Systems 4 Ushering in the AI enabled era today. Our customers move faster. They deploy the power of AI and Deep Learning to discover new opportunities, unlock new revenue, and boost efficiency. Industry leading AI-as-a-Service Pretrained Models Next Generation Architecture Chris Ré Professor CS Stanford University Kunle Olukotun Professor EE/CS Stanford University Rodrigo Liang CEO AI is here. SambaNova was founded to solve challenges in applying AI technology to drive impact in unprecedented ways. We are democratizing AI, by providing purpose built Deep Learning solutions, delivered as a service, to enable the future of AI today.
Snapshot 5 GLOBAL PRESENCE HEADQUARTERS IN PALO ALTO, CA AI/ML SOFTWARE-DEFINED HARDWARE 2017 ESTABLISHED THE COMPANY AI PROCESSOR (RDU) OPTIMIZED FOR DATAFLOW 500+ EMPLOYEES CLOUD-READY HIGH-VALUE MISSION-CRITICAL ENTERPRISE WORKLOADS SERIES D FUNDED $1.1B RAISED $5B VALUATION INTEGRATED AI PLATFORM OFFERED AS A SUBSCRIPTION
is a Generational Shift 6 AI TODAY MOBILE 2008 WEB 1998 SCALE-OUT COMPUTING EMBEDDED DEVICES DATAFLOW COMPUTING DEVELOPER DRIVEN CONSUMER DRIVEN DATA DRIVEN
CEOS BELIEVE AI WILL HAVE A LARGER IMPACT THAN THE INTERNET 63% ADDED TO THE GLOBAL ECONOMY IN THE NEXT DECADE BY AI TRILLION $15 18 BY 2025, EVERY CONNECTED PERSON WILL HAVE A DIGITIAL INTERACTION EVERY… SECONDS IDC, The Digitization of the World PwC’s Annual Global CEO Survey, McKinsey Report: The Executive’s AI Playbook PwC’s Annual Global CEO Survey, McKinsey Report: The Executive’s AI Playbook Artificial intelligence in Business by McKinsey & Co
Study: Key Learnings 13 plan to allocate more than $100 million of IT budget toward strategic technology goals. 70% Main reason why your organization is investing in AI.ML? 40% 25% 22% Powers innovation Operational efficiency Keep up with competitors 75% say improving access to deep learning is very important for fostering competition and innovation in their industry. What are your organization’s biggest challenges in scaling your AI/ML efforts? 50% Restrictive computing architectures 35% Lack of trained talent 28% Difficulty customizing models
is here. AI is here. 28% of enterprises are deploying AI at Scale 72% are being left behind. What are you going to do next? Reach out to us, the leader in AI deep learning to jump start your journey today.
is here. AI is here. The state of the art AI is increasing at 10x per year Is your organization prepared? Reach out to us, the leader in AI deep learning to jump start your journey today.
Dawn of the Deep Learning Era 18 • “Industrialization of Computer Vision” • Highly Flexible, General-Purpose Models • Surpassing Human Level Performance at Some Tasks Dawn of the Deep Learning Era AI 2.0 Digital Transformation & Machine Learning AI 1.0
Deep Learning Deployment Gap AI transformation: Time to hire a team, build infrastructure, train and deploy model Increase in compute requirements 2 32X Increase The Deep Learning Deployment Gap Increase in model size3 15X Increase 1: “Most AI transformations take 18-36 months to complete” from “Building the AI-Powered Organization”, Harvard Business Review, 2019 2: Based on datapoint that compute demands of top models doubled every 3.4 months between 2012-2018 from AI and Computer research paper from Georgetown Center for Security and Emerging Technology, January 2022 3: Based on the increase of the number of parameters of large language models, which have increased at a pace of roughly 10X per year 18 months¹ Increase 10X every 12 months Increase 2X every 3.4 months Compounding Compounding
SambaNova Difference: AI is here. 21 Advanced Deep Learning Capabilities • We solve the hard problems • We solve the large problems • We solve the scale problems Faster Time to Value • Deploy in weeks, not years • Pre-trained models reduce the need for heavy data-science effort Reduced Complexity • Simplified and ease of use Purpose Built for Innovation • Ability to handle the pace of AI innovations: 10X increase in compute requirements and model size every year
Industry’s Most Powerful Deep Learning Platform Dataflow-as- a-ServiceTM High Resolution Image Processing Complex 3D Image Analysis Real Time Video Analysis Sentiment Analysis Document Classification Named Entity Recognition COMPUTER VISION AS-A-SERVICE NATURAL LANGUAGE PROCESSING AS-A-SERVICE Public Sector BFSI Manufacturing Energy Healthcare Customer Intent Product Recommendation Personalization RECOMMENDATION SYSTEMS AS-A-SERVICE Pre-trained models | Domain Specific Fine Tuning Integrated Hardware/Software AI Platform Deep Learning Models-as-a-Service, Deployed in Weeks Industries Powered by Reconfigurable Dataflow Unit (RDU)
the weeks since its arrival, GPT-3 has spawned dozens of other experiments that raise the eyebrows in much the same way. It generates tweets, pens poetry, summarizes emails, answers trivia questions, translates languages and even writes its own computer programs, all with very little prompting.” State-of-the-Art AI is the Province of the Elite Few NYTimes, 11/24/2020 ML & systems PhDs Integration & deployment Massive data center resources
RDU and SambaFlow Are the New Engine for AI 27 The old way: Kernel-by-kernel Bottlenecked by memory bandwidth and host overhead Memory Data & Weights Result Memory Data & Weights Result Memory Data & Weights Result Memory Data & Weights Result Memory Batch & Weights Memory Result Conv 1 Cache Cores Conv 2 Cache Cores Pool Cache Cores Norm Cache Cores Sum Cache Cores Content Switching Content Switching Content Switching Content Switching
RDU and SambaFlow Are the New Engine for AI 31 Weights Weights Sample Conv 1 Pool Conv 2 Norm Sum The Journey of a PyTorch App With SambaFlow SambaFlow Spatial Compilation Whole program optimization reconfigures the RDU to be optimal for this model…instantly
RDU and SambaFlow Are the New Engine for AI The compiler handles partitioning so no changes are required to go from 1 socket to 1000 The Journey of a PyTorch App With SambaFlow
full-stack state-of-the-art AI deployable instantly 33 GPT3 Conv3d Rescale ... Dataflow-as-a- Service On-prem Hosted Cloud BFSI Solution PubSec/ Govt solution Healthcare solution Oil and Gas Solution Manufacturing Solution No/low code access to latest models on scalable infrastructure Vertical Solutions use common Software platform components Platform includes enterprise AI functions like Mlops, manageability, Observability, BI, labelling mgmt. Base Models, compiler and compute functions Hardware
AI world is constrained in the Habitable Zone Fine grained (Too Hot) Large Models (Too Cold) Performance Habitable Zone GPT-3 Large embeddings True-Res CV BERT-Large ResNet RNNs Sparsity AI4S
has expanded the Habitable Zone Fine grained Large Models GPU-optimized GPT-3 Large embeddings True-Res CV BERT-Large ResNet RNNs Sparsity AI4S 1x GPU Performance
SambaNova Solutions: Enterprise grade AI 39 Banking, Financial Services and Insurance • McKinsey estimates that the potential annual value unlocked by Advanced AI is $683.7 Billion. • Disruption from fintech, DeFi and crypto. • Large scale NLP problems for customer service, fraud prevention, anti-money laundering, document processing. Healthcare • McKinsey estimates that the potential annual value unlocked by Advanced AI is $306.9 Billion. • Disruption from COVID, cheap sequencing, pervasive data sharing. • CV and NLP for cancer screening, drug discovery, Point of care diagnoses. Manufacturing • McKinsey estimates potential annual value unlocked by Advanced AI is $491B Billion • Disruption from great resignation, supply chain disruptions, remote work and robotics. • CV for defect detection, time series / recommendation systems for supply chain, predictive maintenance, and Agritech. Energy and Sustainability • BCG claims applying AI to corporate sustainability will add $1.3T -$2.6T of value. • Disruption from Increasing Energy demand, climate change, net zero pledge. • RL for efficient datacenters, fusion reactors. 3Dconv for efficient oil extraction.
grade AI Use cases in Banking, Financial Services and Insurance • Customer fulfilment and retention - Chatbots for faster MTTR in customer service, customer churn prediction and prevention. • NLP for insurance cost reduction - Can take into account data points from vast amounts of data that humans simply can't. • AI based credit scoring - Increases speed, accuracy of credit risk models and gives quick access to credit for BNPL. • Algorithmic trading - Reduce compliance risks, analyze more signals, aid in alpha generation strategies, • Roboadvisers - $8T AUM in wealth management is slowly being moved to roboadvisers. • AI based fraud detection - 5% of global GDP transactions are money laundering and $80-100B lost in fraud transactions. KYC scenarios • AI based Regtech - Respond to regulatory requirement quickly and effectively. To become an AI-first institution, a bank must streamline its capability stack for value creation Reimagined engagement AI-powered decision making Core technology and data Operating model
grade AI Use cases in Healthcare • Disease detection – Data analysis using NLP for diagnosis, high-res imaging analysis for disease diagnosis in radiology, pathology image analytics. • Drug Discovery – ML analysis of patient data, clinical trials, publications for new drug applications. Protein folding prediction. • Targeting and detection of evolving pathogens – detect and reduce antibiotic resistance • Real time patient monitoring • Healthcare supply chain management • Dynamic pricing of health insurance • Supplementing Underequipped point of care in developing/underdeveloped/disaster stuck places
grade AI Use cases in Manufacturing • Defect detection – High-res imaging to detect defect on assembly and QA lines. • Predictive maintenance – time series models to predict machine failures in advance. • Yield optimization/enhancement - RL used for robotic control, extruder operations and chemical process optimization. • Autonomous machine control – various control inputs can be optimized in a closed loop system for reduced energy, better quality, sustained reactions. • Generative design – Use generative models to design better than traditional topology optimization. • Inventory management – predictive models for inventory replenishment Share of respondents
SiteGround - Reliable hosting with speed, security, and support you can count on.
itpresstour
sakaik
brainpadpr
edeandrea
ry
subroh0508
line_developers_tw
lycorptech_jp
seafay
gotok365
uhyo
junki
yamaguchitk333
reverentgeek
elsirapls
notwaldorf
jonoalderson
geoffreycrofte
varn
lfama
eileencodes
jdejongh
inesmontani
schroneko
paigeccino
