Emergence Capital

Are we at the top of the S-Curve for AI? Likely, yes. Performance is plateauing. AI companies are scavenging for data. For example, with Whisper, OpenAI has transcribed a million hours of YouTube videos for GPT-4. Model providers can keep following this path but they won’t escape the flattening S-Curve this way. The next great training source for AI models is data produced in a work context. It's of far higher quality than what’s left of public data for training purposes, especially compared to running the dregs of the internet through the transformer mill. We have written a deeply researched piece about how startups can help us make the jump to the next S-Curve: By helping companies tap the brilliance of their people, and by building the next great AI tools with business data. We've identified four areas for startups who want to solve for breaking out of the AI plateau to build something new, useful, and powerful. Take a read, and if you're working on this problem, please get in touch! https://lnkd.in/g6M8UvzF

💰 Funding Alert 💰 Adaptive 💲 Amount Raised: $19M 🚀 Series: Series A 💡 Founded: 2023 👥 Founders: Matt Calvano, Francisco Enriquez, Henry Bradlow 🔍 Problem: The construction industry faces chronic payment delays and cash flow issues due to outdated and manual back-office processes. ✅ Solution: Automates financial management for SMB construction companies using AI, reducing errors and accelerating payments. 🤝 Investors: Emergence Capital, Andreessen Horowitz, Definition, Exponent, 3kvc, BoxGroup, Gokul Rajaram Subscribe to our newsletter launching soon in the comments! 👇 #venturecapital #constructiontech #fintech #AI #automation

Here at Arcee AI, we've been committed to providing tools to enable you to easily customize the best open source #LLMs for your own needs – by leveraging them to create your own Small Langauge Models (#SLMs), using our cutting-edge model merging and training platform. So with the release of Llama-3.1 this week – another major step forward in open source #AI closing in on closed source AI – we're so excited about the possibilities. How do you get started? We show you how in our latest blog. Here's just one of the examples: ✔ Continual Pre-Training involves retraining a language model's next token prediction on your own text data (this allows you to extend Llama's knowledge of tokens within its parameters – not externally, with prompting, reasoning, or RAG). ✔ For Llama-3.1, you can Continuously Pre-Train with a smaller context window (e.g., 8K token stacks) and merge back into the extended context window without knowledge transfer loss. ✔ We scanned Llama-3.1 with Arcee Spectrum (our technique for optimizing training). ✔ We integrated the model into the Arcee Continual Pre-Training routines. ✔ Then, we injected a few text tokens (1.83 million) from our internal Arcee Slack. ✔ We ran Continuous Pre-Training on Llama-3.1-8B. ✔RESULT: even with just that *small sample set of tokens*, we can observe knowledge injection (in other words, the model of Llama-3.1-8B that we've trained is 🧠 MUCH SMARTER 🧠 about the text in Arcee AI's Slack). This is pretty incredible when you think about efficiently creating accurate Small Language Models that can bring 💴 immediate business value 💴 in your organization. Learn more in the full article: https://lnkd.in/eZzDbpCk #NLP #GenAI

Demo software is now a 🔑 part of the sales cycle. Saleo is here to help you understand and strategize how to use demo software to blow away buyer expectations and seal the deal. Check it out!

Wondering if you should get started with Model Merging, leveraging the best open source #LLMs to create your own custom Small Language Models (#SLMs)? No better endorsement of the technique that we have pioneered (and productized via our easy-to-use platform) than these words from the Llama-3 team: 💖 💖 💖 "Oh, and seriously, don't throw out your fine-tuned models -- merge them!!💖 💖 💖 #GenAI #NLP #Llama3

Good brief article by Gordon Ritter and Wendy Lu at Emergence Capital on the need for businesses to own and control their own data, which of course has been baked into the KOS from inception. The language model is actually all but irrelevant (commoditized--take your pick) when compared to high quality data tailored to the business. Knowledge workers do indeed continuously generate new data specific to their role and organization. It would be absolutely foolish to license business data to LLM chatbots or anyone else in my view. Knowledge worker data essentially IS the business -- it represents the value of the organization, products, services, and competitive advantage. A good example of the value of knowledge workers are the recent cases when two teams from LLM firms were 'acquired' by Big Techs rather than buying the company, obviously in part in an attempt to avoid being declined in antitrust review, but also presumably due to significant cost savings. Why pay investors an ROI for taking risk when you can just poach the team? It's much the same with scraped data -- LLM firms and their partners are attempting to transfer the knowledge base of the planet on every topic to a small oligopoly, with a strong preference (if not need) for not paying for it. We have the EAI OS businesses need. No question in my mind. The only question is how best to deliver it.

“AI should be based on human discovery and knowledge and crafted with human-centric attributes of privacy and quality in mind. In the dance of cocreation with AI, humans must always take the lead.” Read more from Gordon Ritter and Wendy Lu in Fast Company: https://lnkd.in/grwxADqF

“Startups that unlock and harness business data will create significant value and tools that enterprises actually want to adopt. The potential for AI in the B2B space is vast and largely untapped.” –Gordon Ritter and Wendy Lu write today about the next wave of AI Innovation in Fast Company

I and my partners at Emergence Capital are excited to be partners with the Adaptive team! The world-class financial tools they're creating for construction firms are sorely needed -- and are already helping their customers run more efficiently and profitably. Thank you, Matt, Francisco, Henry, and the whole team at Adaptive for trusting us to lead your Series A!

The AI wave is plateauing. The next wave is just starting…and it’s 20 times bigger. Enterprise data, especially operational data, captured in the flow of work and from internal subject matter experts represents a treasure trove for AI vendors. The closer vendors get to this treasure trove, the more security, ownership, and privacy become a concern. While there’s a clear technology dislocation in play, the real dislocation is enterprise trust. Great overview of the massive opportunity for enterprise AI by the Emergence Capital team