Auta verkostoasi oppimaan uutta!
At the Dun & Bradstreet 2022 Power of Data event, Future Today Institute CEO Amy Webb wanted to talk about cats. Ten years ago, a neural network taught itself how to recognise cats. Today AI can generate an infinite number of cats.
Webb went through this timeline to illustrate that data and AI are picking up speed. She called these technologies enablers that will give business opportunities to design the future.
However, while (most) cats are cute, the ramped-up capacity to generate fake images can be put into play in endless arenas. Just tell the AI what you'd like a picture of and even when. But surely we could tell the difference? Yes, was Webb's answer, we could – emphasis on the past tense. Whether we can do so today is the big question.
Up until recently, one could spot a real photo quite easily. In family photo albums, for examples, the hair styles, clothes and even the hue of aged analog prints tell us what year, or thereabouts, an image was taken. As the Power of Data audience looked up at a faded snapshot on the screen, Webb pointed out that AI can now replicate not just the content but the feeling of a scene.
That can pose problems. The quality of the resulting pictures, according to Webb, will be so convincing that there's a risk of creating fake memories.
The same goes for film. A neural rendering can capture and generate 3D video from just a few 2D snapshots. The result is video that looks convincingly real. AI can create fake scenes that Webb described as immersive.
Those who master these kinds of processes will have immense power, she cautioned. The audience was asked to ponder what would happen if a country starts generating fake memories to brainwash its citizens? The Power of Data audience were both enthralled and quite visibly disquieted.
For business, fake images and videos could pose a reputational risk. What if a video tells a bad story about your company or your CEO?
Yet all was not doom and gloom as Webb ended her keynote address by pointing out that we can and should focus on the multiple potential versions of a future where data is harnessed for positive ends.