Written by Corey Bakhtiary – Sr. Director, Strategy and Innovation 

What’s the difference between generative AI and AI? Is there a difference? (Other than generative AI clearly having a better PR agency, as it’s getting headline after headline after headline…)

What’s the difference between generative AI and Microsoft Copilot? Is there a difference?

How did all of this get started? How can it be put to the best possible use? What’s the governance plan for AI? What happens next?

Let’s first take a step back and look at the roots of generative AI. By no means is this a comprehensive look at the computer science history of AI — we’re not going back to Ada Lovelace or Charles Babbage — but let’s start with some classifications.

• Artificial Intelligence – field of computer science that seeks to create intelligent machines that can replicate or exceed human intelligence
• Machine Learning – subset of AI that enabled machines to learn from existing data and improve upon that data to make decisions or predictions (at the core of most predictive analytics tools)
• Deep Learning – a ML technique in which layers of neural networks are used to process data deliver benefits such as image recognition and speech recognition
• Generative AI – where ML works off existing data, generative AI creates new data — written, visual, and auditory content — off user prompts

What about Copilot?

Copilot is a Microsoft tool comprising workload-specific generative AI assistants. Copilot has the capability to synthesize large amounts of data and learn and iterate through interactions with humans in the loop. It finds applications in various creative fields such as image generation, visual design, content creation, summarization, search, and composition.

Additionally, Copilot is used in the development of documentation, code generation, unit testing, drug discovery, data augmentation, and reasoning. It is important to note that Copilot is not an auto-pilot system. It functions “alongside” users.

Where it all started: language models

Now let’s take a look at language models, a key component in deep learning and the bedrock of generative AI. These models are typically built upon deep learning architectures, particularly transformer-based architectures, which have revolutionized natural language processing (NLP).

At the core of language models is the capacity to learn patterns and structures within vast amounts of text data. For example, the “bag of words.” In the context of NLP, “bag of words” is a model for representing text data where a document is treated as a collection of words, rather than a narrative (i.e., a “bag”). Grammar and usage don’t really matter as each word is treated as a separate entity, and the frequency of each word’s occurrence in the document is counted.

What does this mean in the NLP context? Consider spam emails. It does not matter where “Urgent, act now” occurs, that message will be filtered as spam. This example is AI in its most basic form. But by processing massive amounts of text over time, these models can now extract intricate relationships between words, phrases, and context, enabling them to predict and generate coherent sequences of text themselves, responding appropriately to queries or generating novel content. I.e., generative AI.

The effectiveness of generative AI is in its ability to capture semantic meaning and syntactic structure. They employ techniques such as attention mechanisms to weigh the importance of different words in a sentence, allowing for more contextually relevant responses. Additionally, AI can be fine-tuned on specific tasks or web domains, further enhancing adaptability and performance.

The continuous advancements in language model architectures, along with the availability of large-scale datasets and computational resources, contribute to their improved capabilities over time.

And with 85% of knowledge industry workers wanting digital tools to automate information collection and accessibility, and the fact that ChatGPT — the first generative AI tool widely adopted by the public — acquired more 100 million users in just two months, it’s safe to say that this story has only just begun.

Where are you on your AI journey? Are you leveraging Microsoft generative AI across customer service? Marketing and sales? Product development? We can help create an AI roadmap for your business with a dual emphasis on governance and strategic growth. Contact us to get started.