Data Science Conversations

Early Interest in Generative AI

• Martin's initial exposure to Generative AI in 2016 through a conference talk in Milano, Italy, and his early work with Generative Adversarial Networks (GANs).

Development of GANs and Early Language Models since 2016

• The evolution of Generative AI from visual content generation to text generation with models like Google's Bard and the increasing popularity of GANs in 2018.

Launch of GenerativeAI.net and Online Course

• Martin's creation of GenerativeAI.net and an online course, which gained traction after being promoted on platforms like Reddit and Hacker News.

Defining Generative AI

• Martin’s explanation of Generative AI as a technology focused on generating content, contrasting it with Discriminative AI, which focuses on classification and selection.

Evolution of GenAI Technologies

• The shift from LSTM models to Transformer models, highlighting key developments like the "Attention Is All You Need" paper and the impact of Transformer architecture on language models.

Impact of Computing Power on GenAI

• The role of increasing computing power and larger datasets in improving the capabilities of Generative AI

Generative AI in Business Applications

• Martin’s insights into the real-world applications of GenAI, including customer service automation, marketing, and software development.

Retrieval Augmented Generation (RAG) Architecture

• The use of RAG architecture in enterprise AI applications, where documents are chunked and queried to provide accurate and relevant responses using large language models.

Technological Drivers of GenAI

• The advancements in chip design, including Nvidia’s focus on GPU improvements and the emergence of new processing unit architectures like the LPU.

Small vs. Large Language Models

• A comparison between small and large language models, discussing their relative efficiency, cost, and performance, especially in specific use cases.

Challenges in Implementing GenAI Systems

• Common challenges faced in deploying GenAI systems, including the costs associated with training and fine-tuning large language models and the importance of clean data.

Measuring GenAI Performance

• Martin’s explanation of the complexities in measuring the performance of GenAI systems, including the use of the Hallucination Leaderboard for evaluating language models.

Emerging Trends in GenAI

• Discussion of future trends such as the rise of multi-agent frameworks, the potential for AI-driven humanoid robots, and the path towards Artificial General Intelligence (AGI).

In this episode, we sit down with Steve Orrin, Federal Chief Technology Officer at Intel Corporation. Steve shares his extensive experience and insights on the transformative power of AI and its parallels with past technological revolutions. He discusses Intel’s pioneering role in enabling these shifts through innovations in microprocessors, wireless connectivity, and more.

Steve highlights the pervasive role of AI in various industries and everyday technology, emphasizing the importance of a heterogeneous computing architecture to support diverse AI environments. He talks about the challenges of operationalizing AI, ensuring real-world reliability, and the critical need for robust AI security. Confidential computing emerges as a key solution for protecting AI workloads across different platforms.

The episode also explores Intel’s strategic tools like oneAPI and OpenVINO, which streamline AI development and deployment. This episode is a must-listen for anyone interested in the evolving landscape of AI and its real-world applications.

Intel's Legacy and Technological Revolutions

• Historical parallels between past tech revolutions (PC era, internet era) and current AI era.
• Intel's contributions to major technological shifts, including the development of wireless technology, USB, and cloud computing.

AI's Current and Future Landscape

• AI's pervasive role in everyday technology and various industries.
• Importance of computing hardware in facilitating AI advancements.
• AI's integration across different environments: cloud, network, edge, and personal devices.

Intel's Approach to AI

• Focus on heterogeneous computing architectures for diverse AI needs.
• Development of software tools like oneAPI and OpenVINO to enable cross-platform AI development.

Challenges and Solutions in AI Deployment

• Scaling AI from lab experiments to real-world applications.
• Ensuring AI security and trustworthiness through transparency and lifecycle management.
• Addressing biases in AI datasets and continuous monitoring for maintaining AI integrity.

AI Security Concerns

• Protection of AI models and data through hardware security measures like confidential computing.
• Importance of data privacy and regulatory compliance in AI deployments.
• Emerging threats such as AI model poisoning, prompt injection attacks, and adversarial attacks.

Innovations in AI Hardware and Software

• Confidential computing as a critical technology for securing AI.
• Research into using AI for chip layout optimization and process improvements in various industries.
• Future trends in AI applications, including generative AI for fault detection and process optimization.

Collaboration and Standards in AI Security

• Intel's involvement in developing industry standards and collaborating with competitors and other stakeholders.
• The role of industry forums and standards bodies like NIST in advancing AI security.

Advice for Aspiring AI Security Professionals

• Importance of hands-on experience with AI technologies.
• Networking and collaboration with peers and industry experts.
• Staying informed through industry news, conferences, and educational resources.

Exciting Developments in AI

• Fusion of multiple AI applications for complex problem-solving.
• Advancements in AI hardware, such as AI PCs and edge devices.
• 
  
• Potential transformative impacts of AI on everyday life and business operations.

In this episode we talk to Kirk Marple about the power of Knowledge Graphs when combined with GenAI models.  Kirk explained the growing relevance of knowledge graphs in the AI era, the practical applications, their integration with LLMs, and the future potential of Graph RAG.

Kirk Marple a veteran of Microsoft and General Motors, Kirk has spent the last 30 years in software development and data leadership roles. He also successfully exited the first startup he founded, RadiantGrid, acquired by Wohler Technologies.

Now, as the technical founder and CEO of Graphlit, Kirk and his team are streamlining the development of vertical AI apps with their end-to-end, cloud based offering that ingests unstructured data and leverages retrieval augmented generation to improve accuracy, domain specificity, adaptability, and context understanding – all while expediting development.

Episode Summary -

• Introduction to Knowledge Graphs:
• Knowledge graphs extract relationships between entities like people, places, and things, facilitating efficient information retrieval.
• They represent intricate interactions and interrelationships, enabling users to "walk the graph" and uncover deeper insights.
• 
  
• Importance in the AI Era:
• Knowledge graphs enhance data retrieval and filtering, crucial for feeding accurate data into large language models (LLMs) and multimodal models.
• They provide an additional axis for information retrieval, complementing vector search.
• Industry Use Cases:
• Commonly used in customer data platforms and CRM models to map relationships within and between companies.
• Knowledge graphs can convert complex datasets into structured, easily queryable formats.
• Challenges and Limitations:
• Familiarity with graph databases and the ETL process for graph data integration is still developing.
• Graph structures are less common and more complex than traditional relational models.
• Integrating Knowledge Graphs with LLMs:
• Knowledge graphs enrich data integration and semantic understanding, adding context to text retrieved by LLMs.
• They can help reduce hallucinations in LLMs by grounding responses with more accurate and comprehensive context.
• Graph RAG (Retrieval Augmented Generation):
• Combines knowledge graphs with RAG to provide additional context for LLM-generated responses.
• Allows retrieval of data not directly cited in the text, enhancing the breadth of information available for queries.
• Scalability and Efficiency:
• Effective graph database architectures can handle large-scale graph data efficiently.
• Graph RAG requires a robust ingestion pipeline and careful management of data freshness and retrieval processes.
• Future Developments:
• Growing interest and implementation of knowledge graphs and Graph RAG in various industries.
• Potential for new tools and standardization efforts to make these technologies more accessible and effective.
• Graphlit: Simplifying Knowledge Graphs:
• The platform focuses on simplifying the creation and use of knowledge graphs for developers.
• Provides APIs for easy integration, supporting domain-specific vertical AI applications.
• Offers a unified pipeline for data ingestion, extraction, and knowledge graph construction.
• Open Source and Community Contributions:
• Recommendations for libraries and projects in the knowledge graph space.
• Notable contributors and projects include data extraction libraries and AI agent initiatives.

At LanguageTool, Bartmoss St Clair (Head of AI) is pioneering the use of Large Language Models (LLMs) for grammatical error correction (GEC), moving away from the tool's initial non-AI approach to create a system capable of catching and correcting errors across multiple languages.

LanguageTool supports over 30 languages, has several million users, and over 4 million installations of its browser add-on, benefiting from a diverse team of employees from around the world.

Episode Summary -

1. LanguageTool decided against using existing LLMs like GPT-3 or GPT-4 due to cost, speed, and accuracy benefits of developing their own models, focusing on creating a balance between performance, speed, and cost.
2. The tool is designed to work with low latency for real-time applications, catering to a wide range of users including academics and businesses, with the aim to balance accurate grammar correction without being intrusive.
3. Bartmoss discussed the nuanced approach to grammar correction, acknowledging that language evolves and user preferences may vary, necessitating a balance between strict grammatical rules and user acceptability.
4. The company employs a mix of decoder and encoder-decoder models depending on the task, with a focus on contextual understanding and the challenges of maintaining the original meaning of text while correcting grammar.
5. A hybrid system that combines rule-based algorithms with machine learning is used to provide nuanced grammar corrections and explanations for the corrections, enhancing user understanding and trust.
6. LanguageTool is developing a generalized GEC system, incorporating legacy rules and machine learning for comprehensive error correction across various types of text.
7. Training models involve a mix of user data, expert-annotated data, and synthetic data, aiming to reflect real user error patterns for effective correction.
8. The company has built tools to benchmark GEC tasks, focusing on precision, recall, and user feedback to guide quality improvements.
9. Introduction of LLMs has expanded LanguageTool's capabilities, including rewriting and rephrasing, and improved error detection beyond simple grammatical rules.
10. Despite the higher costs associated with LLMs and hosting infrastructure, the investment is seen as worthwhile for improving user experience and conversion rates for premium products.
11. Bartmoss speculates on the future impact of LLMs on language evolution, noting their current influence and the importance of adapting to changes in language use over time.
12. LanguageTool prioritizes privacy and data security, avoiding external APIs for grammatical error correction and developing their systems in-house with open-source models.

In this enlightening episode, Dr. Julia Stoyanovich delves into the world of responsible AI, exploring the ethical, societal, and technological implications of AI systems. She underscores the importance of global regulations, human-centric decision-making, and the proactive management of biases and risks associated with AI deployment. Through her expert lens, Dr. Stoyanovich advocates for a future where AI is not only innovative but also equitable, transparent, and aligned with human values.

Julia is an Institute Associate Professor at NYU in both the Tandon School of Engineering, and the Center for Data Science.  In addition she is Director of the Center for Responsible AI also at NYU.  Her research focuses on responsible data management, fairness, diversity, transparency, and data protection in all stages of the data science lifecycle. 

Episode Summary -

1. The Definition of Responsible AI
2. Example of ethical AI in the medical world - Fast MRI technology
3. Fairness and Diversity in AI
4. The role of regulation - What it can and can’t do
5. Transparency, Bias in AI models and Data Protection
6. The dangers of Gen AI Hype and problematic AI narratives from the tech industry
7. The impotence of humans in ensuring ethical development 
8. Why “Responsible AI” is actually a bit of a misleading term
9. What Data & AI leaders can do to practise Responsible AI

Luis Moreira-Matias is Senior Director of Artificial Intelligence at sennder, Europe’s leading digital freight forwarder. At sennder, Luis founded sennAI: sennder’s organization that oversees the creation (from R&D to real-world productization) of proprietary AI technology for the road logistics industry.

During his 15 years of career, Luis led 50+ FTEs across 4+ organisations to develop award-winning ML solutions to address real-world problems in various fields such as e-commerce, travel, logistics, and finance. 

Luis holds a Ph.D. in Machine Learning from the U. Porto, Portugal. He possesses a world-class academic track with high impact publications at top tier venues in ML/AI fundamentals, 5 patents and multiple keynotes worldwide - ranging from Brisbane (Australia) to Las Palmas (Spain).

In this episode Tarush Aggarwal, formerly of Salesforce and WeWork is back on the podcast to discuss the evolution of the Semantic layer and how that can help practitioners get results from LLMs.  We also discuss how smaller ELMs (expert language models) might be the future when it comes to consistent reliable outputs from Generative AI and also the impact of all of this on traditional BI tools.

In this episode Patrick McQuillan shares his innovative Biological Model - a concept you can use to enhance data outcome in large enterprises.  The concept takes the idea that the best way to design a data strategy is to align it closely with a biological system.

He discusses the power of centralized information, importance of data governance, and the necessity for a common performance narrative across an organization.

Episode Summary -

- Biological Model Concept

- Centralized vs. Decentralized Data

- Data Collection and Maturity

- Horizontal translation layer 

- Partnership with vertical leaders

 - Curated data layers 

- Data dictionary for consistency

- Focusing on vital metrics

- Data Flow in Organizations

- Biological Model Governance

- Overcoming Inconsistency and Inaccuracy

In this episode Heidi Hurst returns to talk to us about how in her current role at Pachama she is using the power of machine learning to fight climate change.  She discusses her work in measuring the capacity of existing forests and reforestation projects using satellite imagery.

Episode Summary

1. The importance of carbon credits verification in mitigating climate change

2. How Pachama is using machine learning and satellite imagery to verify carbon projects

3. Three types of carbon projects: avoided deforestation, reforestation, and improved forest management

4. Challenges in using satellite imagery to measure the capacity of existing forests

5. The role of multispectral imaging in measuring density of forests

6. Challenges in collecting data from dense rainforests and weather obstructions

7. The impact of machine learning on scaling up carbon verification

8. Advancements in the field of satellite imaging, particularly in small satellite constellations

Ágnes Horvát is an Assistant Professor in Communication and Computer Science at Northwestern University. Her work focuses on understanding how online networks induce biased information production, sharing and processing across digital platforms. 

- The new Post-normal era for science - Having an awareness of the context and values that impact scientific research

• Where is science communication in relation to digital platforms? - Scholars work hard on discovering scientific findings, but information doesn’t always reach the public appropriately.  

• How to communicate scientific research - it’s not just about communicating with scientists and general audiences. News needs to reach policymakers and governments too for real change.

• The production of scientific research has exploded recently thanks to decision-making demands - and the pandemic had a lot to do with this. Scientists were under pressure to carry out research quickly and at the expense of quality. 

• Misinformation can have detrimental consequences - even leading to vaccine hesitancy in some communities.

• The surprising effect of retracting papers - papers that get retracted in the future are more likely to receive more engagement before getting withdrawn.

• Why are paper retractions on the rise?  - again, the recent pandemic has caused an increase in retractions. 

• Is social media helping or hindering science research? - while the platforms are helping to  spread real news, social media also helps the spread of false information. 

• As long as you have quality data and robust trends - regardless of the method, you will identity that trend. 

Reducing the problem of miscommunication - with whom does the responsibility lie?