New AI improves itself through Darwinian-style evolution

AutoML-Zero is a proof-of-concept project that suggests the future of machine learning may be machine-created algorithms.

New AI improves itself through Darwinian-style evolution
Pixabay
  • Automatic machine learning is a fast-developing branch of deep learning.
  • It seeks to vastly reduce the amount of human input and energy needed to apply machine learning to real-world problems.
  • AutoML-Zero, developed by scientists at Google, serves as a simple proof-of-concept that shows how this kind of technology might someday be scaled up and applied to more complex problems.

Machine learning has fundamentally changed how we engage with technology. Today, it's able to curate social media feeds, recognize complex images, drive cars down the interstate, and even diagnose medical conditions, to name a few tasks.

But while machine learning technology can do some things automatically, it still requires a lot of input from human engineers to set it up, and point it in the right direction. Inevitably, that means human biases and limitations are baked into the technology.

So, what if scientists could minimize their influence on the process by creating a system that generates its own machine-learning algorithms? Could it discover new solutions that humans never considered?

To answer these questions, a team of computer scientists at Google developed a project called AutoML-Zero, which is described in a preprint paper published on arXiv.

"Human-designed components bias the search results in favor of human-designed algorithms, possibly reducing the innovation potential of AutoML," the paper states. "Innovation is also limited by having fewer options: you cannot discover what you cannot search for."

Automatic machine learning (AutoML) is a fast-growing area of deep learning. In simple terms, AutoML seeks to automate the end-to-end process of applying machine learning to real-world problems. Unlike other machine-learning techniques, AutoML requires relatively little human effort, which means companies might soon be able to utilize it without having to hire a team of data scientists.

AutoML-Zero is unique because it uses simple mathematical concepts to generate algorithms "from scratch," as the paper states. Then, it selects the best ones, and mutates them through a process that's similar to Darwinian evolution.

AutoML-Zero first randomly generates 100 candidate algorithms, each of which then performs a task, like recognizing an image. The performance of these algorithms is compared to hand-designed algorithms. AutoML-Zero then selects the top-performing algorithm to be the "parent."

"This parent is then copied and mutated to produce a child algorithm that is added to the population, while the oldest algorithm in the population is removed," the paper states.

The system can create thousands of populations at once, which are mutated through random procedures. Over enough cycles, these self-generated algorithms get better at performing tasks.

"The nice thing about this kind of AI is that it can be left to its own devices without any pre-defined parameters, and is able to plug away 24/7 working on developing new algorithms," Ray Walsh, a computer expert and digital researcher at ProPrivacy, told Newsweek.

If computer scientists can scale up this kind of automated machine-learning to complete more complex tasks, it could usher in a new era of machine learning where systems are designed by machines instead of humans. This would likely make it much cheaper to reap the benefits of deep learning, while also leading to novel solutions to real-world problems.

Still, the recent paper was a small-scale proof of concept, and the researchers note that much more research is needed.

"Starting from empty component functions and using only basic mathematical operations, we evolved linear regressors, neural networks, gradient descent... multiplicative interactions. These results are promising, but there is still much work to be done," the scientists' preprint paper noted.

U.S. Navy controls inventions that claim to change "fabric of reality"

Inventions with revolutionary potential made by a mysterious aerospace engineer for the U.S. Navy come to light.

U.S. Navy ships

Credit: Getty Images
Surprising Science
  • U.S. Navy holds patents for enigmatic inventions by aerospace engineer Dr. Salvatore Pais.
  • Pais came up with technology that can "engineer" reality, devising an ultrafast craft, a fusion reactor, and more.
  • While mostly theoretical at this point, the inventions could transform energy, space, and military sectors.
Keep reading Show less

China's "artificial sun" sets new record for fusion power

China has reached a new record for nuclear fusion at 120 million degrees Celsius.

Credit: STR via Getty Images
Technology & Innovation

This article was originally published on our sister site, Freethink.

China wants to build a mini-star on Earth and house it in a reactor. Many teams across the globe have this same bold goal --- which would create unlimited clean energy via nuclear fusion.

But according to Chinese state media, New Atlas reports, the team at the Experimental Advanced Superconducting Tokamak (EAST) has set a new world record: temperatures of 120 million degrees Celsius for 101 seconds.

Yeah, that's hot. So what? Nuclear fusion reactions require an insane amount of heat and pressure --- a temperature environment similar to the sun, which is approximately 150 million degrees C.

If scientists can essentially build a sun on Earth, they can create endless energy by mimicking how the sun does it.

If scientists can essentially build a sun on Earth, they can create endless energy by mimicking how the sun does it. In nuclear fusion, the extreme heat and pressure create a plasma. Then, within that plasma, two or more hydrogen nuclei crash together, merge into a heavier atom, and release a ton of energy in the process.

Nuclear fusion milestones: The team at EAST built a giant metal torus (similar in shape to a giant donut) with a series of magnetic coils. The coils hold hot plasma where the reactions occur. They've reached many milestones along the way.

According to New Atlas, in 2016, the scientists at EAST could heat hydrogen plasma to roughly 50 million degrees C for 102 seconds. Two years later, they reached 100 million degrees for 10 seconds.

The temperatures are impressive, but the short reaction times, and lack of pressure are another obstacle. Fusion is simple for the sun, because stars are massive and gravity provides even pressure all over the surface. The pressure squeezes hydrogen gas in the sun's core so immensely that several nuclei combine to form one atom, releasing energy.

But on Earth, we have to supply all of the pressure to keep the reaction going, and it has to be perfectly even. It's hard to do this for any length of time, and it uses a ton of energy. So the reactions usually fizzle out in minutes or seconds.

Still, the latest record of 120 million degrees and 101 seconds is one more step toward sustaining longer and hotter reactions.

Why does this matter? No one denies that humankind needs a clean, unlimited source of energy.

We all recognize that oil and gas are limited resources. But even wind and solar power --- renewable energies --- are fundamentally limited. They are dependent upon a breezy day or a cloudless sky, which we can't always count on.

Nuclear fusion is clean, safe, and environmentally sustainable --- its fuel is a nearly limitless resource since it is simply hydrogen (which can be easily made from water).

With each new milestone, we are creeping closer and closer to a breakthrough for unlimited, clean energy.

The science of sex, love, attraction, and obsession

The symbol for love is the heart, but the brain may be more accurate.

Videos
  • How love makes us feel can only be defined on an individual basis, but what it does to the body, specifically the brain, is now less abstract thanks to science.
  • One of the problems with early-stage attraction, according to anthropologist Helen Fisher, is that it activates parts of the brain that are linked to drive, craving, obsession, and motivation, while other regions that deal with decision-making shut down.
  • Dr. Fisher, professor Ted Fischer, and psychiatrist Gail Saltz explain the different types of love, explore the neuroscience of love and attraction, and share tips for sustaining relationships that are healthy and mutually beneficial.

Sex & Relationships

There never was a male fertility crisis

A new study suggests that reports of the impending infertility of the human male are greatly exaggerated.

Quantcast