How Pepsi is using AI to improve internal workflows

💡 1 USE CASE

PepsiCo's generative AI integration and how other enterprises can learn from them

Consumer goods giant PepsiCo has gone all-in on artificial intelligence investing hundreds of millions of dollars to integrate AI across its operations according to PepsiCo executive Dr. Athina Kanioura. PepsiCo's wide deployment of AI provides valuable lessons for companies looking to leverage generative AI's capabilities.

PepsiCo is utilizing AI for a diverse array of use cases:

In product development - AI helps create new flavors and variations.

For logistics - AI optimizes product availability by predicting demand at stores.

In marketing - AI enhances campaigns through data analysis. It can also bolster sustainability efforts and warehouse safety.

This wide application of AI has delivered tangible benefits for PepsiCo, including improved efficiency and increased customer responsiveness. PepsiCo has also implemented a responsible AI framework to help guide deployment, ensure model prediction equity, and avoid pitfalls like bias.

Dr. Kanioura has noted that AI development and integration is going to advance far beyond generative / chatbot style AI. "I don't believe that generative is the panacea for everything," she said. "I think you can also do things with much simpler AI practices or process automation."

Other companies can follow PepsiCo's lead in strategically applying AI across operations outside of chat.

Here are some potential enterprise use cases:

Customer churn prevention - Apply machine learning to user data to identify customers likely to churn, & proactively offer promotions.

Predictive analytics - Apply machine learning to forecast sales, detect trends, and gain deeper market insights.

Automated data validation - Use AI techniques to clean, structure, & validate data for analysis.

Sentiment analysis - Leverage natural language processing to extract emotional tone from customer feedback and reviews.

Through optimizing strategic applications of AI, companies can significantly automate core workflows and provide data-driven insights, enhance decision-making, and elevate human capabilities across an organization.

The key is determining where AI can augment human strengths focused on the real work that delivers value to customers.

🛠️ 2 TOOLS

GPT-4 Vision (🔗 link)

GPT-4 Vision is OpenAI's latest advancement, integrating visual capabilities into ChatGPT. It combines the power of GPT-3.5 and GPT-4 models to proficiently analyze various visual inputs, including photographs, & documents with text and images. GPT-4 vision assists users with tasks like: object detection, visual question answering, data analysis, and deciphering text. It is currently available for GPT plus users.

Tammy.ai (🔗 link)

Tammy AI provides AI-driven summarization capabilities for text and video materials. Users can effortlessly create concise summaries for any written content and delve into trending YouTube channels through its AI-generated video synopses. Tammy AI offers a convenient way to explore YouTube channels and videos by simply inputting a link and receiving a speedy summary.

🥽 3 TRENDS

Apple reportedly planning generative AI Features for iOS 18 in late 2024 (🔗 link)

Apple is rumored to be working on integrating generative AI capabilities into iOS potentially starting with iOS 18 in late 2024. Analyst Jeff Pu predicts Apple will build hundreds of AI servers this year and more in 2024.

The features would likely combine cloud-based and on-device processing. But Apple's rollout may be gradual, as it determines how to align generative AI with its privacy commitment. iOS 18 could see Apple debut chatbot-like functionality in Siri using large language models.

It’s unclear if a 2024 timeframe is realistic given Apple's reported lag behind competitors in generative AI.

New index highlights major gaps in the transparency of leading generative AI companies (🔗 link)

Stanford researchers have created an index evaluating transparency in the generative AI industry. The Foundation Model Transparency Index assesses companies across 100 indicators on how they build, deploy, and govern AI systems.

Initial scores highlight substantial room for improvement. The highest scores ranged from 47 to 54 out of 100. Low transparency poses risks around safety, accountability, and responsible AI practices. It also hinders policymaking and public understanding.

The index aims to benchmark transparency while spurring progress through voluntary company disclosures or new regulations like the EU's AI Act.

OpenAI reportedly ends development of chatbot model Arrakis (🔗 link)

OpenAI has reportedly ceased work on Arrakis, an experimental chatbot model once seen as a potential breakthrough. Arrakis was intended as a smaller, more efficient alternative to models like GPT-4. Sources suggest it failed to meet expectations during training.

Reasons for ending Arrakis may include massive training costs topping $1 million daily and inferior performance by simply scaling up GPT-4. Microsoft is also advancing its own small models amid rising enterprise demand.

Some technicians also worry Arrakis had flaws making it vulnerable to manipulation. While disappointing, ending ineffective projects may be prudent for OpenAI as it balances costs, ethics, and quality. The company's choices will influence the AI landscape as developers are desperate for safer, more affordable generative models.