Scientists have created artificial intelligence in a test tube using DNA molecules, and they are confident that it will soon begin to form its own "memories."
An artificial neural network made entirely from DNA and mimicking the way the brain works, was created by scientists in the laboratory.
Test-tube AI could solve the classic machine learning problem by correctly identifying handwritten numbers.
The work is a significant step in demonstrating the ability to program AI into artificial organic circuits, scientists say.
This could one day lead to humanoid robots made from completely organic materials, rather than the shiny metal cybermen popular in show culture.
The researchers are confident that the device will soon begin to form its own “memories” from the samples added to the test tube.
Their ultimate goal is to program intelligent behaviors such as the ability to compute, make choices and more, using artificial neural networks made from DNA.
The artist's drawing is an artificial neural network that is created from DNA.
Promotional video:
The California Institute of Technology chose a problem that is a classic challenge for solving the problem of electronic artificial neural networks that recognize handwritten text.
It was one of the first problems solved by computer vision researchers, and an ideal method for illustrating the capabilities of DNA-based neural networks.
A person's handwriting can vary widely, and therefore, when a person studies a written sequence of numbers, the brain performs complex computational tasks to identify them.
Since it is difficult even for humans to recognize each other's sloppy handwriting, identifying handwritten numbers is a common test for programming intelligence in AI neural networks.
These networks must be “trained” to recognize numbers, account for differences in handwriting, and then compare the unknown number with their so-called memories and determine the identification of the number.
The team demonstrated that a neural network of elaborate DNA sequences can carry out chemical reactions indicating that it correctly identified "molecular handwriting."
When an unknown number is given, this so-called "smart soup" undergoes a series of reactions and emits two fluorescent signals, for example, green and yellow to represent five, or green and red to represent nine.
WHY RESEARCHERS HAVE USED DNA FOR
CREATING AI IN A TUBE?
The key to creating bio-molecular strands from DNA is the strict rules for the binding between DNA molecules.
A single-stranded DNA molecule is made up of smaller molecules called nucleotides - abbreviated A, T, C, and G - located in a string or sequence.
Nucleotides in a single-stranded DNA molecule can bind to nucleotides on another single-stranded strand to form double-stranded DNA, but nucleotides only bind in very specific ways.
Nucleotide A always binds to T, and C to G.
Using these predictable binding rules, the researchers were able to design short strands of DNA to undergo predictable chemical reactions in a test tube and thus compute tasks such as recognizing molecular structures.
In 2011, they created the first artificial neural network of DNA molecules that could recognize four simple patterns.
In July 2018, they unveiled in vitro artificial intelligence that can solve the classic machine learning problem by correctly identifying handwritten numbers.
Lead researcher Lulu Qian, associate professor in the Department of Bioengineering, said: “Although scientists have just begun to investigate the creation of artificial intelligence in molecular machines, its potential is already undeniable.
Just as electronic computers and smartphones made humans more capable than a hundred years ago, artificial molecular machines will be able to make anything made from molecules - including even paints and bandages - and become more capable and more responsive to the environment in the next hundred years.."
HOW DOES ARTIFICIAL INTELLIGENCE LEARN?
AI systems rely on artificial neural networks (ANNs) that try to mimic the way the brain works in order to learn.
ANNs will learn to recognize patterns in information, including speech, textual data or visual imagery, and are the basis for a large number of AI developments in recent years.
Conventional AI uses inputs to train an algorithm about a particular subject, feeding it a huge amount of information.
Practical applications include Google language translation services, Facebook face recognition software, and Snapchat image editing filters.
The process of entering this data can be extremely time consuming and limited to one type of knowledge.
A new generation of ANNs, called Adversarial Neural Networks, pits the wits of two AI bots against each other, allowing them to learn from each other.
This approach aims to speed up the learning process as well as improve the inferences generated by AI systems.
