IBM THINK 2023: AI and Business

Introduction

• Opening Remarks: IBM THINK 2023
  • AI in business needs to be reliable, secure, and adaptable.
  • Hybrid-ready AI that scales across systems.
  • Emphasis on AI’s transparency and its potential for extensive business applications.

Keynote Speaker: Dr. Dario Gil

• Position: Senior Vice President and Director of Research, IBM.
• Topic: The significant impact of AI and the need for businesses to become AI value creators.

AI’s Impact on Industries

• AI is transforming various industries: customer care, data centers, logistics, medicine, manufacturing, energy, automotive, aerospace, and communications.
• Foundation models and generative AI are pivotal technologies.

Being an AI Value Creator

• AI User vs. AI Value Creator
  • AI User: Limited to prompting pre-made models, no control over the models or data.
  • AI Value Creator: Control over data, model training, and tuning. Ability to own customized foundation models.
• Watsonx: Integrated data and AI platform comprising: Watsonx.data, Watsonx.ai, and Watsonx.governance.

Watsonx Components

• Watsonx.data: Massive curated data repository for training and fine-tuning models.
• Watsonx.ai: Enterprise studio for training, validating, tuning, and deploying models.
• Watsonx.governance: Tools for ensuring responsible AI execution.
• Red Hat OpenShift: Underpinning hybrid cloud architecture for seamless integration.

AI Workflow with Watsonx

1. Data Preparation
   • Connects and accesses diverse data sources.
   • Filters, categorizes, annotates, and tags data for model training.
   • Uses IBM data pile combining public and proprietary data.
2. Model Training
   • Selection from IBM’s model architectures (e.g., encoder-decoder, etc.).
   • Utilizes IBM Vela, a cloud-native AI supercomputer.
   • Tokenization of data and model training using scalable resources.
3. Validation
   • Running benchmarks and creating detailed model cards.
4. Governance
   • Combining data cards (data provenance) and model cards into fact sheets.
   • Monitoring models in production and ensuring compliance with changes.
5. Tuning and Deployment
   • Adapting models to specific tasks using business data.
   • Deployment of models in various IT environments.
   • Continuous monitoring with Watsonx.governance and updating models as needed.

Case Studies and Applications

• SAP: Uses Watson capabilities for digital assistant in enterprise solutions.
• Red Hat: Embeds Watson Code Assistant in Ansible Automation Platform.
• BBVA: Applies proprietary data to own foundation model for NLP.
• Moderna: Utilizes IBM models for predicting mRNA medicines.
• NASA: Enhances scientific understanding with language and spatial models.
• WiX: Gains insights for customer care through Watsonx.

Partnership with Hugging Face

• Guest Speaker: Clem Delangue, CEO of Hugging Face.
  • Discussing the role of open community and the collaboration with IBM.
  • Emphasizing the need for customized AI models over singular dominant models.
• Hugging Face Contributions
  • Repository of over 250,000 open models, 50,000 datasets, and 100,000 demos.
  • Integration into Watsonx to provide extensive AI resources and community support.

Strategic Recommendations for Businesses

1. Act with Urgency: Embrace transformative technology now.
2. Be a Value Creator: Build and own your AI models and data.
3. Bet on Community: Leverage the innovations of the open AI community.
4. Run Everywhere Efficiently: Optimize performance across hybrid environments.
5. Be Responsible: Integrate transparency and governance throughout the AI lifecycle.

Conclusion

• Watsonx: A comprehensive platform for businesses to create, govern, and scale AI solutions.
• Encouragement for businesses to adopt AI and join IBM on this technological journey.

Key Message: Capture the moment, be an AI value creator, and build responsibly with IBM’s Watsonx platform. (Applause)