Shulou(Shulou.com)06/02 Report--

The relationship between biology and technology is very delicate. All advanced intelligence comes from biology, but we still don't understand why caterpillars curl up because of external stimuli. Because of this, we will intentionally or unintentionally imitate the characteristics of living things when we develop technology. Of course, the results are not entirely true. Brain-like computing that imitates the human brain is not going well, but a series of swarm algorithms learned from ants and other organisms have high application value.

Arrange and combine the cells like the electronic original, plug in the wire to "supply energy" to the human body, and seamlessly fit the robotic arm with the body. Biochemical punk can be said to be our ultimate fantasy about technology.

However, on the way to open up the natural world and the technological world, we are still making difficult steps forward. Today in Friends of AI, let's take a look at the biotechnology that AI has been studying and what progress has been made this year.

What did the best teacher in AI teach us in a year?

Biotechnology is a very broadly defined discipline in itself, but what we are focusing on today is the progress that can be used by AI technology.

Generally speaking, how AI uses biotechnology has been quite vague for a long time. As a result, academia has done a lot of useless work, for example, we use a lot of arithmetic to simulate an amoeba protozoa, just to observe how nerve cells transmit signals. But judging from this year's research progress, the uselessness of bioscience to AI is being rewritten.

First of all, we have a deeper understanding and control of biological perception.

We associate brain science, neuroscience and AI because we all face the same ultimate question-how on earth does consciousness come about? Why do people and other animals recognize that you are you and I am me, and why do they know to avoid danger?

In biology, although we still do not understand these ultimate problems, we have taken a step forward in our understanding of biological perception systems.

For example, in the sense of smell, this animal is far better than humans, and researchers at Duke University have found a way to replicate. By extracting olfactory receptors from animal stem cells, the researchers found that 2% of genes in the human genome are used in smell, compared with 5% in dogs and mice. By conquering animal genes, researchers have been able to grow living animal olfactory cells and keep them active in a liquid medium.

In other words, one day in the future, maybe we only need a few drops of liquid to achieve the same sense of smell as "drug detection dog" and "truffle pig".

At the same time, in terms of bionics, our imitation of biological abilities is becoming more and more lifelike.

Although many people believe that bionics is absolutely the wrong direction for AI, it has to be admitted that the development of AI can not do without the imitation of animals. Take the well-known Boston Dynamics, for example, the appearance of a robot dog climbing a hill and running is a complete copy of dogs and insects.

This year, our imitation of animals is not limited to conventional species such as cats and dogs, but has begun to reach out to reptiles.

When they saw a lizard accidentally vacating its upper body and walking on two feet, researchers at the University of Queensland School of Biological Sciences suddenly had an idea and began to study why the animal, which is used to walking on four legs, could suddenly change its posture and move faster while maintaining balance.

The final conclusion is that lizard tails play a great role in changing posture, and adding tails to quadruped robots may also help them cope with richer terrain.

Stanford University is interested in the gecko's hearing system: unlike humans, the heads of animals such as geckos are too small to triangulate the position of noise by the distance between two ears. But the gecko's head has a small tunnel that can be used to determine which direction they come from by reflecting incident sound waves.

Researchers have tried this method of miniature reflection on photodetectors and are trying to create a way to record the angle of light. Tiny sensors of color and polarity. In the future, such sensors are likely to be used in self-driving cars to replace the giant radar above today's cars.

In addition to these, our understanding of other aspects of biology is also being combined with science and technology.

Like the "biological light source" that we have introduced to you, our understanding and replication ability of biological DNA are enabling us to replicate more of the biological characteristics of luminous power generation, which are beneficial to human beings. For example, recently, Pennsylvania State University in the United States in the study of plants, found a protein that closely binds to metals, the use of this protein can help human beings to find more rare earth resources.

In short, nature is the creator of human beings, and human beings are the creators of AI. Bioscience can give AI much more "intergenerational education" than we think.

Open the door of biochemical punk

Since biochemical punk's blueprint is not out of reach, what progress will be made in the integration of AI and biotechnology in 2019?

First of all, the in-depth understanding of biology will make the dynamic mode of the robot more advanced.

All the birds, animals and insects we see today have gone through countless rounds of screening and evolution to become what they are today. Their behavior patterns are highly adapted to nature, and AI wants to be more close to the real application scene, so it naturally has to learn from these "predecessors" in its behavior patterns. In 2018, we have seen many related cases, such as micro-robots imitating the movement of insects to help them jump on the surface of the water. It is believed that with the understanding of the biological motion mode itself, the future robot motion system will adapt to more scenes: pipes, gaps, vertical walls.

Combined with biotechnology, AI will be able to make more use of biology itself in the future.

In many biochemical punk literary works, the end of technological development is to transform life with technology, rather than creating an intelligent body similar to life. In the fantasy of biochemical punk, it is not the omnipotent robot that will serve mankind a few years later, but the monkey that has been transformed by the brain. In fact, in 2018, a paper on simulating "dog brain" attracted people's attention. With our in-depth understanding of biological behavior and thinking patterns, barrier-free communication between humans and animals is likely to become a reality in the future-recently, scientists at the University of Washington have developed a sensor that does not interfere with the daily actions of bees. wear them to help collect more comprehensive agricultural data.

The most significant thing is to expand the range of training data needed by AI through the understanding of other biological perception patterns.

Nowadays, the most widely used image recognition and speech recognition essentially adopt the perception mode that human beings are familiar with. But for those perception methods that human beings are not familiar with, such as stronger sense of smell, richer color perception and so on, they are also not used by Al. For example, a Canadian research institute once tried to study the visual style of fruit flies and established a neural network that mimics the vision of fruit flies and found that this neural network can accurately identify fly species in a very small proportion. In the past, it was impossible for humans to label data.

In fact, from another point of view, when most of the "technology" in our eyes has not yet been born, the greatest wisdom of human beings is to domesticate animals, develop their wisdom and abilities, and establish connections with each other. The development and utilization of the wisdom of animals and the wisdom of building machines from scratch are not two parallel lines that do not intersect.

When the simulation of human intelligence enters the bottleneck, lowering the line of sight and learning to use and simulate the wisdom of animals may be an extraordinary breakthrough.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.