TD SYNNEX Corporation · NYSE:SNX
TD SYNNEX — IBM — IBM Innovation Day Lausanne (2023)
70 slides from TD SYNNEX's conference presentation, 2023
Slide 1: AI for Business
AI for Business Philipp Späti IBM Distinguished Engineer Director Technical Sales and SW Lab Services Switzerland philipp.spaeti@ch.ibm.com
Slide 2: AI in the news
@ - Reviews - Others Stable Diffusion Gets an Upgrade —Here’s How Its AI Model Stacks Up Against MidJourney Stable Diffusion XL v0.9 is out. And the results are impressive. NZZ [Lamaconina amoral Virtuellen Stars gehért die Zukunft der Musik. In der Branche wollen das viele nicht wahrhaben, aber sic
Slide 3: Video link: super powers of ChatGPT
https://www.rsi.ch/play/tv/patti-chiari/video/i-super-poteri-di-chat-gpt-?urn=urn:rsi:video:1885174 3
Slide 4: Foundation models establish a new paradigm for AI capabilities
Foundation models establish a new paradigm for AI capabilities Traditional AI models Foundation Models Tasks Enhanced capabilities: Training Tasks Summarization A AI1 Prompting Conversational Knowledge B AI2 Content Creation AI3 Enterprise Code co-creation Massive Data in- AI4 Pre Trained Foundation
Slide 5: Three proven, high impact use cases as starting points
The impact of generative AI |The opportunity Three proven, Talent Customer service App modernization high impact use cases as starting points with IBM 40% 70% 30% Improvement in HR Contact center cases Productivity gain in productivity contained by conversational AI application modernization • Talen
Slide 6: What will Generative AI bring us
What will Generative AI bring us Controllable Output Generation • Trusted Data Acquisition Every company needs an AI strategy • Domain Adaptation • Prompt Engineering • Bias & Discrimination • Bias Mitigation • Privacy & Confidentiality • Human-in-the-loop • Post-Generation Validation • Transparency
Slide 7: Introducing watsonx
What IBM offers Introducing… watsonx
Slide 8: watsonx: The platform for AI and data
The platform watsonx.ai watsonx.data watsonx.governance for AI and data Train, validate, tune and Scale AI workloads, Enable responsible, deploy AI models for all your data, transparent and explainable watsonx anywhere data and AI workflows A next generation enterprise Fit-for-purpose data store End
Slide 9: Put AI to work with watsonx
Put AI to work with watsonx Scale and accelerate the impact of AI with trusted data on hybrid cloud watsonx.ai watsonx.data watsonx.governance Train, validate, tune Scale AI workloads, for Enable responsible, transparent and and deploy AI all your data, anywhere explainable data and AI workflows mod
Slide 10: watsonx.ai is based on foundation models, multi-model on multi-cloud
watsonx.ai is based on foundation models that are multi- model on multi-cloud with no lock-in Use IBM models Use IBM + open Try Tune Build and Deploy* + other models Evaluate *on any cloud Bring your BYO own models Model 10
Slide 11: watsonx.data
What IBM offers watsonx.data Scale AI workloads, for all your data, anywhere A fit-for-purpose data store based on an open lakehouse Public cloud architecture Cloud warehouse Cloud data lake • Access all your data through a single point of entry across all clouds and on-prem environments Hybrid IBM
Slide 12: Trustworthy Technology: watsonx.governance
What IBM offers watsonx.governance Tamper proof cap (Adversarial) Trustworthy Technology: Robustness What todays Clear Glass Bottle Transparency AI Governance Food label Factsheets / can learn from the Explainability Pure Food history of producing and Drug Act Sanitary Production Trustworthy Tomato
Slide 13: watsonx.governance
watsonx. governance Enable responsible, transparent and explainable AI Govern across the AI Manage risk and protect Help adhere to workflows lifecycle by automating reputation by regulations by and consolidating tools, automating workflows to translating growing An end-to-end toolkit applications an
Slide 14: Introducing Granite
Introducing Granite granite.13b Intended Use: Available today in English language tasks including watsonx.ai generation, summarization, classification, extraction, and question answering LLM = Architecture + Data + Compute + Alignment Model Architecture Data Alignment granite.13b • 13b 1T tokens of
Slide 15: Why IBM? Open, Trusted, Targeted
What IBM offers Why IBM? Open IBM’s AI is based on the best open technologies available Trusted IBM’s AI is transparent, responsible, and governed Targeted IBM’s AI is designed for enterprise and targeted at business domains Empowering IBM’s AI is for value creators, not just users
Slide 16: Registration & More Information
Registration & More Information: https://www.ibm.com/community/ibm-techxchange-summit- emea/
Slide 17: Slide 17
Slide 18: The future of intelligent applications is open and now
The future of intelligent applications is open and now Red Hat makes AI/ML workloads more productive and faster Mohamed Ramdani Senior Solution Architect 20
Slide 19: What we'll discuss today
Agenda What we’ll ▸ The potential and challenges of AI/ML discuss today ▸ Adding intelligence to your workloads ▸ Why Red Hat? ▸ Edge use cases for AI 21
Slide 20: The potential and challenges of AI/ML
The future of intelligent applications is open and now The potential and challenges of AI/ML 22
Slide 21: What are the challenges impeding your progress?
The potential of AI/ML What are the challenges that are impeding your progress? Bridging environments How do you bridge the data scientist and app developer environments? 87% Infrastructure complexity How do you make the complexity of infrastructure invisible? 87% of data science projects Realizing
Slide 22: Complexities of operationalizing models
Realizing Value from AI/ML Complexities of operationalizing models "a consistent application platform for the management of existing, data machine modernized, and cloud-native verification resource monitoring management data collection applications that runs on any cloud." configuration serving larg
Slide 23: Adding intelligence to your workloads
The future of intelligent applications is open and now Adding intelligence to your workloads 25
Slide 24: Enterprises use a mix of open source and cloud-based software for AI
AI/ML Background and Challenges 69% of enterprises use a mix of open source and cloud-based software to power AI initiatives. 26 Source: Data collected from Dec. 16, 2020 - Jan. 22, 2021 | Respondents: 100 enterprise IT and data leaders
Slide 25: Enterprises deploying on hybrid cloud today
Open source and cloud-based software to power AI initiatives 66% of enterprises are deploying on hybrid cloud today. 27 Source: Data collected from Sep 5, 2022 - Feb 14, 2023 | Respondents: 300 enterprise IT and data leaders
Slide 26: AI adopters are keen on using containers
AI adopters are keen on using containers to improve AI workloads 94% of AI adopters are using or plan to use containers within one year1 28 Q: Are you using a containerized environment to deploy machine learning models to production? (n=493) Q: What benefits [do or will] containerized environments b
Slide 27: Enter MLOps
Enter MLOps MLOps incorporates DevOps and GitOps to improve the lifecycle management of the ML application TRAIN VALIDATE DEVELOP APP BUILD TEST SERVE CODE Cross Functional Git as Single Repeatability Automation Security Observability Collaboration Source of Truth DEVELOP DRIFT/OUTLIER MONITOR ML CO
Slide 28: From DevOps to MLOps
From DevOps to MLOps codifying problem data collection feature model model training model model model serving monitoring, and metrics and cleaning engineering development and pipeline tuning validation deployment pipeline validation codifying project problem data requirements collection architecture
Slide 29: Use the power of enterprise-ready open source
Adding intelligence to your workloads Use the power of enterprise-ready open source Most of innovation in AI/ML is happening in open source communities The AI/ML ecosystem is complex ▸ Technologies are rapidly evolving ▸ Vendor landscape is constantly changing ▸ No single vendor can provide everythi
Slide 30: Tools to get AI/ML models into production faster
Adding intelligence to your workloads The tools and technology to get AI/ML models into production faster A self-service platform for data scientists using the DevOps/MLOps discipline Gather and prepare data Develop model Deploy models in an application Model monitoring and management Data engineers
Slide 31: Hybrid MLOps platform
Why Red Hat? Model development Conduct exploratory data science in JupyterLab with access to core AI/ML libraries and frameworks including TensorFlow and PyTorch using our notebook images or your own. Model serving and monitoring Deploy models across any cloud, fully managed, and self-managed OpenSh
Slide 32: Why Red Hat?
The future of intelligent applications is open and now Why Red Hat? 34
Slide 33: Data science and edge services solutions
Why Red Hat? Data science and edge services solutions Establish Adopt Deploy a data science platform MLOps in your organization models to the edge ▸ Data science platform on-premise or ▸ Establishes a foundation for ▸ Flexibility and freedom to run edge using a managed service considering production
Slide 34: Tradeoffs at the edge
The optimal location of these services is Tradeoffs at the edge STRICTLY INTERNAL ONLY determined by the tradeoffs between the constraints. ion ect onn ec Last mile network abl reli Un k or w et il en La m st st La mile ne tw or k enterprise k device edge far edge near edge wor net m ile n Unreliabl
Slide 35: Come visit us at the RedHat booth
Come visit us at linkedin.com/company/red-hat the RedHat youtube.com/user/RedHatVideos booth facebook.com/redhatinc twitter.com/RedHat 37
Slide 36: IBM Quantum Computing
IBM Quantum Computing Sasha Lazarevic IBM Quantum Ambassador IBM Quantum © 2023 IBM Corporation
Slide 37: We are a small planet with big problems to solve
We are a small planet … with big problems … to solve IBM Quantum © 2023 IBM Corporation
Slide 38: Slide 38
Slide 39: Problems quantum could help address
Improved nitrogen-fixation New catalysts to make CO2 process for creating ammonia- conversion into hydrocarbons based fertilizer more efficient and selective New classes of antibiotics to Better financial models to improve stability, predictability counter the emergence of multidrug-resistant bacter
Slide 40: Why quantum?
Why quantum? Problems we can’t adequately address today Problems we hope to Problems we can address with quantum address today and classical classically computing Despite how sophisticated digital “classical ” computing has become, there are many scientific and business problems for which we’ve bare
Slide 41: Computing with caffeine
Computing with caffeine If our best classical computers are so powerful, shouldn’t we be able to perfectly simulate molecules and chemical reactions? This would allow us to accelerate discovery of new compounds and processes for healthcare, materials, alloys, and sustainable energy creation. Let’s c
Slide 42: Computing with caffeine
Computing with caffeine We would need approximately 1048 bits to represent the energy configuration of a single caffeine molecule at a single instant in a classical computer. This is 1 to 10% of the total number of atoms in the Earth. 1048 = 1,000,000,000,000,000, 000,000,000,000,000,000, 000,000,00
Slide 43: Quantum applications span three general areas
Quantum applications span three general areas Simulating Quantum Systems Artificial Intelligence Optimization / Monte Carlo Improved battery Genomic Analysis Options pricing Product Portfolio Manufacturing Supply materials Chemical product design Derivatives Pricing Optimization Process Chain Manufa
Slide 44: ExxonMobil
ExxonMobil Maritime Routing’s Mind-Boggling Math In 2021 more than 500 LNG (liquified natural gas) ships are used to transport critical fuel supplies across the oceans. Together, they make thousands of journeys per year to destination ports where the LNG is deployed to power critical infrastructure.
Slide 45: JP Morgan Chase
JP Morgan Chase Quantum Computing for the Financial Services Industry Recently, JPMC and IBM used Quantum Amplitude Estimation, a Monte Carlo-like sampling algorithm, to compute European European derivative pricing on a Quantum computer implements option pricing, pricing path depend options, the Bla
Slide 46: CERN
CERN Quantum Machine Learning to understand what sews the universe together CERN’s partnership with IBM Quantum seeks new ways of finding patterns in data of the Large Hadron Collider. A recent collaboration with IBM scientists involves the detection and analysis of the Higgs boson, a recently disco
Slide 47: IBM Quantum – On the cloud since May 2016
IBM Quantum – On the cloud since May 2016 Over 460,000 registered users have run … over 2 TRILLION hardware quantum circuits in total, and users run … over 4 BILLION hardware quantum circuits on a typical day on ... more than 25 quantum computing systems on the IBM Cloud, and written over 1750+ scie
Slide 48: Development Roadmap
Development Roadmap Executed by IBM On target 2019 2020 2021 2022 2023 2024 2025 Beyond 2026 Run quantum circuits Demonstrate and Run quantum Bring dynamic circuits to Enhancing applications Improve accuracy of Scale quantum applica- Increase accuracy and on the IBM cloud prototype quantum programs
Slide 49: IBM Quantum Computation Centers (QCC)
IBM Quantum Computation Centers (QCC) Centers with dedicated Quantum Systems committed to advancing industry-specific initiatives or regional quantum ecosystems IBM Quantum Fraunhofer University of Tokyo Cleveland Clinic PINQ2 Yonsei BasQ datacenter in NY Dec 2020 Jun 2021 Mar 2023 Projected 2023 Pr
Slide 50: A snapshot of global collaboration within the IBM Quantum Network
A snapshot of global collaboration within the IBM Quantum Network IBM Quantum Keio University JPMorgan Chase Qiskit Advances Publishes Applies the Contributed method, algorithm paper method to code into open publishes with on Amplitude options pricing source Qiskit Mizuho and Estimation. and publish
Slide 51: 4 distinct IBM Quantum offerings
4 distinct and simple IBM Quantum offerings to enable clients on their Quantum journey Open Plan Free Qiskit Runtime service on a limited set of small quantum systems and simulators. Available to all. Pay-As-You-Go Plan Qiskit Runtime service on IBM Cloud with access to two 27 Qbit IBM quantum syste
Slide 52: IBM Quantum Accelerator
IBM Quantum Accelerator Our offering to prepare your organization for quantum advantage... Business readiness: Define your company’s compelling case for quantum. Set priorities to capture new opportunities and minimize risk of competitive Use case framing & path to Readiness assessment & Guided prot
Slide 53: Quantum Computing and the Quantum Threat to Digital Security
Quantum Computing and the Quantum Threat to Digital Security Dr. Luca De Feo IBM Research Europe | Zurich
Slide 54: Why Quantum Computing?
Why Quantum Computing? Problems we can’t Applications: Chemistry adequately address today Materials Machine learning Optimization Problems we hope to Problems we can address with quantum address today and classical classically computing Simulating quantum mechanics Factoring Despite how sophisticate
Slide 55: A mathematical warm-up
A mathematical warm-up… Multiplication Challenge: find product Difficulty 7·3 =? Very simple Factorization Challenge: find prime factors Difficulty 91 =?·? Pretty simple IBM Research / © 2023 IBM Corporation 57
Slide 56: A mathematical warm-up
A mathematical warm-up… Multiplication Challenge: find product Difficulty 15050852373840701695841662 13890227180513831490647757 55272608400921082198800517 64178654364320490994679695 02049474405698177982391685 04616070004815403073796078 · =? 90672334150961421847090241 96495907462695502861006886 < 1se
Slide 57: Today's cryptography relies on hard mathematical problems
Much of today’s cryptography relies on hard mathematical problems Public key encryption • Digital signatures • Key exchange algorithms RSA • DSA • ECC • ECDSA • DH Factorization Challenge: find prime factors Difficulty 23227875644355491648343614430281496129940316847274172637862 Most powerful 9043050
Slide 58: Cryptography can be broken with a strong quantum computer
Much of today’s cryptography relies on hard mathematical problems … can be broken with a strong Quantum Computer Public key encryption • Digital signatures • Key exchange algorithms RSA • DSA • ECC • ECDSA • DH Factorization Challenge: find prime factors Difficulty 2322787564435549164834361443028149
Slide 59: Q-Day: What will attackers be able to do?
Q-Day: What will attackers be able to do? Q-Day (availability of “cryptographically relevant” Quantum Computer) …before …after Harvest Decrypt Disrupt Manipulate confidential data lost or harvested daily business with digitally signed confidential manipulation and contracts and legal historical data
Slide 60: What is at risk?
What is at risk? Infrastructure update/replacement cycles Internet Passports – 10 years from issue Hyper-text Transfer Protocol (HTTP/TLS), Secure Shell (SSH), Road Vehicles – 15-20 Years Domain Name Service (DNS), Virtual Private Networks (VPN) Critical Infrastructure – 25-30 Years Digital Signatur
Slide 61: When is Q-Day?
When is Q-Day? IBM Quantum Status and Roadmap Physical Qubits Required to break RSA 2048 2012: 1,000,000,000 IBM “Condor” (1121 / 2023) 2017: 230,000,000 IBM “Osprey” (433 / 2022) 2019: 170,000,000 IBM “Eagle” (127 / 2021) 2019: 16,000,000 2021: 13,426 * Elliptic Curve Cryptography (ECC) requires fe
Slide 62: The good news
The good news Quantum safe Governments issue Awareness of the cryptography exists advisories and extent and urgency and gets standardized directives is growing … and first quantum-safe systems are already available! IBM Research / © 2023 IBM Corporation 64
Slide 63: NIST standardization for quantum-safe cryptography
NIST standardization for quantum-safe cryptography 2016 … 2019 2020 2021 2022 2023 2024 Call for Final Standardization Proposals Round 2 Round 3 Early 2024 Deadline for submissions NIST algorithm selection Public-key Encryption → CRYSTALS-Kyber Key establishment → CRYSTALS-Dilithium Digital Signatur
Slide 64: Governments Advisories and Directives
Governments Advisories and Directives Commercial National Security Algorithm Suite 2.0 NCSC (UK) CNSA 2.0, Sept 7, 2022 National Cyber Security Centre 2) ‘Large organisations should factor the threat of quantum Algorithms recommended by NSA as computer attacks into their long-term roadmaps.’ cryptog
Slide 65: Cryptography is everywhere, except in an inventory
“Cryptography is everywhere, except in an inventory.” IBM Research / © 2023 IBM Corporation 67
Slide 66: How to manage cryptography migration? Two distinct use cases
How to manage crytpography migration? Two distinct use cases. IT Operations CIO / CISO / Cryptographic Software CRO / CDO SaaS Posture Network Documents PKI Data Line of Business Devices/IoT “Things that I use” Dependencies App & Product Cryptographic Product owner Posture Source code Software depen
Slide 67: IBM Quantum Safe Strategy and Technology
IBM Quantum Safe Strategy and Technology External systems Transform Remediator Studio | Automation | Crypto Agility Business applications Continuous integration/ continuous delivery Observe Advisor Compliance policies | Vulnerabilities | Prioritization Ticketing system Vulnerability management syste
Slide 68: IBM Quantum Safe Strategy and Technology
IBM Quantum Safe Strategy and Technology Transform Remediator Studio | Automation | Crypto Agility Observe Advisor Compliance policies | Vulnerabilities | Prioritization Discover Explorer Scanners | Cryptography Bill of Materials (CBOM) | Dependency analysis Help clients to – understand the impact o
Slide 69: The time to start is now
The time to start is now Understand the Identify cryptography Begin transformation quantum risks and prioritize actions with crypto agility IBM Research / © 2023 IBM Corporation 71
Slide 70: Thank you
Thank you IBM Research / © 2023 IBM Corporation 72