Photobucket, once a leading image-hosting site, now seeks a resurgence through AI.

CEO Ted Leonard discusses licensing 13 billion photos/videos for AI training.

Photobucket's content could be worth billions, reflecting a booming AI data market.

Tech giants like Google, Meta, and Microsoft quietly buy data for AI training.

The hidden trade includes chat logs and old social media photos.

An estimated $2.5 billion AI data market could grow to $30 billion in a decade.

Deals with content owners aim to secure data for AI model training.

Shutterstock, Freepik among those licensing vast image libraries to tech companies.

Concerns rise over personal data's use in AI without explicit consent.

The FTC investigates data licensing practices amid privacy, copyright concerns.

Apple was quiet about its GenAI activities.

Last week, ReALM (Reference Resolution As Language Modeling) was announced.

ReALM uses screen content for context, improving Siri's functionality for cases like "Hey Siri, call this number".

Four model sizes available, from 80 million to 3 billion parameters. Suggests ReALM might process data locally.

Privacy concerns regarding this feature are yet to be addressed.

Apple's stance on privacy with ReALM remains to be seen.

Overall, Apple introduces context-aware AI enhancements for Siri.

Salesforce's Winter '24 release enhances various clouds with new features.

Automates post-call and meeting notes with editable summaries for sales calls.

Personalizes sales emails using AI, leveraging sales data for impactful content.

Provides real-time response recommendations for customer support agents.

Predicts and fills case summaries for service agents, improving documentation.

Offers AI-generated work briefings for Field Service mobile workers.

Enhances Commerce product fields with AI for engaging descriptions.

Introduces a Summary tab in call records for easy information capture.

Improves customer service efficiency with AI-powered assistance.

Elevates Field Service operations with comprehensive AI-generated briefings.

1. Content Generation: AI crafts new articles, enhancing knowledge bases.

2. Summarization: AI condenses reviews, improving knowledge accessibility.

3. Categorization: AI sorts content, enhancing knowledge base navigation.

4. Q&A Generation: AI creates relevant questions and answers.

5. Chatbots Creation: AI develops assistants, streamlining routine inquiries.

6. Customer Insights: AI analyzes feedback, enhancing customer understanding.

7. Sales Support: AI personalizes recommendations, boosting customer satisfaction.

8. Personalized Content: AI tailors content, improving user engagement.

9. Product Recommendations: AI suggests products, enhancing shopping experiences.

10. Onboarding Experience: AI customizes welcome materials for newcomers.

11. Customer Engagement: AI delivers tailored content, fostering loyalty.

12. Customer Support: AI automates routine service queries.

13. Customer Retention: AI provides targeted promotions, encouraging loyalty.

14. Content Freshness: AI generates topical content regularly.

15. Content Optimization: AI refines content, improving relevance.

16. Translation Services: AI breaks language barriers, expanding reach.

17. Metadata Creation: AI enhances content discoverability through SEO.

18. Content Organization: AI categorizes and tags for better findability.

19. Behavior Analysis: AI uncovers trends, tailoring customer engagement.

20. Persona Development: AI creates detailed customer profiles.

21. Customized Marketing: AI crafts targeted promotional campaigns.

22. Predictive Modeling: AI predicts customer behaviors, guiding marketing.

Transitioned from Data Lab to Analytics Factory, scaling data use across Carrefour.

Analytics Factory aims to combine intimate business knowledge with data science.

Maintains balance between strategic priorities and adaptability to ensure relevance.

Carrefour's digital transformation focuses on integrating data analytics into core processes.

Data team, Analytics Factory, consists of 150 people, expanding across Europe.

Challenges include maintaining focus on strategic projects and balancing centralization with decentralization.

Introduced role of "data translator" to bridge data science and business knowledge.

Example projects: optimizing promotions using AI and personalizing e-commerce recommendations.

Future focuses include expanding Analytics Factory's approach to all European countries.

Emphasizes the importance of data culture and leveraging generative AI for new opportunities.

Generative AI impacts business growth, margins, and daily operations.

Investors, customers, and employees have evolving expectations around GenAI.

Executives must integrate GenAI into business models and plan for future disruptions.

By 2025, 70% of enterprises will prioritize AI's sustainable and ethical use.

GenAI-powered chatbots' mistrust will increase service costs by 40% by 2025.

Synthetic data will reduce machine learning's real data needs by 70% by 2025.

30% of large organizations' marketing messages will be synthetically generated by 2025.

AI's impact on global jobs will be neutral through 2026.

By 2030, AI could reduce CO2 emissions by 5-15% and increase global electricity consumption.

AI without human oversight could cause $100 billion in asset damage by 2030.

Generative AI requires a combination of models to function effectively.

Companies find it easy to create proofs of concept, but difficult to scale them up.

Challenges include aligning business needs with the technical reality of generative AIs.

Successful projects often combine open source and proprietary solutions.

Hybrid approaches and strategic use of specific models are crucial for success.

The AI team needs to blend technical skills with business objectives under a strong, dual-knowledge leader.

The push for generative AI in enterprises includes both internal productivity apps and customer-facing solutions, prioritizing data security and customization.

Enterprises focus on customizing AI models through fine-tuning rather than building from scratch, leveraging cloud services for model access.

There's a growing trend towards using open-source models in both countries, but more pronounced in the US.

Joanna Stern explores replacing herself with AI for a day.

Uses smiling and breathing for an AI avatar's creation.

Stern's AI avatar developed by startup Synthesia and ElevenLabs.

AI avatar made with Stern's recorded head movements and voice.

Stern tests AI in four challenges: phone calls, TikTok, bank biometrics, video calls.

AI voice fools people in phone calls but not in video calls.

Challenge with TikTok: avatar's lack of movement and facial expressions.

Bank biometrics passed using AI voice for verification.

Video call test reveals AI's inability to mimic human interactions fully.

Concludes AI voices are effective but raises concerns about misuse.

RAG improves LLM accuracy

RAG combines current, sourced information with large language models' generative capabilities.

LLMs can give outdated or unsourced answers, unlike RAG-enhanced responses.

RAG accesses updated data stores for current answers, reducing misinformation risks.

It instructs LLMs to verify information before responding, enhancing reliability.

RAG allows LLMs to admit ignorance when data is insufficient, avoiding speculation.

RAG's goal is to make LLMs more accurate and trustworthy by grounding responses in verified data.

Watsonx.ai: A studio for building and deploying AI models tailored to business needs.

Features Prompt Lab for creating and testing generative AI prompts easily.

Offers a variety of pre-trained models, including Llama 2 and Granite series.

Supports prompt engineering with natural language instructions and AI guardrails.

Enables sharing and further testing of prompts via API or Jupyter Notebooks.

Includes a Tuning Studio for customizing models with proprietary data.

Allows building traditional machine learning models in a collaborative workspace.

AutoAI feature for no-code ML model creation, experimentation, and deployment.

Synthetic data generator tool for creating datasets up to 2.1 billion rows.

Alan uses AI to enhance employee efficiency. 

AI helps reduce variable costs at Alan.

AI aids in developing new product types.

Custom AI assistants tailored to specific data and team needs.

Data analysis tool "@metabase" used for growth metrics.

Automation improves customer service and reduces service costs.

AI document parsing for efficient claim processing.

Tools for sales, including account summaries and meeting recaps.

AI enhances customer care with immediate responses and document handling.