Scale AI

Today, we’re announcing Scale has closed $1B of financing at a $13.8B valuation, led by existing investor Accel. For 8 years, Scale has been the leading AI data foundry helping fuel the most exciting advancements in AI, including autonomous vehicles, defense applications, and generative AI. With today’s funding, we’re moving into the next phase of our journey: accelerating the abundance of frontier data to pave the road to Artificial General Intelligence (AGI). “Our vision is one of data abundance, where we have the means of production to continue scaling frontier LLMs many more orders of magnitude. We should not be data-constrained in getting to GPT-10.” - Alexandr Wang, CEO and founder of Scale AI. This new funding also enables Scale to build upon our prior model evaluation work with enterprise customers, the U.S. Department of Defense, and collaboration with the White House to deepen our capabilities and offerings for both public and private evaluations. There’s a lot left to do. If this challenge excites you, join us: https://scale.com/careers Read the full announcement: https://lnkd.in/gVBhaPZ5

Hosted by Scale CEO Alexandr Wang and entrepreneur and investor, Nat Friedman, the AI Base Camp brought together the world’s foremost AI leaders and researchers to discuss the challenges and opportunities we face and surface insights for creating the blueprint to develop and implement AI responsibly. Attendees discussed challenges to achieving AGI, the progress towards and need for better model evals, and what’s on the horizon for the industry.

On the latest episode of The Twenty Minute VC from Harry Stebbings and 20VC, Scale CEO Alexandr Wang touches on how data, not compute, is the main bottleneck to improving foundation model performance. He also emphasizes the critical role of AI as a military asset, with the potential to surpass nuclear weapons and discusses the evolving dynamics of AI competition between China and the US. 👉 Tune in here: https://lnkd.in/g8pqE2eH

“AI is going to get better at certain things that allow us to have better output without just adding bodies to the staff. If you have fewer people who spend their days dumpster diving in the share drive to uncover information or understand what’s going on, there's a lot of value there.” Former Marine and Scale Head of Federal Delivery Dan T. joined the Phoenix Cast, a veteran-run podcast about cyber, tech, and innovation in the military, for a discussion on “Real Actual AI” applications in the military. Tune in to hear him discuss why AI assistants augment military analysts and planners instead of replacing them, how to implement AI in the field, and the limitations of AI for certain military applications. Listen to the episode here: https://lnkd.in/gS34HKzD

Say hello to Scale Evaluation for the public sector. Ensure AI can be trusted for critical missions and workflows. As the government looks to adopt and scale their use of AI, it is critical that they acquire and use safe, ethical and responsible AI systems. Scale Evaluation equips public sector organizations with the tools required to fulfill the requirements outlined by the Biden-Harris Administration’s AI Executive Order and OMB Circular. Agencies can: • identify AI risks • simulate and measure AI in real-world context • understand model performance • ensure compliance with federal mandates In the latest blog from Scale, we dive into Scale Evaluation for the public sector and how our work with organizations like DoD Chief Digital and Artificial Intelligence Office, National Institute of Standards and Technology (NIST), and Cybersecurity and Infrastructure Security Agency shaped Scale’s approach to test and evaluation: https://lnkd.in/gsCuVXCZ

✴ Upcoming Webinar Alert! We're discussing key AI adoption strategies for enterprises from our recently released AI Readiness Report which surveyed over 1,800 AI practitioners and organizations. ➡ Topic: AI Readiness Report 2024 - Insights and Strategies for Enterprise Success 🗓 When: June 20, 10AM PT 👥 Who: Raghvender Arni, Director of Worldwide GenAI GTM Cross Portfolio, Amazon Web Services (AWS) and Vijay Karunamurthy, Field CTO Scale AI 🎓 What you’ll learn: • latest trends in enterprise AI adoption • best practices for optimizing AI performance and handling data transformations • how to ensure model safety and reliability using industry standards and benchmarks 👉 Save your spot today! https://lnkd.in/geSm9t4N 

“It is imperative for our nation’s economic and national security that the United States continues to lead the world in AI.” Scale Managing Director and former Chief Technology Officer of the United States, Michael Kratsios joined the Stanford Institute for Human-Centered Artificial Intelligence (HAI)'s five-year anniversary celebration along with Condoleezza Rice, Susan Rice, and Jason Matheny. They discussed strategic competition, cooperation, and regulatory strategies in the age of AI. Watch: https://lnkd.in/gGmzA-ZA

Last week Scale's SEAL team released leaderboards ranking frontier models across domains like coding and math. Let’s dive into the methodology 👇 1️⃣ We start with human evaluations. Human raters compare the responses of two models to the same prompt and rate which is better along a multitude of domains / capabilities 2️⃣ We then use the Bradley-Terry mathematical model to re-weight these ratings and then convert them to an Elo-scale. This is the same method used by Chatbot Arena. The idea is to find the set of Elo ratings that best explains the observed outcomes of the model vs model matchups, based on maximum likelihood estimation. At the end of the process, each AI model gets a skill rating that represents its overall performance relative to the other models. 3️⃣ From there, we use a technique called bootstrapping to estimate confidence intervals that help us understand the reliability and statistical robustness of our Elo-scale Bradley-Terry ratings. Bootstrapping is a resampling technique that allows us to assess the variability of our estimates by repeatedly sampling from the data with replacement. 📊 And that’s all! After bootstrapping, we get a ranked leaderboard of AI models along with confidence intervals around each model's skill rating. Doing so allows us to track and compare the capabilities of different language models objectively over time, enabling us to maintain a neutral and true measure of model performance. Learn more about our methodology and see the final leaderboard results here: https://lnkd.in/g7N_Hs9p

Today, Cisco announced a $1B global investment fund to expand and develop secure and reliable AI solutions. Over the past several years, we’ve made over 20 AI-focused acquisitions and investments. 🚀 As part of today's news, we welcome Cohere, Mistral AI, and Scale AI to our Cisco Investments portfolio. To learn more, check out VP Janey Hoe's conversation with Cisco EVP and Chief Strategy Officer, Mark Patterson, and Cisco Investments' SVP, Derek Idemoto, about our AI strategy.👇 🔗 https://hubs.ly/Q02zkRlN0 #CiscoInvests #AI

We’re celebrating the one-year anniversary of Donovan today, Scale’s mission-focused AI application for the public sector! 🎉 Over the past year, Donovan has used frontier LLMs to enable government organizations to make faster, more informed decisions and minimize time searching for information. As we celebrate this milestone and look back at Donovan’s major product updates, we’re sharing what’s next for Donovan: https://lnkd.in/gbEudXTW