By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.

Celebrating the AI Pioneers at Decibel's Inaugural AI Summit

We were excited to host our inaugural AI Pioneers Summit in SF on Thursday, October 26, with our friends at Latent Space. Our event was the first to celebrate the technical leaders at the forefront of deploying AI and LLMs in the enterprise today and to learn from those who have overcome the technical and commercial complexity required to ship successful products at scale. We were excited to host over 200 product and engineering execs from 160 companies at our event, and to recognize those who have forged ahead to turn the promise of AI into products that are now widely used throughout the industry.

We have summarized many of the most important findings below.

Our #1 takeaway for all is that deploying AI requires technical acumen, strong leadership, and the courage to ship new products in a rapidly changing domain. It was inspiring to hear from those who have tackled these challenges to pioneer the field, and we look forward to supporting and expanding this community in the years to come.

AI Pioneer Award Recipients

We recognized several leaders in our community for their contributions to the field of AI, including:

Innovation in AI Engineering:

Deploying AI at Scale:

Setting Standards for Responsible AI:

Bringing AI to the Edge:

Largest Gen AI Acquisition:

Chains, Agents, and the Autonomous Enterprise

The field of AI Engineering encompasses traditional software engineering and the rapidly evolving field of LLMs. One of the most promising and provocative areas of innovation is the role that “AI agents'' will play in the design of every modern software application. We were excited to discuss the emerging trend of autonomous software agents in the enterprise with  Harrison Chase (Founder and CEO, Langchain), Jim Fan (Distinguished AI Scientist, Nvidia), and Silen Naihin (Founding AI Engineer, AutoGPT):

  • Most enterprises are comfortable with building applications with multi-step LLM prompts and chains - they are ideal when you are expecting a fixed set of actions, and want a fixed set of responses.
  • Autonomous agents are a possibly more powerful way of handling a wider range of inputs and outputs - they are ideal when you need LLMs to reason but they require a more robust set of test cases to ensure accuracy.
  • In the future, LLM agents may be capable of acquiring skills in order to improve their accuracy and performance. Jim Fan’s “Voyager” paper first introduced the concept of agents that can learn skills, such as mining gold in Minecraft. We expect this concept to ultimately find its way into the enterprise world.  

How AI Works at Scale

Shipping AI-enabled products requires incredible technical acumen, a very strong leadership team, and the navigation of complex commercial requirements including evolving government regulation. There are few people in the industry who have been able to ship. We were excited to be joined by Cheryl Ainoa (EVP Technology, Walmart), Neha Batra (VP Engineering, GitHub), and Clara Shih (CEO, Salesforce AI) to discuss their journey from designing to deploying LLMs to scale to millions of end users:

  • Most large enterprises try multiple foundation models including commercial, open source, and specialized / home grown options. Model selection is based on the needs of the end product.
  • Every AI-enabled application has to optimize for three variables - (i) accuracy, (ii) speed, and (iii) price. Every AI architect must weigh the trade-offs - sometimes you can sacrifice latency for price or vice versa, but rarely can you sacrifice on accuracy.
  • Technology leaders need to embrace responsible and ethical AI in order to ship widespread commercial products. There are no shortcuts, and no way forward without awareness and buy-in across the organization.

Creating >100 “GPU Cities” in 18 months

How do you deploy GPU capacity to more than 100 data centers around the world to accelerate AI inference at the edge? We were excited to be joined by Matthew Prince (Founder and CEO of Cloudflare) to discuss the recent launch of their Workers AI platform and his ambition to enable developers to use serverless GPU inference on Cloudflare’s global network:

  • Why focus on the edge? Deploying AI-enabled applications at scale will ultimately come down to the cost and speed to serve inference. Creating higher utilization of scarce GPU resources at the edge should drive down cost
  • Where we run inference will matter - customers and countries that care about data sovereignty will need to localize their AI products
  • How do you put AI as close to every user as possible? The answer is not just in running a global network that can be flexibly expanded, but also traveling around the world with “suitcases full of GPUs” to get them installed where needed!

Who Will Win the War for AI?

Are there clear winners and losers in the war for AI? We could not think of anyone more qualified to answer this question than Ion Stoica (Chairman and Co-Founder of Databricks) whose numerous contributions to the field including Spark have made him one of the major architects of the modern era of AI. He shared many words of wisdom:

  • The open source community will continue to have a huge role in shaping the AI landscape. As LLMs consume expensive compute resources, we will increasingly see open source communities and major cloud providers continuing to partner to ensure that these projects are competitive with commercial offerings
  • There will be multiple foundation model providers for the foreseeable future. AI is too important and strategic for cloud providers, internet companies, enterprises, and governments - it is in everyone’s interest to ensure that is more than one commercial winner in the end
  • We are in the early innings of a cloud-computing paradigm shift. Because AI workloads are data and GPU-heavy, we will likely see the cloud computing stack be more tightly integrated. We are indeed headed back to the days of supercomputers!

Countdown to the Latent Space Launchpad!

The Latent Space podcast has crossed >500,000 downloads this year and is the #1 community for AI engineers who are pushing the frontier of LLMs in production. We have gotten a lot of interest from early founders in expanding our programming to help startups, and are excited to create our first Latent Space “Launchpad”! We want to build a community for founders, future founders, and early adopters in AI who are interested in starting and launching companies at the intersection of AI and the enterprise. Our program will help with company formation, pre-seed financing, early customer discovery, and ultimately a product launch. If you are interested in joining our program as a startup, a mentor, or a design partner please indicate your interest here. We look forward to seeing you all at our kick off event in January!