Deepak Agarwal

The cost of training top models is increasing 2-3X per year, and at this rate will reach $1 billion per model by 2027. The findings are from Epoch AI and follow their recent analysis of increases in compute for foundation models running at 4-5X per year. They've also provided some interesting data on the relative breakdown of cost components like staff, hardware, and energy. A few thoughts after reading this: - I would assume staff costs won't increase as much as hardware and energy costs, so the relative contribution of human labor to future AI models will decrease, in yet another sign of capital's increasing dominance over labor in the industry (just look at NVIDIA's ever-rising stock price). - If compute is increasing 4-5X per year but cost is increasing 2-3X, is it safe to assume that training runs are getting more efficient? - Some people, like Anthropic CEO Dario Amodei, have said we'll see $1 billion training runs in the next year. Does this mean the 2-3X pace is about to increase? That Epoch is wrong in its estimate? Or that Amodei is overstating the cost, possibly to put fear into competitors who can't keep pace? - Only a few companies are going to be able to train future models. We're likely going to see further consolidation. Perhaps we get to a future with only two large general purpose models, like the Windows/MacOS and Android/iOS duopolies we've seen in the past. Article here: https://lnkd.in/g5SejXen

7

Congrats VERSES (OTCQB:VRSSF) Gabriel René and Dan Mapes and Dr. Karl Friston for introducing an "efficient alternative to deep learning, reinforcement learning and generative AI called Renormalizing Generative Models (RGMs) that address foundational problems in artificial intelligence (AI), namely versatility, efficiency, explainability and accuracy, using a physics based approach." #ai #activeinference #RGM #spatialweb #AGI #cognitivecomputing https://lnkd.in/gcYSM4eN

23

“It’s all about core product creation at the companies that use our AI,” SandboxAQ CEO Jack Hidary told VentureBeat. “And so if you want to create a drug, a diagnostic, a new material or you want to do risk management at a big bank, that’s where quantitative models shine.” OOh and you want an enterprise use case? How's this LQMs for a better version of a Monte Carlo simulation. A Monte Carlo simulation is a classic form of computational algorithm that uses random sampling to get results. With the SandboxAQ LQM approach, a financial services firm can scale in a way that a Monte Carlo simulation can’t enable. “If I have a portfolio and I want to know what the tail risk is given changes in this portfolio,” said Hidary. “What I’d like to do is I’d like to create 300 to 500 million versions of that portfolio with slight changes to it, and then I want to look at the tail risk.”

6

Something that stuck with me during my interview with Naveen Rao, VP of Generative AI at Databricks: "I take the hard path." He built an AI-focused chip company Nervana in 2014 when machine learning was still academic theory and the chip industry was deeply out of favor. Naveen saw computing architecture would need to fundamentally change for ML workloads. And then in 2019, observing the power of LLMs like BERT, Naveen foresaw the need for enterprise-grade services to manage their training, deployment, and customization. That vision led to a $1.3B acquisition by Databricks in 2023. What I love about Naveen's approach is his willingness to build for markets that don't exist yet. "There isn't a market, employees don't even know about it. So how the hell do you build a company?" You educate. You convince. You get people to see what you see. It's brutal work, but it's important work. You're reshaping how people think about problems.

176

8 Comments

https://lnkd.in/ecuZDqKz. https://lnkd.in/ecuZDqKz Liquid AI Introduces New Class of Foundational Models, Advancing AI Performance and Efficiency Boston, September 30, 2024—Liquid AI, a pioneer in artificial intelligence innovation, is set to redefine the AI landscape with the launch of its Liquid Foundational Models (LFMs). These models bring unprecedented levels of performance and efficiency to generative AI, offering powerful solutions for industries looking to harness artificial intelligence more effectively on device or on server. A technical whitepaper about model performance will be available on September 30, 2024, at 11 am EST. Key Highlights: ● LFMs Introduce a New Scaling Law for Generative AI: Markedly improving quality, cost-effectiveness, and power efficiency compared to current GPT models. ● Innovative Architecture: Leveraging dynamical systems and advanced mathematics, LFMs achieve superior performance with fewer parameters. ● Initial Model Release: The lineup includes models with 1B, 3B, and 40B parameters, each setting new benchmarks in their respective categories. ○ LFM-1B: A 1.3-billion-parameter model achieving the highest scores across benchmarks in its class. ○ LFM-3B: A 3.1-billion-parameter model that outperforms many 7B and 13B models, making it ideal for edge deployment and resource-constrained environments. ○ LFM-40B: A 40.3-billion-parameter Mixture of Experts (MoE) model designed for complex tasks, offering performance comparable to much larger models while remaining cost-effective. Real-World Applications: • Enterprise: LFMs enable the design of more powerful AI assistants and analytics tools that can run efficiently on existing infrastructure, reducing operational costs. • On-Device: LFMs prioritize privacy, low latency, security and offline accessibility, making them suitable for mobile and edge deployments. These models have demonstrated impressive results on standard LLM evaluation benchmarks. Additionally, LFMs maintain near-constant inference time and memory complexity, allowing for longer input sequences without significant increases in computational resources. Advancements in AI Architecture: LFMs introduce a new design space for foundational models by employing adaptive linear operators rooted in dynamical systems and signal processing. This approach unifies and extends existing computational methods in deep learning, providing a systematic way to explore architectural possibilities. The models are highly adaptable, optimized for various hardware platforms including NVIDIA, AMD, Qualcomm, Cerebras, and Apple devices.

46

Last thursday, I had a chance to attend Hubspot + Fellows fund sponsored GenAI summit! I got the opportunity to listen to my former bosses Karen Ng and Vijay K Narayanan, and many great leaders! Few key themes that came out at the conference: 1) All enterprises (big/small) have been playing/experimenting with LLMs for past year and half, now is the time for differentiation. Differentiation cannot just be speed (for startups) or data (for incumbents), it has to be a valid business model. 2) Enterprise software as we know today is dead, enterprise software needs to be re-written. Databases are not longer relevant, enterprise flows/workflows are no longer relevant. 3) For enterprises, as co-pilots are beginning to be adopted internally advanced RAG + fine tuning capabilities are coming out to be essential to make the these systems production ready. 4) You are no longer selling Enterprise AI software to a particular group in a company, you sell it to the whole C-suite with CISO, legal, Chief AI officer, Chief Customer officer, CRO, etc.

33

3 Comments

I just read an insightful piece by Eric Broda (published in Towards Data Science) that outlines the Agentic Mesh—a new ecosystem designed for autonomous agents to interact, collaborate, and transact. His work introduces a structured approach for managing the coming surge in autonomous agents, showing how they can function independently yet collectively. Agentic Mesh serves as a networked environment where autonomous agents operate in a safe, transparent, and efficient manner. It allows agents to find each other and securely interact under a framework of policies. Each agent has a specific purpose and autonomy, making it a valuable contributor to the ecosystem without sacrificing security or control. Core Components: 1. Marketplace: A centralized space where users discover agents by tasks and capabilities, offering a trusted environment for engagement. 2. Registry: A metadata repository organizing each agent’s purpose, owner, and policies, allowing agents and users to securely find and connect with each other. 3. DNS Integration: Each agent has a recognizable digital identity, enabling seamless discovery and interaction, just like an internet service. 4. Agent Stack: This layered structure of communication protocols, control services, and interfaces helps agents operate effectively and responsibly. It ensures that agents act independently within controlled parameters. Agentic Mesh’s effectiveness depends on agents sharing consistent attributes, including: • Purpose: A clear objective drives each agent’s actions. • Ownership: Transparency around ownership fosters responsibility and trust. • Autonomy: Agents act independently within set parameters to complete tasks. • Discoverability: Easily locatable, ensuring efficient interaction within the network. • Intelligence: Agents utilize language models to plan and execute tasks, adapting to complexity as needed. Operational Flow: Registration, Discovery, and Execution Agent interactions in the Mesh follow three core flows: 1. Registration: Agents set up purpose, policies, and registration data to gain visibility in the mesh. 2. Discovery: Through the registry, agents and users locate potential collaborators based on defined needs. 3. Execution: Engaged agents use AI models to plan and perform tasks, enlisting collaborators when necessary to complete workflows. Read the full article: https://lnkd.in/eVmq5BFC

16

1 Comment

We are not slowing down on Gen AI, come see why...and how it changes things, the nature of work. https://lnkd.in/gW3ettg3

2

My quick thoughts about the pent-up innovation inside the enterprise that is set to explode with those who make GenAI a priority. This video was in support of a great article from Brett A. Hurt around the same topic. You can read that here: https://lnkd.in/gwBBbX4M https://lnkd.in/gvDyRKZN

9

1 Comment

👉 Howdy LinkedIn! Tomorrow, on "Building Products with Generative AI" - I will be hosting 🚘 Nitin Kartik at 11:30AM CST/12:30 PM EST! We're going to be digging into the intersection of Product Marketing & Generative AI, as well as other topics that will branch from Nitin's experience in the space. I'm excited for another interesting episode with a great thought leader in the space - we hope to see you join for tomorrow's episode! #GenerativeAI #AI #ProductMarketing

15

2 Comments

In today's rapidly evolving business landscape, Generative AI stands out as a transformative technology capable of driving significant impact across various domains. This one-day workshop is designed for executive and departmental leaders, aimed at fostering a deep understanding of GAI, aligning strategic vision, and catalyzing actionable insights. Participants explore practical applications, engage in hands-on activities, and develop strategies to leverage GAI for enhanced business outcomes. While the primary focus of the sample workshop is on marketing, the workshop framework is adaptable for other functional areas such as engineering, operations, finance, administration, HR, and recruiting. #generativeartificialintelligence #generativeAI #GAI #AI #digitaltransformation #CIOs #CTOs #CDOs #CDAOs #CMOs #CFOs #CEOs #EQengineered

27

6 Comments

Entering year three of the post-AI era, and here's my prediction: Enterprises are about to jump headfirst into custom model training and fine-tuning. Generic AI models are great for consumers and some prosumers. But thinking they are suited for industry-specific use cases is a bit.. optimistic. These models weren't built for that and definitely haven't been trained on your vast, private enterprise data. There's a growing gulf between data inside enterprise silos and the data generic AI models are trained on—a gap that's only widening. It's basically the public internet vs. your company's private data. Companies aren't keen on giving up their moat and all their data. So, custom model training and fine-tuning on hardware **you** control is set to become the new norm. Simply deploy the trained, proprietary model to your secure cloud or even on-premises.

18

2 Comments

Yann LeCun's critique of generative AI and advocacy for Objective-Driven AI is an interesting perspective from one of the great pioneers in the space. LeCun's critique underscores the limitations of current AI models, prompting a reevaluation of the direction of AI development. However, while he notes that models lack true intelligence and are subject to hallucinations, hallucination is what provides these models "creativity" that can be leveraged in other ways. #AI

12

Here are some early insights from the survey we are using to inform our Sept. 19 roundtable on generative AI monetization. Michael Mansard, Kyle Poyar and Steven Forth will go deep into the trends in generative AI and how it us being monetized. We include lists, categorized by AI, on what people worry about with generative AI and what gets them excited. https://lnkd.in/gNXDt8Ry #SaaS #pricing #monetization #AI #genAI #generativeAI

2

The state of AI in early 2024: Gen AI adoption spikes and starts to generate value. Recent insights from McKinsey & Company highlight a surge in generative AI adoption across various industries, revealing transformative effects on business operations and strategy. Key Takeaways: - Widespread Adoption: From startups to multinationals, organizations are leveraging AI to innovate and enhance workflows, particularly in sales and marketing. - ROI Generation: Strategic investment in AI is not just about cost savings; it's increasingly about driving revenue growth. - Personal to Professional: The integration of personal AI usage into professional settings is creating new opportunities for efficiency and creativity. My take is that with the latest announcements from Apple, this will accelerate - Challenges Remain: Despite the progress, concerns around data accuracy and legal complexities persist, though they are diminishing as technology advances. This is not surprising as innovation is hard. Hallucinations & inaccuracies in my opinion will disappear more and more as people get used to GenAI, Prompting and models get better - Strategic Approaches: Companies are opting for configurable off-the-shelf AI solutions rather than building from scratch. I'm very convinced that's the best strategy to apply All in all, it shows it is still the early days for adoption but companies are finding success much faster than before. I think the future is bright for companies that adopt faster than others. #GenerativeAI #DigitalTransformation #AIinBusiness #TechLeadership #Innovation #ArtificialIntelligence #TechTrends

21

1 Comment

$2 H100s: How the GPU Bubble Burst Last year, H100s were $8/hr if you could get them. Today, there's 7 different resale markets selling them under $2. What happened? The market has flipped from shortage ($8/hr) to oversupplied ($2/hr), because of reserved compute resales, open model finetuning, and decline in new foundation model co’s. Rent instead. https://lnkd.in/gpQEidmC

2

1 Comment