gemma 2 27b, scheduled for a June '24 release. It’s specifically sized and optimised for on-device edge use cases, with performance comparable to Llama 3 70b, but capable of running efficiently on your laptop. All the pieces are falling into place for some very interesting product, service, and business process use cases, and in reality we are only 18 months into this race... #AiPC #ai #artificialintelligence #genai #generativeAI #gemma2 #llama3 #google #meta #openai #nvidia #apple #opensource #llm #ollama #openwebui

In this article more on the joint-venture news on #AI, launched by OpenAI, direclty involving #Oracle on the new project called #Stargate. https://lnkd.in/dUrY-AYe

Microsoft and Apple Quit OpenAI Board Amid Growing Regulatory Scrutiny In a significant shift within the tech industry, Microsoft and Apple have decided to step back from their observer seats on the OpenAI board. This move comes amid increasing regulatory scrutiny over big tech's influence and partnerships in the artificial intelligence space. Key Takeaways- 🤖 Microsoft and Apple Withdraw from OpenAI Board: Both companies have stepped back from their observer seats on the OpenAI board. The move is in response to increasing regulatory scrutiny over big tech's influence in AI. 🤖 Regulatory Concerns: Driven by mounting concerns about potential conflicts of interest. Reflects worries over the growing power of tech giants in AI development. 🤖 Microsoft's Involvement: Microsoft has invested billions into OpenAI. Their technology is integrated into products like Azure and Office 365. The company will continue its financial support and collaboration with OpenAI. This decision addresses antitrust concerns and demonstrates a commitment to fair competition. 🤖 Apple's Strategy: Apple, known for its independent approach to AI, has also relinquished its observer role. The move aligns with Apple’s strategy to mitigate regulatory risks and maintain its distinct AI development path. 🤖 Future Dynamics: Departure signals a significant moment in the AI industry. Collaboration, transparency, and regulatory adherence are crucial for future AI advancements. Observers will be watching how this decision influences AI development and regulation moving forward. 🤖 Sources for More Information: CNBC, Financial Times, Business Today, The Verge, Axios, Bloomberg, and Forbes provide detailed reports on this development. https://lnkd.in/eKg9QF2A https://lnkd.in/eVC9zj5i https://lnkd.in/eTP5khPJ https://lnkd.in/erfFFNtr https://lnkd.in/eyQ63Y4s https://lnkd.in/evs9THvm https://lnkd.in/eF6NpSHZ #AI #TechNews #Microsoft #Apple #OpenAI #Regulation #Innovation #Antitrust #Technology

That Getty Image is a Teledeck Studio... an immersive room you can stand in. It can be tuned like a channel. IPTV transport ready... multicast globally for tuning to a TV room (Televerse) anywhere on the planet via Starlink. Create any environment on an 8 screen immersive wireframe. (8) 16:9 screens of any size... from cellphone to video wall to 3D projection. (4) Walls are a Televerse. An onsite Teledeck Studio could be built within a couple months... 2,304 Cubic Foot. 16'x9' immersion room. HD, 4K, passive glasses 3D options. A HyperCube is (7) recorded TV rooms or remote live locations. AI created an image for me several weeks ago that had 8x8x8 blocks of rooms like this one... essentially a 512 channel CATV headend... about $1m in hardware... lack of content is a constant roadblock. And... no doubt a $100b Phase 5 Stargate could create AI simulations for millions of Teledeck channels. Singlecast or Multicast. Teledeck on Demand. Headset viewing would be similar to standing in the center of a 2,304 Cubic Foot 3D Teledeck. Cartesian space adds a possibility of adding a Sense of Presence within the 16x16 floorspace/ headspace.

NEW: The FTC dissects three of the big partnerships in AI: Microsoft + OpenAI: $13.75 billion Amazon + Anthropic: $8 billion Google + Anthropic: $2.55 billion The FTC found many areas of concern to watch: locking up AI talent, exclusive services, transfer of proprietary business information https://lnkd.in/eHQhR4xU

Oracle continues to move forward in the Cloud and with AI.

Read about the new Stargate Project. No new worm holes to new planets this time (!) but just as exciting when considering the development of AI, through the collaboration with Oracle, OpenAI and Nivdia. https://lnkd.in/e6TKaErd

We shouldn’t expect this to be the only alliance put together in order to begin fulfilling the AI tsunami which is just starting to take shape. And reminder this is only the training model side - imagine what the interference (EDGE) deployments will look like in every major city greater than 1 M. Welcome to the beginning of the most transformative era in modern history. #ai #gasturbineinfrastructure #llm #digitalinfra

☁️ Recent news on AI with Oracle and OpenAI are truly exciting times for innovation, especially in industries like healthcare, where the implications of AI advancements are revolutionary. Personalized cancer vaccines or faster drug developments are just the beginning of what’s next.

Announcing the Stargate Project: This project will not only support the re-industrialization of the United States but also provide a strategic capability to protect the national security of America and its allies. As part of Stargate, Oracle, NVIDIA, and OpenAI will closely collaborate to build and operate this computing system.