Blog

What you need to know about… Generative AI

This blog will help you come off as a pseudo expert anytime AI comes up in conversation (which is all the time nowadays)

Param Gopalasamy
OneTrust Editorial Team, CIPP/E, CIPP/US, CIPM
November 3, 2023

What is generative AI?
The evolution from predictive to generative AI
The building blocks of generative AI
The business use case for generative AI
Privacy and ethical concerns in generative AI

In a digital tapestry, vibrant and infinite, where data weaves and dances, there exists a maestro known as generative AI. With strokes bold and subtle, it crafts, creates, and conjures images, sounds, and words that echo with the rhythm of invention.

Ugh. That statement was written by generative AI, and it clearly doesn’t mind talking itself up. The art analogy wasn’t bad though – we’ll stick with that throughout the article.

What is generative AI?

Generative AI is like a digital artist. It doesn't merely analyze or process; it creates. With data as its paint and algorithms as its brushes, this form of artificial intelligence generates new content, which sets it apart from its predecessor, predictive AI, by a significant margin. We no longer just get the weather forecast; we can get a detailed summary of what outfits we should wear each week based on the weather.

In this blog, we’ll talk about some of the technical building blocks of generative AI, dive a bit deeper into its different applications, the effect this is having on the creative process across industries, and what the future holds.

Get ready to learn way more about AI than you ever thought you’d need to know.

The evolution from predictive to generative AI

As we previously discussed, AI has been around in our lives for some time. Its previous avatar was predictive AI – which was widely used across industries and use cases. The name is a bit self-explanatory, which is … you guessed it, predicted things. In a nutshell, you give a predictive AI system a bunch of data, you’ll get a prediction on what’s going to happen next. Think stock markets, financial targets, the weather, even Netflix’s recommendations for you.

However, despite its prowess, Predictive AI’s canvas was only half-painted (bringing that art analogy back). It could analyze and predict, but when it came to creation, its hands were tied. Predictive AI could tell you the probability of rain tomorrow but couldn’t paint you a picture of the storm. And as someone who has interacted with ChatGPT quite a bit, let me tell you, generative AI has no problem painting a very verbose picture.

And so, last year, Generative AI made its entrance. Unlike its predictive counterpart, Generative AI doesn’t just analyze; it imagines and creates. It takes the baton from predictive models and runs forward, generating new data that is coherent and relevant (sometimes). There’s still some work to be done on the coherent and relevant part, as Google Bard’s public mishap was a clear sign to the world that we’re not quite there yet.

The leap from analysis to creation

This transition from predictive to generative AI represents a leap from understanding the world to adding something new to it. While Predictive AI might inform a doctor about a patient's likelihood of having a disease, Generative AI could assist in creating personalized treatment plans or even simulate the progression of the disease under different conditions.

Use cases at this level of significance are what’s driving the mass excitement (some may say hysteria) around Generative AI. Given the possibility of having an accurate AI system that can provide relevant, timely solutions to seemingly any problem – it's not too hard to see why the hype train has been rolling.

The building blocks of generative AI

This is the section where you can really differentiate yourself from standard “AI hype train person” to “Wow, they really know what they’re talking about!”.

If you want to know how generative AI models really work, there are three primary elements involved:

GANs – Generative Adversarial Networks
VAEs – Variable Auto-Encoders
Transformer Models – No futuristic-sounding acronym here, although transformer sounds cool enough

When a prompt walks into an AI studio

Imagine you walk into an art studio (the generative AI system) with an idea or a request. It's like telling an artist, "Hey, I want something that looks like a sunset over a city." Let's see how the team of artists and storytellers – the GANs, VAEs, and Transformers – work together from that prompt:

Sketching with GANs

Your idea first goes to the GANs. As the name suggests, the Generative Adversarial Network consists of two entities:

The Drawer (Generator) tries to sketch out a basic picture of a sunset over a city.
The Checker (Discriminator) looks at it and suggests improvements. Maybe the colors need to be warmer, or the buildings need more detail.

After a few rounds of sketching and checking, the GANs come up with a detailed and beautiful sketch of the sunset scene.

Imagine coming up with some copy or a picture and having an annoying proofreader immediately correct you afterwards multiple times. That’s what this stage is (no wonder the potential of AI is said to be more than humans).

Sculpting with VAEs

Now that we have a sketch, the VAEs step in. They interpret this sketch and think about all the different ways it can be enhanced. Maybe they imagine 3D buildings popping out or the light of the sunset casting realistic shadows. Using their knowledge of data, they add depth and dimension to the scene, turning the flat sketch into something more lifelike.

This would be like an architect or sculptor taking a look at a sketch and coming up with ideas of how to bring it to life.

Storytelling with Transformers

With our image getting richer, the Transformers jump in. They want to add a narrative or a backstory. Why is the sunset important? Who lives in those buildings? Think of any marketing professional you know – and this step suddenly makes a lot more sense.

They craft a description or a story: "As the sun dipped below the horizon, the city's lights began to twinkle. Each window told a story, from the baker ending his day to a writer just beginning her novel under the warm glow of a lamp."

This final step is what adds context and the human-like element to generative AI.

These three elements together turn a simple prompt into cohesive instructions, descriptions, images, or more – as each layer continues to add depth to the AI system’s response.

The business use case for generative AI

Now in the business world, the term generative AI has reached “Payphone” by Maroon 5 levels of being overused and overplayed at this point. But as with “Payphone” (which was an absolute banger), it’s not without reason. Here are just a few use cases across different departments that are already starting to use generative AI effectively.

Marketing: Need a tagline? How about 500? Or perhaps the old writers’ block has set in. You now have an outlet to draft that initial piece of content for you (definitely never used this myself).
Design: Whoa that’s a cool logo. Well, it was AI generated. That’s going to happen more and more often, with a lot of corporate images that you see.
Financial forecasting: What’s the next quarter gonna be bro? What are our targets bro? My impression of finance/sales bros asking an AI system these questions.
Coding: As much as every coder loves troubleshooting bugs all day, sadly, they’ve got the option of an AI fixing it. Not that the purists will use it, I’m sure. Even generating new bits of code – turns out Skynet has you covered.
Executive level: Executives, so busy. Who’s got time for paragraphs? Gotta get those meeting notes into five bullet points. Well, now you can with a simple prompt.
Customer support: Ever heard of a chatbot? One that’s not useless and keeps giving you the same structured responses? We’re getting there. The AI voices on customer support phone lines are starting to get more intonation too. Scary.

Again, these are just a few of the applications in the business world that can currently be powered by AI. Presentations, social media, workplace productivity tools, and CRMs are also just a few more ways AI can take over our world.

Privacy and ethical concerns in generative AI

A big concern when it comes to generative AI is the privacy and ethical side of things.

Who’s consented to using this data?
Are they even in a position to give consent for this type of data use?
How do we explain to people what the AI tool does with their data, so that they can understand?
How can I get access to or delete my personal data that AI is (mis)using?
Do original creators that AI pulls from get the proper credit?
Is sensitive data at risk when it comes to AI systems?
What happens when the AI system is just wrong? What are the repercussions of inaccurate answers?
How can biases be handled in a fair way when it comes to AI?

These are just a few of the very real concerns that surface when it comes to generative AI in the realm of privacy and ethics. The three principles below can help you start to prioritize privacy when introducing AI solutions into your organization.

Responsible data use: Consent, legitimate interest, or contract – depending on the use case of the AI system, your organization needs to understand what tools to rely on to use that data.
Anonymized data: When using personal data in AI systems, making sure that this data is unidentifiable is one way to ensure that privacy is honored for your customers or users. Using synthetic data that’s computer-generated is another way to go about training your AI systems without putting personal data at risk.
Privacy-first AI: Ensuring that a privacy framework is “baked in by design” into the AI system that your organization uses is key. Developing an AI system and processes and applying a privacy framework after the fact is always going to be tougher. Think of that idiom about putting toothpaste back in the tube.

Constructing ethical and privacy frameworks

For the canvas of your AI systems to have a place in your organization, they need to rest on an easel of ethical and privacy frameworks (this may be the last art analogy of the article).

Using the following ethical AI guidelines will help you address the primary ethical concerns that an AI system raises:

Human oversight: Having a human-in-the-loop system where the final decision is not solely made by the AI goes a long way to preventing inaccurate, unfair decisions.
Transparency: Communicate with your customers and users exactly what decisions the AI systems are used for to avoid any surprises.
Privacy and data governance: Ensure all data being used by these systems is accurately classified and has the necessary protections in place based on sensitivity.
Technical safeguards: When it comes to people’s data, your security measures need to be top of the line, meeting or exceeding the industry standards.
Bias monitoring: When decisions are made by the AI system, ensure that biases are not creeping into the AI system, with regular feedback.

Just to make things easier for you in conversation, here’s a list of “AI don’ts” for you to spout off whenever the opportunity presents itself.

Never share personal, sensitive, or confidential information with AI systems unless it’s approved by the InfoSec team and is an essential functionality of the system.
Never rely solely on the outcomes that AI systems produce. Always review them, keeping the system’s limitations in mind.
Never take AI system content as a substitute for legal, financial, or professional advice (or relationship advice - ever watched Her?).
Never input information that you want to access, modify, or delete into an AI system.
Never input hateful, offensive, malicious, misleading, or inappropriate information into AI systems.

So now that we’ve covered generative AI today, its building blocks, and concerns around it, hopefully, you’re a bit more qualified to hold a 5-minute discussion at any dinner table and come off as at least the second-most intelligent person there.

Time for a shameless plug: If you actually want to learn more about AI governance beyond sounding mildly informed over a meal – be sure to check out OneTrust AI Governance.

Blog

What you need to know about… Generative AI

Table of contents

What is generative AI?

The evolution from predictive to generative AI

The leap from analysis to creation

The building blocks of generative AI

When a prompt walks into an AI studio

The business use case for generative AI

Privacy and ethical concerns in generative AI

Constructing ethical and privacy frameworks

You may also like

AI Governance

Overcoming the privacy pitfalls of GenAI (APAC)

Join us and learn about the data privacy risks of adopting GenAI and practical strategies on avoiding them.

January 08, 2025

Responsible AI

Overcoming the privacy pitfalls of GenAI

This webinar will explore the key privacy pitfalls organizations face when implementing GenAI, focusing on purpose limitation, data proportionality, and business continuity.

December 10, 2024

Responsible AI

Getting Ready for the EU AI Act, Phase 1: Discover & Catalog, The Gartner® Report​

Getting Ready for the EU AI Act, Phase 1: Discover & Catalog, The Gartner® Report​

October 16, 2024

AI Governance

California's approach to AI: Unpacking new legislation

This webinar unpacks California’s approach to AI and emerging legislations, including legislation on defining AI, AI transparency disclosures, the use of deepfakes, generative AI, and AI models.

October 15, 2024

AI Governance

Securing reliable AI solutions: Strategies for trustworthy procurement

Download this eBook to explore strategies for trustworthy AI procurement and learn how to evaluate vendors, manage risks, and ensure transparency in AI adoption.

September 12, 2024

AI Governance

Ensuring compliance and operational readiness under the EU AI Act

Join our webinar and learn about the EU AI Act's enforcement requirements and practical strategies for achieving compliance and operational readiness.

August 22, 2024

AI Governance

OneTrust AI Governance demo video

Learn how OneTrust AI Governance acts as a unified program center for AI initiatives so you can build and scale your AI governance program

August 13, 2024

Responsible AI

Privacy and AI: Bridging the divide

Watch this webinar for insights on ensuring responsible data use while building effective AI and privacy programs.

July 31, 2024

AI Governance

AI governance masterclass miniseries: EU AI Act

Discover the EU AI Act's impact on your business with our video series on its scope, roles, and assessments for responsible AI governance and innovation.

July 31, 2024

Responsible AI

EU AI Act compliance resource kit

Download this resource kit to help you understand, navigate, and ensure compliance with the EU AI Act.

July 22, 2024

AI Governance

From build to buy: Exploring common approaches to governing AI

In this webinar, we'll navigate the intricate landscape of AI Governance, offering guidance for organizations whether they're developing proprietary AI systems or procuring third-party solutions.

July 10, 2024

AI Governance

Navigating the ISO 42001 framework

Discover the ISO 42001 framework for ethical AI use, risk management, transparency, and continuous improvement. Download our guide for practical implementation steps.

July 03, 2024

AI Governance

AI Governance Leadership Webinar: Best Practices from IAPP AIGG with KPMG

Join out webinar to hear about the challenges and solutions in AI governance as discussed at the IAPP conference, featuring insights and learnings from our industry thought leadership panel.

June 18, 2024

AI Governance

Colorado's Bill on AI: Protecting consumers in interactions with AI systems

Colorado has passed landmark legislation regulating the use of Artificial Intelligence (AI) Systems. In this webinar, our panel of experts will review best practices and practical recommendations for compliance with the new law.

June 11, 2024

AI Governance

Governing data for AI

In this webinar, we’ll break down the AI development lifecycle and the key considerations for teams innovating with AI and ML technologies.

June 04, 2024

AI Governance

Global AI Governance law and policy: Jurisdiction overviews

In this 5-part regulatory article series, OneTrust sponsored the IAPP to uncover the legal frameworks, policies, and historical context pertinent to AI governance across five jurisdictions: Singapore, Canada, the U.K., the U.S., and the EU.

May 08, 2024

AI Governance

Embedding trust by design across the AI lifecycle

In this webinar, we’ll look at the AI development lifecycle and key considerations for governing each phase.

May 07, 2024

Getting Ready for the EU AI Act, Phase 1: Discover & Catalog, The Gartner® Report

Getting Ready for the EU AI Act, Phase 1: Discover & Catalog, The Gartner® Report