How to prepare for the upcoming generative and responsible AI revolution.

This article was originally published on Medium

What do you think of when you hear “Artificial Intelligence”?

AI is now part of our daily lives. Ideas that used to belong to science fiction or advanced academic research are real. Facial recognition technology helps us cross international borders and unlock our phones. Our favorite online stores point us in the direction of products we may want or need that we didn’t even know existed. And autocorrect saves us time when typing and prevents spelling mistakes (although left unchecked it can cause some embarrassing moments — just Google “autocorrect fails”).

We’re constantly hearing about the latest advances made possible by technologies such as computer vision, machine learning, natural language processing…the list is long, there are many use cases and several success stories. We’re already using all these technologies on a daily basis. Whisper it quietly, but some are now so readily available that they’re almost becoming a commodity.

So, what can we expect in the next wave of AI innovation?

While these technologies are here to stay, there’s so much exciting innovation around the corner. And this is why you need to prepare yourself and/or your organization for the next wave. Whether you’re a product manager, algorithm engineer or researcher; a leader in finance, marketing, sales or HR, it’s important to understand what new developments to expect in the next 2–5 years. Early-adopting companies that harness this coming wave smartly will be able to enjoy a significant advantage over their competitors.

I believe that the two most important buckets of innovation are generative AI and responsible AI — and I’ll go into more detail about both shortly. Of course, no one knows exactly what the future will bring. But by looking at dynamics and known models of technology adoption we can get some very strong indicators of what to expect.

A useful model to explore is the Gartner Hype Cycle — a methodology that represents the maturity and adoption of technologies graphically, showing how they will evolve over time.

When technologies are first developed, there’s a lot of hype and expectations can be seriously inflated (weren’t we all supposed to be driving automated cars, shopping in cashier-less supermarkets, living in a virtual reality world and having flawless interactions with our phone’s virtual assistants by now?). As a result, there’s normally some disappointment, but if you stick with the innovation it will eventually reach the plateau of productivity and change the world.

Gartner’s Hype Cycle model for Artificial Intelligence

As you can see in this graphic representing the AI sector, the technologies that most of us associate with artificial intelligence are already after the hype, and some (highlighted in grey) are on the way toward the plateau of productivity. This means that these technologies are appearing everywhere and adopters get a much smaller competitive advantage — if any — because in many cases they’re playing catch-up.

So there’s a huge opportunity here. Look at all those technologies that didn’t yet peak. Most people have never heard of them yet — but they’re going to change the world as well.

What’s the next wave?

In the next wave, we’ll see two major trends: responsible AI (highlighted above in purple) and generative AI (in pink).

Responsible AI is about making sure that we use these new technologies safely, fairly, and ethically — and to do so they need to be explainable. Artificial intelligence is an incredible, life-changing tool that will continue to revolutionize our world. And yet, as it becomes more advanced and more pervasive in our lives, the risks grow.

AI can develop biases or make mistakes. For example, Amazon had to stop using CV screening software that showed a bias against women and an algorithm used in US courts was biased against black people. In extreme cases, errors in sectors such as health care or autonomous cars could be fatal.

The security risks also need to be managed. We need to protect the huge amounts of data that are being used to train the models. And a bad actor could affect an algorithm connected to critical infrastructure with catastrophic results.

These are just some of the risks, which can be mitigated by making the AI explainable (there is still much for us to understand about how neural networks work) and then building “human-centric” AI, considering how best to have a “human in the loop” for control while still harnessing the power of artificial intelligence.

We’ll see more and more conversations about fairness, responsible AI and ethical AI — all of which we take very seriously at Bria. It’s a hot topic right now with the justified concerns of Open AI’s team about the recent release of DALL-E 2, which is not yet fully controllable or predictable. You’ll also hear the term AI TRiSM (AI Trust, Risk & Security Management) which can be used to encompass many of the above issues.

The second trend is that we’ll see more and more technologies harnessing generative AI. This is when the AI creates a completely new set of synthetic data — which has a huge amount of potential use cases. At Bria, we’re using this technology to generate photorealistic images and videos. So, what makes this so different from the AI that we all know so well?

Until recently, artificial intelligence was discriminative

The AI we benefit from on a daily basis is discriminative. Its role is to help us better understand data that already exists.

Discriminative AI can help with recognition (for example, text sentiment identification or facial recognition software), understanding (chatbots understanding user intent), predictions (where and when to expect heavy traffic, for example) or recommendations (which show to watch next on Netflix).

Generative AI — it’s in the name

Generative AI enables computers to learn the underlying pattern related to an input, and then use that to generate similar content completely from scratch. This is known as synthetic data, as it’s been artificially created.

There are already several companies making significant advances in this space. For example, DataGen generates synthetic data that can be used for training in machine learning. Hour One and D-ID create visual bots. In the text space, OpenAI offers the revolutionary text generator GPT3 while AI21 Labs is more focused on short-form text. At Bria we’re powering visual communication by creating photorealistic images and video with no need for a camera, photoshop or graphic designer.

While this technology is still at an early stage, it can already produce some impressive results: 

The future of AI and its use at companies is up in the air,” said Mr. Joffe. “That could help us learn and develop new ways to improve processes.”

In other words, AI will become a tool that humans can use to be smarter about AI. That doesn’t mean everyone gets it. The work of researchers in Japan and in Switzerland has shown how artificial intelligence can be harnessed. Others, however, do not see any reason to expect that AI will someday supplant human beings at work.

Text generated from initial input of “the future of AI” by DeepAI’s text generator Illustration generated from the input “Teddy bears shopping for groceries in the style of ukiyo-e” using Open AI’s Dall-E 2

Predictability and controllability in generative AI can allow you to tell exactly the story you want to tell. In this example we have a different story if the woman is a wife or if she is a daughter; if she is happy with the results of the test or concerned. These visuals were generated by Bria

Generative AI makes responsible AI a must

The moment that AI starts to create and not just discriminate, responsible AI becomes even more important. We have to be able to predict and control what the AI generates to ensure fairness, reduce bias and minimize any ethical risk.

It’s for this reason that at Bria our focus is on predictability and controllability and we recently appointed a responsible AI advocate to help guide us on this journey. This field is so new that we and companies like us are writing the rule book, and we have to get it right — because this technology will soon be everywhere.

The future of visual communication will be powered by generative AI

Think about the creative process today for a visual — whether an illustration, a photograph or a video. You need equipment — whether it’s a camera, a stylus or an old-fashioned pen and paper. You need the skill to translate your idea into an image — either by drawing it, capturing the right frame or being familiar with Photoshop. And you need budget — for the materials, talents or production and post-production.

Generative AI turns all that on its head. Anyone can now create a visual. Creativity is no longer limited by resources, talent or skills. The top creatives and designers of the future will be those with the best imaginations, not the best technical skills, because all they will need to do is articulate their vision and the AI will generate it for them.

At Bria, we call this creative independence; building a world where it’s possible is what drives us forward. We’re already helping thousands of people unleash their creativity — and you’re welcome to try the free beta for yourself.

As AI technology advances, it seems that a new frontier is being explored every day. But this story goes beyond the world of technology. These developments represent huge opportunities for society as a whole — and with those opportunities also come risks that we need to anticipate and manage accordingly.

So, let’s have the conversation. I’d love to hear what you think — share your thoughts in the comments or send me an email.