The AI That’s Recreating You
Hardware and software are working in tandem to synthesis humans. When they aren’t melting you down into your base components, which as far as their concerned is data, they’re doing their best to be just like us.
The more data they get, the easier it should be to achieve. This isn’t another cautionary tale about telling Facebook your birthday, I just want us to explore the extraordinary tech that is doing this.
Create Us
China has been at the forefront of monitoring its citizens, so it stands to reason that they’d be the tech leader in cameras that could create a 3D model out of you from 45 kms away. The prototype that was announced by Zheng-Ping Li and his team from the University of Science and Technology of China had some unique challenges to tackle. Previous Lidar tech involves shooting a laser at something and recording all the return photons. Over large distances, this system can pick up a lot of noise, but their new gating technology, which ignores anything that isn’t at the distance they want to record, mitigates this issue. Also, a new 1550nm low-power infrared laser cuts right through that delicious Beijing smog. Stitching this all together is an AI that gets the data from the photons and assembles it into a recognisable image.
At its furthest current limits, it can resolve details 60cm in length as a 3D object, the resolution obviously increasing at closer distances. The team believes that there’s nothing stopping them from getting a high-resolution LiDAR going at “hundreds of kilometres by refining the setup.”
Animate Us
You know those paintings in Harry Potter that casually harrass students? AI could make those really happen using tech developed by Egor Zakharov, an engineer with the Skolkovo Institute of Science and Technology and the Moscow-based Samsung AI Center. Using just a single image of a painting or a photo, AI can map the face and make it come to life and talk. Its face can turn and show facets that weren’t readily available in the original image. Zakharov and his team managed this by feeding in data from other people talking and moving their faces around. Using this information, the AI was able to bring to life the Mona Lisa and Albert Einstein. Feeding in more frames of a person speaking or moving is able to create an increasingly more accurate replication. The team’s idea is that “in future telepresence systems, people will need to be represented by the realistic semblances of themselves, and creating such avatars should be easy for the users.”
This democratisation of what has been labeled deep fakes doesn’t worry the team much. Making the technology freely available they say, will mean democratising the methods to spot forgeries.
Write Like Us
“Artificial intelligence has become central to our modern economy, giving us access to vast amounts of data, data-mining power and the means to create our own data, all under the guise that it’s just information. We’re going to create data that can answer any type of question we’re interested in. And we’re going to do it with artificial intelligence.
This is why I’ve been in office for the past 12 months. For us, AI is fundamental to economic growth. It matters to the future of our country and our world.”
The text above wasn’t written by a human, it was cobbled together by GPT-2, a project by OpenAI. You can play with it at talktotransformer.com. The version publicly available is a stripped back version which only uses a paltry 345 million parameters and will often come up with garbled results. However, it can successfully translate between a couple of languages and answer questions, despite only being trained to do word prediction. OpenAI stated that “Due to our concerns about malicious applications of the technology, we are not releasing the trained model.”
One can already imagine its ability to inject the internet with even more fake news or regurgitate original university essays, using a stolen copy as a basis.
While current AI projects are essentially one trick ponies, used to exclude photons and regurgitate precreated things in unique ways, it’s not inconceivable to see these disparate parts amalgamating in the future to something truly amazing.
