7 | Enterprises and Large Language Models (LLMs) | AI Monopoly

Happy Friday! Despite the buzz around Amazon's rumored LLM and the unveiling of groundbreaking models like GEMINI and GPT-5, this week's focus shifts towards the intricate relationship between Enterprises and Large Language Models (LLMs).

And as OpenAI kickstarts a new funding round, it's the perfect moment to pause and reflect on the financial currents shaping the LLM landscape.

Our Menu :)

• Weekly Digest:

  • 2023: State of Generative AI in Enterprises, Survey by Menlo Ventures 🏢🤖

  • AI's New Era: Big Tech's Rising Monopoly? 🎩🤖

  • AI Startup Program at STATION F in Paris 🇫🇷🤖

• Actionable Tips:

  • About GPT Security 👮🤖

• Training & Crash Courses:

  • Generative AI for Everyone: A Course Changing Lives. 📖🤖

ARTICLE
2023: State of Generative AI in Enterprises - Survey by Menlo Ventures

Generative AI is revolutionizing the enterprise landscape, but how deeply has it permeated? Menlo Ventures’ comprehensive survey of over 450 enterprise executives in the U.S. and Europe offers a revealing snapshot.

Our Quick Overview of the Current State

Despite the buzz, *generative AI's adoption in enterprises lags behind **traditional AI and cloud software investments. In 2023, investments in generative AI reached around $2.5 billion, modest compared to other technologies.

* Generative AI: This type of AI focuses on creating new content or data. It can generate text, images, music, and other forms of media that didn't exist before.

** Traditional AI: This refers to AI systems that analyze and interpret existing data to make decisions, predictions, or categorizations. They don't create new content but rather process and analyze the data they are given.

2. Market Predictions

1 => Gradual Adoption: Just like in the early days of cloud technology, enterprises are slowly warming up to generative AI. Consumer excitement around tools like ChatGPT is yet to fully translate into widespread enterprise adoption

2 => Favoring Established Players: The market currently favors well-known companies that add AI features to their already existing products.

Microsoft Copilot , an AI tool integrated across its various applications like Microsoft 365 and Bing, is projected to generate over $10 billion in revenue. This tool, part of a suite of established AI integrations, contrasts with newer AI solutions tailored for specific business needs.

Tools like Copilot stand in contrast to a newer wave of AI-native solutions that map to existing SaaS categories (departmental, vertical, and horizontal AI applications). These new AI entrants compete in a crowded market against the deep pockets of category leaders. For every AI CRM, there is a Salesforce Einstein; for every AI design tool, a Figma copilot; and for every contact center agent, an Observe.AI.

They expect the incumbent advantage will hold for the next few years until new and more powerful AI approaches, like agents and multi-step reasoning, become prevalent.

3 => Powerful context-aware, data-rich workflows will be the key to unlocking enterprise generative AI adoption

In the enterprise sector, success demands offering solutions that are significantly superior to the existing norms. Incremental improvements are not enough; startups must showcase major productivity gains and transformative workflows.

This involves leveraging next-gen reasoning techniques, utilizing proprietary data sets for an edge, and creating integrated workflows with feedback loops. Such groundbreaking approaches are key for startups to unlock the enterprise market.

3. Opportunities for Startups

Despite the challenges, there are significant opportunities for startups, particularly in three key areas:

• Vertical AI: Industry-specific applications where AI can drive end-to-end automation.

  

  Two sectors are leading the way for vertical generative AI adoption: Marketing and Legal.

  
  Over time, we will see more conservative industries, like healthcare and finance, embracing the value of generative AI too. Although executives in these industries report fewer use cases for generative AI today

• Horizontal AI: Solutions applicable across various industries and departments, enhancing workflow efficiency.

  

• The Modern AI Stack: Enterprises invested $1.1 billion in the modern AI stack, the largest new market in generative AI.

  In the first half of the year, the modern AI stack was the Wild West. It was under constant construction and revision, making it difficult for buyers to know where to invest.

  
  In the last six months, the industry has converged around some core components and standard practices for enterprise deployment, providing a higher degree of stability and standardization :

Driving Trends:

• Most companies use ready-made software models. Only about 1 in 10 companies create their own software models from scratch.

• Closed-source models like those from Anthropic and OpenAI dominate—comprising upwards of 85% of models in production—compared to open-source options like Llama 2 and Mistral.

• 60% of businesses use several different software models and choose the best one for each task. This multi-model approach eliminates single-model dependency, offers higher controllability, and cuts costs.

• 96% of the money spent in this area is used for operating the software models, not for training them.

• Tools and Middleware for this technology are still improving. The most common way to customize these systems is by designing specific instructions (Prompt engineering) while the most popular evaluation method is human review.

  

Future Outlook:

Menlo Ventures paints an optimistic picture for the future of generative AI, emphasizing that it presents a "once-in-a-generation opportunity" for startups to capitalize on the technology's transformative potential.

Several factors are fueling the growth of the generative AI market: While the modern AI stack was in its early stages of development in the first half of the year, it has since matured and standardized, providing enterprises with a more stable and reliable foundation for deployment.

1. Increasing Data Availability: The abundance of data for training generative AI models is rapidly expanding, enabling the development of more accurate and powerful models.

2. Declining Compute Costs: The cost of compute is decreasing, making it more affordable to train and run generative AI models, further democratizing access to this technology.

3. Soaring Demand for AI Solutions: The demand for AI solutions is skyrocketing, driven by their ability to address a wide range of challenges and unlock new possibilities.

ARTICLE
AI's New Era: Big Tech's Rising Monopoly?

In the evolving landscape of artificial intelligence, a critical question arises: Is the future of AI being monopolized by Big Tech? dominated by a select few - a modern "Magnificent Seven" comprising Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla.

The Investment Paradigm Shift

2023 has witnessed a seismic shift in venture capital trends, with these tech behemoths pouring substantial resources into AI startups. Amazon's involvement with Anthropic and Google's rumored backing of Character.AI exemplify this trend. Such investments are reshaping the venture capital landscape, previously diversified, now increasingly concentrated in the hands of a few.

The Capital Conundrum

AI startups, unlike their SaaS counterparts, face unique capital challenges. Foundation models like GPT-4 demand exorbitant funding for training and scaling.

This financial demand stems not from aggressive growth strategies but from the inherent costs of developing AI products - data collection, training, and fine-tuning.

Each user interaction with AI apps incurs a tangible cost, a stark contrast to the near-zero marginal cost of traditional SaaS applications.

Big Tech as AI's Financial Pillar

Given the uncertain financial returns of AI investments, it’s predominantly big tech companies, with their deep pockets and cloud platforms, who are underwriting this new wave of AI development. Microsoft’s hefty investment in OpenAI exemplifies this trend, highlighting the tech giant's strategic bet on AI's future.

The Open-Source Dimension and Meta's Strategy

Meta has emerged as an unexpected obstacle to companies like Microsoft, Google, and others controlling next-gen AI tech. Rather than acquiring AI startups, Meta openly releases its cutting-edge AI models like Llama 2.

Despite not having prominent AI products until recently, Meta has quietly become an AI research powerhouse under Yann LeCun. It has published over 1,000 AI papers.

After leaks, Meta began publishing not just code but model weights, enabling others to replicate models like Llama 2. Meta is eroding the competitive advantage of OpenAI, Anthropic, DeepMind.

By democratizing AI, Meta aims to decrease the value of investments by competitors. But Meta still controls training data/RLHF process, and its licenses aren't fully open source.

Ironically, Meta/Zuckerberg may be all that stands between AI dominance by Amazon, Google, and Microsoft. But some are uneasy about Meta's prominent role.

NEWS
AI Startup Program at STATION F in Paris 🇫🇷

What’s New:

Meta's exciting project is the "AI Startup Program" at STATION F in Paris , a famous place for new tech companies to grow. They've teamed up with Hugging Face 🤗 and Scaleway to help these startups use advanced AI in their products and services.

Why Does It Matter?

The program is designed to support startups focusing on projects built on open foundation models, encouraging the integration of these models into their products and services.

This initiative aligns with the broader objective of promoting the economic and technological benefits of open, state-of-the-art models within the French ecosystem.

Selected startups will benefit from mentoring by Meta's researchers and engineers, access to Hugging Face's platforms and tools, and computing resources from Scaleway​​.

🤗 Hugging Face CEO Clément Delangue emphasizes the importance of open-source AI as a cornerstone for democratizing ethical AI.

ACTIONABLE TIP
About GPT Security

Mastering GPT security isn't as hard as it seems! 🚀 Dive into this thread where experts share their tried-and-true tips after pushing limits and hitting rate caps. #GPTSecurity #TechTips 👩‍💻👨‍💻 Stay safe, stay informed!"

TRAINING
Generative AI for Everyone: A Course Changing Lives

The newly launched course, Generative AI for Everyone, instructed by Andrew Ng, is creating waves of enthusiasm among learners.

Check out what learners had to say about our new course, instructed by Andrew Ng:

Inspiration and Practicality Combined: 

“A remarkable journey of learning, 'Generative AI for Everyone' stood out for its rich inspirational content and practical applications. The course was so engaging that I was completely absorbed, foregoing breaks.”

The Perfect Introduction to AI and Machine Learning:

 “As a beginner, this course was a perfect entry point for me. It demystified the complex jargon of the AI and machine learning world, making the concepts accessible and understandable.”

Essential for Entrepreneurs and Technophiles: 

“I strongly endorse this course for every entrepreneur, regardless of their tech focus. It lays down a solid foundation, equipping learners with insights to navigate the future – a future that’s already here. Beyond entrepreneurship, it’s a vital resource for anyone aiming to be an informed participant in our tech-driven society. And of course, the invaluable contribution of Andrew Ng to this field and humanity cannot be overstated.”

Link to learn more 👉

OUR SPONSOR
The Newsletter Platform Built for Growth

When starting a newsletter, there are plenty of choices. But there’s only one publishing tool built to help you grow your publications as quickly and sustainably as possible.

beehiiv was founded by some of the earliest employees of the Morning Brew, and they know what it takes to grow a newsletter from zero to millions.

Try beehiiv absolutely free with no credit card required.

Thank you, see you next week!