- AI For All
- Posts
- NVIDIA vs The World
Hello readers,
Welcome to another edition of This Week in the Future! The AI industry no longer wants to depend on NVIDIA. AMD, Microsoft, and OpenAI are looking to develop their own AI chips. Plus, Air Street Capital has released their State of AI 2023 Report!
As always, thanks for being a subscriber! We hope you enjoy this week’s content — for a video breakdown, check out the episode on YouTube.
Let’s get into it!
NVIDIA vs The World
The AI landscape is undergoing a dynamic shift, especially in the hardware realm. With NVIDIA firmly holding the reins as the top dog in the AI chips market, other tech titans are making concerted efforts to challenge this supremacy, aiming for more self-reliance and reduced costs.
AMD Steps Up
AMD, traditionally recognized as NVIDIA's rival in the broader hardware domain, but not necessarily in the AI sector, has thrown its hat into the ring. By acquiring open-source AI software giant nod.ai, AMD is looking to bolster its AI prowess. The core objective behind this acquisition is to equip AI customers with open software solutions that are specifically optimized for AMD's hardware. Nod’s compiler-based platform, SHARK, is anticipated to be integrated into AMD's product lineup, paving the way for AI chips and associated software solutions that optimize their performance.
OpenAI's Ambitions
There's chatter in the industry about OpenAI developing its own AI chips. However, the journey to chip manufacturing is fraught with challenges such as securing manufacturing facilities, acquiring essential materials from global sources, and ensuring smooth production. All of these challenges, coupled with NVIDIA's dominant stance in setting chip prices, make this venture risky. Yet, the allure of saving billions in hardware costs, especially if OpenAI scales its operations akin to Google, might make the gamble appealing. Envisaging a future where traditional search engines could be dethroned by tools like ChatGPT, OpenAI might see this as a worthy bet.
Microsoft's Play
Microsoft, too, seems to be gearing up for some major announcements. Speculations are rife about the tech behemoth unveiling a custom-designed AI chip at their Ignite event. This move would likely be a collaborative effort with an established chip manufacturer, potentially AMD. Similar to other industry players, Microsoft's current AI operations rely heavily on NVIDIA chips.
Our Take
The AI chip market is sizzling. Despite NVIDIA's unparalleled product excellence and leadership, the demand far outstrips the supply. This overwhelming reliance on NVIDIA, combined with their steep pricing, makes the quest for alternative chip sources crucial. With tech mammoths like Google and Amazon already touting their proprietary chips, the impending wave of competition promises to benefit the industry at large. As more contenders enter the fray, we might witness a democratization of AI hardware, potentially leading to more accessible prices and diversified vendor options.
The Absolute State of AI
A recent report from the VC firm Air Street Capital provides an insightful overview of AI's progression, focusing on its impact in various sectors like industry, research, government, and safety. The comprehensive study, based on rigorous statistical research, offers a retrospective on the past year's developments and forecasts for the upcoming year. Here's just some of what was reported:
Context Length is the New Parameter Count: The report suggests that context length is the new aspiration of LLM companies, rather than the sheer amount of parameters.
Depleting Data Reservoirs: A bold prediction suggests that by 2026, all high-quality language data sources, including books and the broader internet, will be exhausted. Other, less refined language data sources might last until 2030-2050, while vision data could stretch till 2060. Notably, some sources like transcribed audio data and AI-generated datasets haven't been considered. This implies that AI's future might not solely revolve around escalating parameter counts; optimizing existing models will be paramount.
GPT-4’s Coding Supremacy: GPT-4 stands out as the unrivaled leader in coding abilities. While other models are on par in benchmarks, in terms of real-world coding tasks — like writing, running, and interpreting code along with natural language prompting — GPT-4 is unmatched.
Trending AI Topics: The AI community's spotlight this year was majorly on prompt engineering and reinforcement learning with human feedback (RLHF). However, due to frequent model updates, prompt engineering poses challenges. RLHF, while essential for tuning large models for effectiveness, remains costly and time-intensive.
Computer Vision Evolution: DINOv2, a self-supervised vision transformer by Meta, is revolutionizing the computer vision domain. This tool, blending the power of transformers, optimizes the data collection and training pipeline for computer vision applications.
Boosting Productivity: AI tools, especially ChatGPT, are enhancing productivity across various sectors. They're proving invaluable in coding and automating repetitive tasks like content generation or email drafting.
Potential Catastrophic Risks: The dangers associated with AI have surged to the forefront of public discourse. The AI community's top echelons are emphasizing these risks, garnering attention from governments globally. The ease of "jailbreaking" models and their occasionally unsafe capabilities are alarming. Before AI systems assume more responsibilities, addressing these vulnerabilities is crucial.
🔥 Rapid Fire
Booz Allen aims to bring AI to government offices
EY and IBM launch AI solution for HR
Anthropic is making language models understandable
EU and Japan are converging on generative AI
Tesla is building a new home for AI supercomputer
Dropbox releases AI-powered Dash
Adobe introduces the next generation of Firefly
Startup demos decentralized AI infrastructure
ElevenLabs releases AI voice dubbing
Replit releases open source AI dev tools
Google is solving traffic with AI
Startup raises $55M for autonomous ships
Electricity of small country needed for powering AI
🎙️ The AI For All Podcast
This week’s episode featured Deep Dhillon, the Founder and Chief Data Scientist at Xyonix, who discussed the fascinating evolution of natural language processing (NLP) and the valuable insights businesses can extract using NLP and AI.
📖 What We’re Reading
This week’s handpicked insights include how to improve security with computer vision. And while you’re deploying a CV application with DINOv2 as the backbone, you’ll need to know strategies for smooth scaling. Thankfully, MIT Technology Review, Adobe, EY, and Owkin have us covered.
Improving the Security of Business Systems with Computer Vision (link)
“According to BusinessWire, the value of the investigation and security services market will climb as high as $417.16 billion by 2025. But it’s still challenging for security teams to minimize losses. Fortunately, thanks to evolving computer vision technologies, maintaining security can be more efficient.”
Generative AI deployment: Strategies for smooth scaling (link)
“Executives recognize the transformational potential of generative AI, but they are moving cautiously to deploy. Nearly all firms believe generative AI will affect their business, with a mere 4% saying it will not affect them.”