AI and Its Impact on the Modern Enterprise - (First Part)

LEVERAGING AI FOR ENTERPRISE SUCCESS

Just in case you haven’t checked your email for the last 18 months, Artificial Intelligence (AI) and Machine Learning (ML) have rapidly emerged as transformative tools for modern organizations. As a senior ML consultant, I’ve observed firsthand the impact of these technologies across various sectors. Let’s take a look at a comprehensive overview of AI’s role in modern enterprises, focusing on practical applications, potential benefits, and key considerations for implementation.

UNDERSTANDING AI AND ML IN THE ENTERPRISE

The original intent of these tools many decades ago was to duplicate human intelligence using mechanical processes.  At its core, AI/ML refers to the development of computer systems capable of performing tasks that typically require human intelligence.  There are ongoing debates if these systems truly exhibit true reasoning or intelligence (or even if humans are uniquely intelligent!), but the surprising utility of the current AI/ML wave has huge implications for business organizations and the wider enterprise.

Just to be clear, ML is a broad concept encompassing various techniques for task-specific learning.  Deep learning, on the other hand, is a more sophisticated and specific type of machine learning that involves artificial neural networks with multiple layers to learn data patterns.  In the general trade press, it’s important to note that the term “machine learning” is more commonly used, even though the more accurate term is “deep learning.”

PREDICTIVE AI AND GENERATIVE AI

Within Deep Learning, there are two primary and important subsets: Predictive AI and Generative AI.

Predictive AI focuses on making predictions or forecasts based on structured, historical data. It aims to identify patterns in the data and use them to predict future outcomes.  Predictive AI models are trained on labeled datasets, where the algorithm learns the relationships between input characteristics and the corresponding target variable.  For example, the impact of rising temperatures on predicting the weather, or the color of a worn truck tire to anticipate a dangerous failure.  Predictive AI is particularly useful when there is a need to anticipate specific outcomes based on available structured data.  Weather prediction, image classification, recommendation systems, autonomous vehicles, anticipating hardware/software failures, detecting email anomalies, etc, are examples of predictive AI.  For many use cases, Predictive AI is an excellent technique to use and has been successfully deployed in production for at least the past 10 to 15 years.  There are several excellent production-quality Java toolkits for deploying Predictive AI, such as JSR 381 Visual Recognition, Amazon’s DLJ, and Deep Netts.

Another very important subset of Deep Learning is Generative AI or “GenAI”.  This is the type of AI that the world is currently excited about over the past 18 months. GenAI focuses on creating new data samples that resemble the input data it was trained on. Instead of predicting from existing data, GenAI generates novel, synthetic data based on learned patterns.  Generative AI is currently extremely popular for its ability to create human-like text, images, and sound.  Generative AI typically deals with a type of natural language processing (NLP) that uses an innovative architecture called “Transformers,” which was developed (and patented) at Google in 2017.  These systems use Large Language Models (LLMs) that are trained on huge amounts of data on the Internet to extract patterns.  While you will find popular LLMs available from large companies such as OpenAI, Google, and Microsoft, there currently are many thousands of open-source models that you can run directly on your laptop or in the cloud.  The growing popularity of these open-source LLMs is a significant trend to monitor, especially with distributed AI intelligence at the edge potentially being a big business opportunity.

Both forms of Deep Learning are powering new types of computing systems for enterprises.  Predictive AI is centered around making predictions based on existing patterns, while generative AI is focused on creating new, realistic data. While Generative AI certainly has captured the attention of the world’s enterprises (and countries!), both approaches have their own set of applications and are valuable in different contexts within the field of artificial intelligence.  There are also use cases that combine the strengths of each type of deep learning into a powerful combination.

SHORT HISTORY OF GENERATIVE AI

GenAI refers to artificial intelligence systems capable of generating new content.  This content can be text, images, audio, or other media, based on existing unstructured data. Unlike Predictive AI, which typically analyzes structured (and labeled) data to find patterns and make predictions, GenAI creates stochastic, non-deterministic outputs from learned patterns, opening up a world of exciting possibilities (and significant risks) for businesses. To fully grasp its potential, it’s essential to understand and appreciate the historical context and evolution of AI. Early AI has its roots in the 1940’s with artificial neural networks (ANNs).  This decade followed a period of intensive research in mathematical biology in the 1930s, which certainly influenced the development of ANNs.

It is sometimes hard to fathom the notion that ANNs were created before the invention of the software subroutine!

Nascent AI systems were focused on rule-based systems and simple decision-making algorithms, particularly during the 1980s.  With the advent of modern machine learning over the past 10-15 years, AI systems began to learn from data, improving their performance over time. Deep learning, a subset of machine learning, further advanced AI capabilities, enabling more complex tasks like image and speech recognition. GenAI represents the latest leap, capable of creating entirely new content. This feature is not totally new, since stochastic music generation systems based on musical styles existed in the 1970s and early 1980s.  However, with the advent of cloud computing resources and the recent breakthrough of the transformer architecture (2017) for natural language processing (NLP), the benefits (and risks) of GenAI are available for all organizations and enterprises.

MACHINE LEARNING MODELS AND TECHNIQUES

As with most ML systems, Generative AI involves using algorithms to create new data points based on the patterns from existing data. These ML models learn patterns and structures in the training data and use this knowledge to generate new, similar data. For instance, a Generative AI trained on text can produce coherent paragraphs, while one trained on images can create realistic pictures.  Once they are trained, large language models, such as GPT-4, generate text by predicting the next word in a sequence. Similarly, image synthesis models, like DALL-E, and Midjourney, generate images from textual descriptions.

Effective interaction with Generative AI requires understanding how to craft textual prompts that yield the desired outputs called “completions”. This involves using clear, specific language and providing sufficient context. Experimentation and iteration are key to refining specific prompts for optimal results.  It is not surprising that human teachers are excellent at crafting prompting techniques. 

Additional techniques, such as Retrieval Augmented Generation (RAG) and Fine-tuning, can further refine an LLM’s results. RAG combines generative models with retrieval systems to improve the quality and relevance of generated content. RAG techniques are a popular mechanism for allowing LLMs to understand the context of private enterprise documents. Fine-tuning models on domain-specific data can also enhance their performance, making them more useful for specific business applications. However, fine-tuning can be expensive and requires deep data science expertise, although new tools are making it easier.

In addition to these features, many of these AI systems allow extended functionality using user-defined functions and autonomous control of models (agents). While these features are quite powerful and show great promise, they are now in the early evolutionary stages of enterprise use.

More recently, sophisticated GenAI “multimodal” models have emerged.  Instead of using a collection of different models to handle multiple types of data, a multimodal model can natively understand the patterns of multiple types of data, such as text and images, and can generate similar data based on those diverse input types. OpenAI’s GPT-4o, Anthropic’s Claude 3, and Google’s Gemini 1.5 are examples of powerful multimodal GenAI models. It is fascinating to envision what innovative types of devices and applications will soon be created using multimodal AI models.

ENTERPRISE USE CASES

The potential of Generative AI spans numerous industries. In healthcare, it can generate synthetic data to enhance medical research. In finance, it can create realistic market simulations for better risk management. The entertainment industry can use Generative AI to produce original content, from music to movies. Several enterprises have already successfully integrated Generative AI. For example, IBM Watson uses Generative AI to assist in medical diagnoses by generating potential diagnoses based on patient data. Similarly, OpenAI’s Codex powers GitHub Copilot, helping developers by generating code snippets.

Generative AI can significantly enhance customer experience. AI-driven chatbots can provide personalized, real-time support, which can improve customer satisfaction. Generative models can also create customized product recommendations, potentially boosting sales and customer loyalty. Additionally, Generative AI can streamline operations by automating repetitive tasks. For instance, it can generate reports, summaries, and other documents, freeing up employees to focus on higher-value activities. This last benefit alone is hugely valuable to most organizations.

Just to be clear, GenAI and PredAI are not mutually exclusive.  PredAI models typically focus on classification, prediction, or recommendation tasks.  Currently, the business value of PredAI overwhelms GenAi due to existing ecommerce, recommendation systems, and production visual recognition systems. In contrast, Generative AI, on the other hand, creates new content, enabling a different range of applications. PredAI might predict customer churn, and GenAI could generate personalized marketing content to retain those customers.

One major concern is the generation of biased or harmful content, reflecting biases in the training data. There’s also the risk of explicit misuse, such as generating deepfakes or misleading information. To mitigate these risks, it’s crucial to address biases in training data and ensure robust data privacy measures. Techniques like bias mitigation algorithms and differential privacy can help.

Monitoring, regular audits, and transparency in AI processes are also essential. Implementing comprehensive risk management strategies is also vital. This includes establishing clear guidelines for AI use, continuous monitoring of AI outputs, and creating feedback mechanisms to address any issues promptly.  Vendors such as ValidMind, ZenGRC, and guidance/regulation from government groups such as NIST and the EU AI Act are now laser-focused on the risks from generative AI models.  Collaborating with these external vendors/experts and business stakeholders can also provide valuable insights and support.