The researchers taught an AI to create pretend DNA
File this at: While the rest of the world was busy getting virtual assistants to tell papa jokes, researchers in Estonia figured out how to fake people at the molecular level.
We’ve seen a delightful onslaught of AI-generated content from Generative Adversary Networks (GANs) in the past few years trying to create what are known as “deepfake” images.
There is that person doesn’t exist. And that cat doesn’t exist. You can even generate feet, resumes, and waifu that aren’t there. It’s amazing how realistic some of the aerial imagery created by the AI can be.
However, this is the first time an AI has generated the recipe for a viable, unique human by creating synthetic genomes.
A team of researchers from Estonia has developed a machine learning system that can be used to generate unique genome sequences. These computer-generated counterfeits could play an important role in the future of DNA research.
According to the team’s research report:
Generative neural networks have been used effectively in many different fields over the past decade, including machine-dreamed photo-realistic images. In our work we apply a similar concept to genetic data to automatically learn its structure and produce, for the first time, realistic genomes of high quality.
These AI sequences, known as “artificial genomes,” are indistinguishable from actual human genomes, except that they are entirely synthetic. This means researchers do not have to deal with ethical privacy concerns.
According to the current research paradigm, researchers must protect DNA to ensure the privacy of the people who own it. This comes with added accuracy and, in many cases, a drought in available data as facilities cannot share their data sets. Synthetic genomes should go a long way in solving these problems.
Take quickly: This is fantastic news for medical researchers, and a clear case in point where GAN technology will be used forever. However, this work highlights some ethical problems in the near and distant future that we will one day have to face.
In the near future, it will be easier for bad actors to create fake personas that can withstand even the strictest of inspections. Not that we imagine a scenario in which a fraudster has to produce a fake copy of his genome, but it is in the unknown unknowns that vulnerabilities grow fastest.
In the long run … if Skynet had this GAN it wouldn’t have had to make so many machines that look like Arnold Schwarznegger. Someday this technology could reach a point where the philosophical question “Am I a robot?” Becomes a very valid question for any person.
Read the whole paper here.
Published on February 8, 2021 – 22:07 UTC