Navigating AI Terrain: Generative vs. Non-Generative Perspectives

Hey! 👋

As a Product Manager and previous Founder, I have always been fascinated by the ever-evolving landscape of artificial intelligence (AI) and what a better time to explore than before the AI Summit?!

I managed to speak to one of the workshop speakers at the Summit who is an experienced ML and Data Scientist for their thoughts too.

One of the most intriguing aspects of AI is the distinction between Generative and Non-Generative approaches. So, grab your metaphorical compass, because this blog post explores the two differences within the realm of AI!

First things first, let's break it down. Generative AI, as the name suggests, has a knack for creativity. It's like having an AI artist in your toolkit, capable of generating new data based on existing patterns. Think stunningly realistic images, captivating pieces of text, and even melodies that tug at your heartstrings. On the flip side, Non-Generative AI operates within predefined boundaries, sticking to the rules and regulations set by its human overlords.

An example of generative AI:

These images were created by generative AI using the prompts light bulb, flower, pastel, geometric shapes, simplicity, clean lines, and minimal still life.  Midjourney. Source: Harvard Business Review, 2023

Now, let's explore applications. Generative AI isn't just about creating pretty pictures (although it certainly has flair!). From generating synthetic data for training models to enhancing content creation and even assisting in drug discovery, the possibilities are endless. Meanwhile, Non-Generative AI holds its own in tasks like predictive analytics, fraud detection, and personalised recommendations. It's the workhorse of the AI world, churning out insights faster than you can say "data-driven decision-making."

Speaking to David Dixon, Machine Learning and Data Scientist, it was interesting to learn that traditional AI (aka Non-Generative AI) is much easier to implement in workplaces and can actually speed up many of the common tasks employees do on a day-to-day basis. Take Hertzian’s Player XP AI solution, which can process more data volume than chat GPT for the same cost, x16 as much! This means employees can spend their time do the more interesting tasks, rather than collating and processing data.

Traditional AI methods, with their clear-cut rules and regulations, are often easier to implement. Generative AI, on the other hand, can be a bit of a wild card. It requires copious amounts of data and hefty computational resources to truly shine.

David Dixon, Machine Learning and Data Scientist, explains: ‘Generative AI is, at it's core, unpredictable in its output, and while these large language models like ChatGPT can give the illusion of intelligence, they act closer to more sophisticated predictive text. Because of this, automating generative AI for business-related tasks is near impossible to do with complete accuracy. On the other hand, traditional AI is constrained by design to give a limited range of results and as such is much easier to implement.’

When navigating the AI landscape, it's essential to weigh up both options. Consider:

1. Your project goals

2. Available resources

3. Data requirements

There is no one-size-fits-all solution. It's all about finding the right tool for the job.

A top tip from an expert:  “Not everything needs to be AI. There are often lots of more economical solutions and the simpler solution can also produce better results. AI is the hot thing but it doesn't mean it's right for you.”

With advancements in Generative AI like GANs (Generative Adversarial Networks) pushing the boundaries of creativity, and the integration of Generative and Non-Generative approaches opening up new possibilities, the sky's the limit. So if this has piqued your interest find out more at the Hertzian AI Summit (6-7th June 2024). You can book tickets here.

Written by Tanuvi Ethunandan

Product Manager at Hertzian

‍