QuantumBlack and McKinsey discuss the trade-offs of buying vs building GenAI solutions, including considerations of black box solutions and transparency. They explore the roles of traditional AI and JAN-AI in messaging channels and the generative nature of AI. The challenges and considerations in using APIs for machine learning models are also discussed.
Read more
AI Summary
Highlights
AI Chapters
Episode notes
auto_awesome
Podcast summary created with Snipd AI
Quick takeaways
Organizations can choose to build or buy generative AI solutions, depending on their needs and resources.
The type and format of data used for generative AI models significantly impact their performance and creativity.
Generative AI and traditional AI approaches complement each other, and organizations should leverage the strengths of both for comprehensive solutions.
Deep dives
Generative AI: Building vs Buying
Generative AI, specifically focused on chatbot applications, can be effectively built or bought depending on the organization's needs and resources. Organizations can start with off-the-shelf solutions like OpenAI to quickly create proof-of-concepts (POCs) and gauge the return on investment (ROI). However, when moving towards productionization, organizations may require more transparency and control, prompting a shift towards open source solutions. When considering the transition from build to buy, it is important to mitigate migration risks and ensure model interoperability. This involves adopting solid technology principles, abstracting prompts, and conducting thorough testing to ensure compatibility between different models and versions.
Data Readiness and Nutrition for Generative AI
Generative AI technology, such as language models, requires proper data readiness and nutrition. Organizations must carefully consider the type and format of data that the models feed on to ensure desired outputs and levels of creativity. It is crucial to understand which data sources (e.g., Word documents, PDFs, structured or unstructured data) best serve the generative AI applications. Additionally, organizations should factor in the sensitivity of generative AI to data type and format when planning data governance, testing, and risk mitigation strategies.
Complementarity between Generative AI and Traditional AI
Generative AI, with its focus on spontaneity and creativity, complements traditional AI approaches. Generative AI excels in tasks such as summarizing information, generating insights, and customizing responses based on specific inputs. However, traditional AI techniques, such as analytics and machine learning, are essential for handling numerical data, complex reasoning, and traditional ML applications. Organizations should consider both generative AI and traditional AI approaches for comprehensive solutions, leveraging the strengths of each approach.
Considerations for Model Interoperability
When transitioning from off-the-shelf solutions to open source models, organizations should carefully consider model interoperability and migration risks. Adopting architectural principles that abstract prompts, store metadata, and allow testing is beneficial for addressing model compatibility. Organizations must document and validate prompt-response pairs to ensure consistent behavior across different models and versions. Additionally, proper testing and monitoring mechanisms should be established to identify any discrepancies and ensure ongoing model performance.
Transparency, Data Security, and Risk Considerations
Organizations need to navigate the balance between transparency, data security, and risk in generative AI applications. Closed enterprise versions of generative AI systems provide data confidentiality, as data does not leave the organization. However, these versions may lack transparency regarding model training and may require additional enforcement layers to avoid undesirable outputs. Open source models offer more transparency but require organizations to oversee data security and address migration risks. Regulatory compatibility, responsible AI principles, and vendor alignment with organizational values are vital considerations for risk management in generative AI applications.
Transform faster. Innovate smarter. Anticipate the future. At QuantumBlack, we unlock the power of artificial intelligence (AI) to help organizations reinvent themselves from the ground up—and accelerate sustainable and inclusive growth.
MLOps Coffee Sessions Special episode with QuantumBlack, AI by McKinsey, GenAI Buy vs Build, Commercial vs Open Source, fueled by our Premium Brand Partner, QuantumBlack, AI by McKinsey.
// Abstract
Do you build or buy?
Check the QuantumBlack team discussing the different sides of buying vs building your own GenAI solution.
Let's look at the trade-offs companies need to make - including some of the considerations of using black box solutions that do not provide transparency on what data sources were used.
Whether you are a business leader or a developer exploring the space of GenAI, this talk provides you with valuable insights to prepare you for how you can be more informed and prepared for navigating this fast-moving space.
// Bio
Ilona Logvinova
Ilona Logvinova is the Head of Innovation for McKinsey Legal, working across the legal department to identify, lead, and implement cross-cutting and impactful innovation initiatives, covering legal technologies and reimagination of the profession initiatives. At McKinsey Ilona is also Managing Counsel for McKinsey Digital, working closely with emerging technologies across use cases and industries.
Mohamed Abusaid
Am Mohamed, a tech enthusiast, hacker, avid traveler, and foodie all rolled into one individual. Built his first website when he was 9 and fell in love with computers and the internet ever since. Graduated with a computer science from university although dabbled in electrical, electronic, and network engineering before that. When he's not reading up on the latest tech conversations and products on Hacker News, Mohamed spends his time traveling to new destinations and exploring their cuisine and culture.
Mohamed works with different companies helping them tackle challenges in developing, deploying, and scaling their analytics to reach its potential. Some topics he's enthusiastic about include MLOps, DataOps, GenerativeAI, Product thinking, and building cross-functional teams to deliver user-first products.
Nayur Khan
Nayur is a partner within McKinsey and part of the QuantumBlack, AI by McKinsey leadership team.
He predominantly focuses on helping organizations build capabilities to industrialize and scale artificial intelligence (AI), including the newer Generative AI. He helps companies navigate innovations, technologies, processes, and digital skills as needed to run at scale. He is a keynote speaker and is recognized in the DataIQ 100 - a list of the top 100 influential people in data.
Nayur also leads the firm’s diversity and inclusion efforts within QuantumBlack to promote a more equitable environment for all. He speaks with organizations on the importance of diversity and diverse team building—especially when working with data and building AI.
// MLOps Jobs board
https://mlops.pallet.xyz/jobs
// MLOps Swag/Merch
https://mlops-community.myshopify.com/
// Related Links
--------------- ✌️Connect With Us ✌️ -------------
Join our slack community: https://go.mlops.community/slack
Follow us on Twitter: @mlopscommunity
Sign up for the next meetup: https://go.mlops.community/register
Catch all episodes, blogs, newsletters, and more: https://mlops.community/
Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Ilona on LinkedIn: www.linkedin.com/in/ilonalogvinova
Connect with Mo on LinkedIn: https://www.linkedin.com/in/mabusaid/
Connect with Nayur on LinkedIn: https://www.linkedin.com/in/nayur/
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode