Go Summarize

No Priors Ep. 50 | With Stripe Head of Information Emily Glassberg Sands

1K views|3 months ago
💫 Short Summary

The video discusses how Stripe leverages data science and machine learning to improve businesses' success. They use AI tools like LLMS and generative AI for fraud detection, with a focus on internal product adoption and promotion. Stripe introduces new products like Radar Assistant and Sigma Assistant to automate code writing and accelerate information retrieval, benefiting various domains. They emphasize the potential of generative AI in enhancing user experience and business operations, aiming to empower both technical and non-technical users. The video also touches on the importance of integrating user-specific solutions without extensive developer work and leveraging payment data for deep insights.

✨ Highlights
📊 Transcript
Emily's role at Stripe involves leading data science and machine learning initiatives to improve businesses' long-term success.
She supports teams in utilizing data effectively and oversees self-served businesses, focusing on user integration and product experiences.
Stripe has early adopted machine learning models like LLMS and generative AI for fraud detection.
The company's focus extends beyond payments to include invoicing, subscriptions, billing, tax, and revenue management based on users' business needs.
Stripe's evolution from financial infrastructure to incorporating machine learning in various contexts.
Stripe's culture encourages bottom-up experimentation, exemplified by the implementation of llm Explorer for internal use.
The three main areas of llm adoption at Stripe: external products, internal tools, and vendor integration.
Teams at Stripe prioritize external applications before focusing on internal and vendor use cases.
Importance of diverse backgrounds in internal product adoption and promotion within companies.
Involving individuals with diverse backgrounds and expertise is crucial for effective product application.
Example of llm Explorer shows how a front end tool supporting multiple models can quickly gain internal adoption.
Emphasis on creating community and sharing within tools to enhance collaboration and productivity.
Implementation of features like presets for prompt sharing is a successful strategy for increasing user engagement and interaction patterns.
LM Explorer has almost 3,000 weekly active users among engineers, salespeople, and marketers.
Accelerator teams, funded for 6 months, are behind LM Explorer to develop new AI capabilities.
The teams operate with a one to two Pizza team structure and focus on experimentation and iteration.
Accelerators aim to create homegrown solutions for internal applications, providing growth opportunities for internal talent.
Overall, accelerators play a crucial role in driving innovation and creating impactful solutions within the company.
Stripe introduces Radar Assistant and Sigma Assistant to automate code writing and accelerate information retrieval.
Radar Assistant generates custom fraud rules from natural language, enhancing fraud detection capabilities.
Sigma Assistant applies natural language to create explicit policy descriptions, improving decision-making processes.
These tools empower a wide range of users to implement rules quickly without developer assistance.
The applications have potential beyond fraud detection, benefiting various domains such as underwriting.
Integration of code for payment options and hosted invoices is highlighted for convenience in companies without developer skills.
Sigma Assistant, an AI tool, is introduced to generate business insights from natural language queries on stripe data.
The tool aims to simplify access to revenue data and customer insights without requiring SQL knowledge.
The discussion extends to the potential impact of generative AI on businesses and the economy, emphasizing the need for innovative models to optimize payments and address fraud effectively.
The potential of generative AI in utilizing payments data for enhancing user experience and business operations.
Real-time data can speed up insights and responses, aiming to help users grow their businesses beyond just payments.
Challenges include deciding on the right operating model for implementing AI technologies within product engineering teams.
Stripe focuses on leveraging AI to empower technical and non-technical users, automating tasks like code writing and accelerating processes.
Approach to building information retrieval systems.
Utilize existing vertical teams and accelerators for developing internal models for different use cases.
Individual teams are allowed to choose models based on cost and latency considerations.
Emphasis on quality, scaling, and performance requirements when selecting models.
Developing an internal API for more programmatic use of models, with multiple applications built on it.
Importance of building internal and external ML infrastructure and experimentation solutions.
Emphasis on identifying unique needs and building internally when necessary.
Significance of understanding business identity, particularly in fintech, for credit lending decisions and meeting card network requirements.
Opportunity seen to streamline financial integrations, potentially through products like Stripe's suite of no-code solutions.
Importance of integrating user-specific solutions without extensive developer work.
Success cases exist, but there is room for improvement in building automated integrations.
Leveraging payment data for deep insights to improve businesses and drive economic growth.
Stripe aligns incentives with user success and uses financial data to help businesses thrive.
Machine learning optimizes authorization requests, reducing false declines and leading to significant cost savings globally.
Using data analytics to optimize transactions and prevent fraud.
Stripe Radar can reduce declines in transactions by 30% by identifying the best day and time to retry payments.
The speaker's background in labor economics influences their approach to data science and decision-making.
Data analysis can improve user experience, increase revenue, and contribute to economic growth.
Early experiences with data analysis led the speaker to pursue a PhD in economics and join Corsair at Stripe.
Stripe aims to become an economic operating system for users by providing economic insights to aid decision-making and growth.
They prioritize long-term investments based on serving users rather than micromanaging.
The approach allows them to take a long-sighted view in choosing where and how to invest in the business.
Impact of AI on Education
AI can provide personalized learning experiences tailored to individual needs.
AI helps in closing skills gaps and matching individuals with suitable jobs.
Utilizing education data is crucial for benefiting both learners and companies.
Developing skills through AI and signaling learning outcomes in the labor market through credentials is valuable.
AI has the potential to enhance elementary school education and improve global access to quality instruction.
Challenges faced by AI startups include high initial compute costs and global demand for products from the start.
Many AI startups are adopting subscription-based business models to monetize early.
Lean teams and financial pressures are driving the need for revenue and financial automation tools.
The adoption of these tools is helping AI startups manage growth effectively.
Rapid monetization in the AI wave.
Many generative AI companies are scaling up quickly, with startups monetizing early in foundation and application layers.
Top companies like OpenAI and Mistrall are leading the way, alongside others like Moon Beam and Runway.
Strong product-market fit and high demand from users for AI products and services are driving the trend.