Hot topic: The fourth person who has been “cured” of AIDS may have appeared

"Hotspot Review | Major Technology Events of the Week" is a new column with pictures and texts launched by Academic Headlines, which aims to help readers quickly understand the hot technology news of the past week. Due to the limited space (and knowledge), everyone is welcome to add more~

Hi, my lovely “ academic bacteria ”, I am your lovely Academic King .

Last week was another "last week" that has passed.

Last week, the fourth AIDS patient in the world was declared "cured". This is really exciting news. We look forward to the fifth, sixth... In the AI ​​circle, DeepMind has made another big move. They used AlphaFold to predict almost all protein structures on the earth. Oh my God, so awesome!

But... Besides these, what other hot news happened last week? Let's review them together:

When a dog wags his tail, does he wag it to the left or to the right? Scientists: It depends on the person

Tail wagging in domestic dogs is an important social behavior in interactions between dogs and dogs and humans.

Recently, the team led by Zhang Yongqing, a researcher at the Institute of Genetics and Developmental Biology of the Chinese Academy of Sciences, together with Yu Shan, a researcher at the Institute of Automation, and Wei Pengfei, a researcher at the Shenzhen Institute of Advanced Technology, conducted a study on the behavior of dogs wagging their tails. Based on deep learning motion tracking technology, they extracted and analyzed the movement trajectory of the dog's tail during the interaction between the dog and the human.

The study showed that similar to the characteristic fingerprints and gait of individual humans, each domestic dog exhibited a unique and stable tail wagging feature. During the three-day interaction between domestic dogs and people, the tail wagging gradually shifted from the left side to the right side, indicating that the lateralization of domestic dogs' tail wagging is a sensitive indicator of social familiarity. When interacting with unfamiliar people, the dog's tail wags tended to the left, while when interacting with familiar people, the tail wags tended to the right.

In addition, the study found for the first time that the tail wagging of domestic dogs is composed of an attractor state with a stable wagging trajectory and an unstable transition state. The study revealed the complex characteristics and composition of the tail wagging behavior of domestic dogs during their interactions with humans, and provided a unique experimental paradigm for exploring the social behavior of domestic dogs and the underlying neural mechanisms.

Source: Chinese Academy of Sciences

Paper link:

https://www.sciencedirect.com/science/article/pii/S2589004222010197

The fourth person who was “cured” of AIDS may have appeared

According to foreign media reports, a 66-year-old leukemia patient has achieved long-term remission of AIDS after receiving a stem cell transplant, and may have been "cured." This is the fourth AIDS patient in the world to be declared "cured" and the oldest patient so far.

In addition to being the oldest, this patient has also been infected with HIV the longest. He was diagnosed with AIDS in 1988. For 30 years, he has been taking ART to control his disease, but ART has not been effective in treating AIDS.

Today, the patient has not received antiretroviral treatment (ART) for 17 months, but still has no signs of HIV replication in his body.

Sharon Lewin, president-elect of the International AIDS Society, said the case provided "continued hope and inspiration" to people living with HIV and the wider scientific community, although it was unlikely to be an option for most people living with HIV because of the risks of the surgery.

Source: Science and Technology Daily

How did the Earth survive? Scientists have found new clues

The Earth is the only celestial body in the universe known to mankind to have life, and it can be regarded as one of the greatest wonders in the universe.

However, hundreds of millions of years ago, the Earth experienced a potentially devastating disaster - the magnetic field almost disappeared, leaving only a pitiful 10%.

You know, one of the important reasons why there is no life on Mars is that the magnetic field of Mars has "really" disappeared and cannot be protected from the direct impact of the solar wind.

Earth's magnetic field mysteriously began to "bounce" 550 million years ago, allowing the planet to avoid a fate similar to that of Mars, according to a study by University of Rochester scientists and collaborators.

Figure | There is no core; about 550 million years ago, the inner core began to grow; about 450 million years ago, the boundary between the outermost and innermost cores was formed.

This "bounce" occurred over tens of millions of years -- very fast on geological timescales -- and coincides with the formation of Earth's solid inner core, suggesting that the inner core could have been a direct cause.

Paper link: https://www.nature.com/articles/s41467-022-31677-7

Question: How do you say "Thank you for listening to me" in ancient poetry?

"We have never met in this world, but I will always remember the kindness we have received." This line of poetry clearly reflects the speaker's intention to express gratitude, and can be regarded as the ultimate upper-class version of "Listen to me say thank you."

It comes from WantQuotes, a magical tool for "looking up sentences by meaning" launched by Tsinghua University.

It is understood that the reason why WantQuotes can quickly find the sentences you want from a large number of corpora around the world is mainly due to an extremely large data set and a recommendation model.

The dataset consists of three parts: English, modern Chinese and classical Chinese. The English part contains 6108 famous sayings and 126713 context text segments; the modern Chinese part contains 3004 famous sayings and 408433 related contexts; the classical Chinese part contains 4438 classical Chinese allusions (including ancient poems) and 116537 related contexts.

In terms of recommendation models, WantQuotes uses BERT as a sentence encoder to learn famous quotes and related text representations, and introduces semantics into the encoder.

It is understood that WantQuotes was born from the Natural Language Processing and Social Humanities Computing Laboratory (THUNLP) of the Department of Computer Science at Tsinghua University, and the project instructors are Professor Sun Maosong and Associate Professor Liu Zhiyuan.

Content source: Quantum Bit

"WantQuotes" official website: https://wantquotes.net/

Science: 48-hour non-stop monitoring, wearable ultrasound imaging patch is here

Ultrasound imaging is a safe, non-invasive way to examine the body, providing clinicians with real-time images of a patient's internal organs. However, ultrasound imaging relies on bulky, specialized equipment that is generally only available in hospitals.

A research result from MIT is expected to break this situation. They have developed a wearable ultrasound patch that is about the size of a postage stamp but can be attached to the skin and provide imaging for 48 hours continuously.

The researchers attached the patches to volunteers and showed them high-resolution, real-time images produced by the devices of blood vessels and deep-lying organs such as the heart, lungs, and stomach.

The patches maintained strong adhesion and captured changes in the organs as the volunteers performed a variety of activities, including sitting, standing, jogging and cycling.

Source: Synced

Paper link: https://www.science.org/doi/10.1126/science.abo2542

ICML2022 awards announced: 15 outstanding papers, Fudan University, Xiamen University, and Shanghai Jiaotong University selected

Recently, the ICML 2022 conference was held in Baltimore, Maryland, USA in a combination of online and offline methods, and all awards were announced, including 15 outstanding paper awards and 1 test of time award.

Among them, the research of several Chinese teams including Fudan University, Shanghai Jiaotong University, Xiamen University, and Rice University Hu Xia team won the Outstanding Paper Award. The ICML 2012 paper on "Poisoning Attacks against Support Vector Machines" won the Time Test Award of this conference.

Source: Synced

ICML2022 official website:
https://icml.cc/virtual/2022/awards_detail

Two drops of water can generate electricity for an hour. How do scientists do it?

Using more environmentally friendly materials and improving resource recycling rates are important methods to solve the problem of the proliferation of electronic waste, and biodegradable batteries are a hot topic in current battery research.

Recently, a research team from the Swiss Federal Laboratories for Materials Science and Technology (Empa) has made a new breakthrough in biodegradable batteries. In a proof-of-concept study, they proposed a disposable paper battery that can be activated by water.

It is reported that with just two drops of water, this new type of battery can continuously power an alarm clock with LED for 1 hour. After 1 hour, add two more drops of water and it can continue to be used . It can also be made into any shape and size, and its potential application scenarios are very broad.

Figure | Two-unit paper battery is driving an alarm clock with LED. (Source: Research team)

The research team said that this paper battery can be used to power a variety of low-power, disposable electronic devices (such as smart tags for tracking items, environmental sensors and medical diagnostic equipment) and minimize their environmental impact.

Paper link:

https://www.nature.com/articles/s41598-022-15900-5

Former Google CEO: AI is like nuclear weapons

In an interview, former Google CEO Eric Schmidt compared artificial intelligence to nuclear weapons and called for the establishment of a similar Mutually Assured Destruction deterrence mechanism to prevent the world's most powerful countries from destroying each other because of AI.

Mutually assured destruction is a "mutual destruction" concept. If one of the two opposing parties uses it fully, both will be destroyed. This is called the "balance of terror."

Schmidt compared AI to nuclear weapons, saying that in the future major powers might have to conclude a treaty similar to the ban on "nuclear testing" to prevent AI from destroying the world.

Source: Big Data Digest

A Nature journal revealed that multiple papers had fake author identities.

Dental experts publish papers in the field of mechanical engineering, experts in the fields of medicine, economics, and finance publish papers related to fluid mechanics... Can normal scholars really conduct cross-disciplinary research in this way?

Recently, the well-known open access journal Scientific Reports expressed "editorial concerns" about two papers in the journal. The main reason was the identity issue of the authors.

A reporter from China Science Daily found that most of the authors of the two articles were from Indonesia, Iran and Russia, and one of the articles even had a Chinese author.

Anna Abalkina, a sociologist at the Free University of Berlin in Germany, mentioned in a previously published study that there is a brokerage company in Russia that specializes in selling paper authorships. The authorship is priced differently depending on the position of the authorship and the impact factor of the journal. The fees range from 1,700 yuan to nearly 50,000 yuan. The first author position is usually more expensive, starting at 3,000 yuan.

It is reported that the scope of this type of transnational business has been expanding in recent years, and related papers have penetrated into the journals of many large academic publishing institutions.

Source: ScienceNet

AlphaFold has made another major leap! It predicts almost all protein structures on Earth

Recently, a collaborative team of DeepMind and the European Bioinformatics Institute (EMBL-EBI) announced a major leap in the field of biology. They used the artificial intelligence (AI) system AlphaFold to predict 214 million protein structures of more than 1 million species, covering almost all known proteins on Earth. This breakthrough will accelerate the development of new drugs and bring a new revolution to basic science.

Regarding the release of new data, the research team stated that among more than 200 million protein structure predictions, approximately 35% of the structures have high precision, reaching the structural accuracy obtained by experimental means; 80% of the structures are reliable enough for multiple subsequent analyses.

Content source: Academic Journal

Original link:

https://www.nature.com/articles/d41586-022-02083-2

https://www.deepmind.com/blog

New progress on the cause of the fifth mass extinction

The fifth mass extinction event of the Phanerozoic Eon, the end-Cretaceous (about 66 million years ago), fundamentally reshaped Earth's ecosystems, ending the more than 150 million-year era of dinosaurs and paving the way for the rise of mammals.

At present, the cause of the mass extinction is still under debate. Among them, the two mainstream views are the Deccan basalt eruption in India and the asteroid impact. The timing of the Deccan basalt eruption and its impact on the global environment are the current research hotspots.

Recently, Associate Researcher Li Sha, Researcher Wang Bo, Researcher Zhang Haichun, Researcher Wang Qifei from Nanjing Institute of Geology and Paleontology, Chinese Academy of Sciences, in collaboration with Professor Chen Jiubin from Tianjin University and Professor Wan Xiaoqiao from China University of Geosciences (Beijing), used core data from two wells in Shandong Province that span the Cretaceous/Paleogene (K/Pg) boundary to study the record of the first eruption of the Deccan volcano in East Asia and the response of the lake environment and biota to this event based on high-precision biostratigraphy, paleoecology and geochemistry methods.

The results provide strong evidence for the coupling relationship between the late Maastrichtian warming event and the first eruption of the Deccan volcanoes, and show that the warming event had a significant impact on the lake environment and biota.

Source: Chinese Academy of Sciences

Paper link:

https://pubs.geoscienceworld.org/gsa/geology/article-abstract/doi/10.1130/G50016.1/615407/Mercury-evidence-of-Deccan-volcanism-driving-the

AI Passes the Turing Test, But Scientists Say: Maybe It's Not Necessary

Can machines think? Can artificial intelligence (AI) be as smart as humans? A study suggests that it may be able to do so.

In a non-verbal Turing test, a research team led by Professor Agnieszka Wykowska of the Italian Institute of Technology found that behavioral variability can blur the distinction between humans and machines, that is, it can help robots look more like humans.

Figure | Participants were asked to judge whether the robot's behavior was pre-programmed or controlled by humans. (Source: The paper)

However, in response to this study, Tom Ziemke, professor of cognitive systems at Linköping University, and Sam Thellman, a postdoctoral researcher, believe that human similarity is not necessarily the ideal goal for the development of artificial intelligence and robotics. Making artificial intelligence less like humans may be a wiser approach, and they used the examples of self-driving cars and chatbots to illustrate their point.

Paper link:

https://www.science.org/doi/10.1126/scirobotics.abo1241

https://www.science.org/doi/10.1126/scirobotics.add0641

Yu Longfei's team at Shenzhen International Graduate School has made new progress in global nitrous oxide emissions

Nitrous oxide is an important greenhouse gas with a global warming potential nearly 300 times that of carbon dioxide (on a 100-year scale). Soil is the main source of nitrous oxide emissions, and with additional nitrogen input from anthropogenic activities, terrestrial nitrous oxide emissions are increasing.

In recent years, the concentration of nitrous oxide in the atmosphere has shown an accelerating upward trend. Model inversion results show that the main reason may be the increase in the global nitrous oxide emission coefficient (flux/nitrogen input). However, the changing pattern and driving factors of the nitrous oxide emission coefficient at the global scale are still unclear.

In view of the above problems, the research group of Yu Longfei from Tsinghua University Shenzhen International Graduate School and the research group of Eliza Harris from Swiss Federal Institute of Technology in Zurich have worked closely together to develop a soil-atmosphere nitrogen isotope coupling process model to assess the spatiotemporal distribution of nitrous oxide emissions on a global scale.

Model results show that the global average nitrous oxide emission coefficient (weighted average of nitrogen input) in 2020 is as high as 4.3%, which is significantly higher than the default assessment coefficient of 1.4% by the Intergovernmental Panel on Climate Change (IPCC). At the same time, the global distribution of nitrous oxide emission coefficients is very uneven, with the highest emission coefficients in warm and humid regions.

Over the past century, the global nitrous oxide emission coefficient has increased significantly, with the main driving factors being climate warming and the redistribution of nitrogen fertilizer input (from temperate to tropical zones and from developed to developing countries). The research team recommends that in future nitrous oxide emission forecasts and the formulation of emission reduction policies, great attention should be paid to the feedback effect of climate change and the contribution of regional emission hotspots.

Source: Tsinghua University

Paper link:
https://www.nature.com/articles/s41467-022-32001-z

Scientists achieve quantum simulation of topological time crystals in superconducting system for the first time

Recently, the research group of Assistant Professor Deng Dongling from the Institute of Cross-Disciplinary Information Sciences at Tsinghua University, in collaboration with the research groups of Wang Zhen and Wang Haohua from the School of Physics at Zhejiang University, experimentally realized the first fully digital quantum simulation of topological time crystals in a superconducting system.

Time crystals extend the characteristics of "crystals" to the time dimension, that is, certain characteristics of the system repeat periodically in time, destroying the translational symmetry of time. However, due to the special topological properties of topological time crystals, the breaking of time translation symmetry only occurs at the system boundary.

Realizing topological time crystals is a very challenging problem. Deng Dongling's research group proposed a theoretical model of topological time crystals and found an optimization solution for digital quantum simulation of topological time crystals through artificial intelligence algorithms. The research group cooperated with the superconducting quantum computing team of Zhejiang University and successfully observed this novel phenomenon in experiments.

The experiment simulated a one-dimensional chain of 26 superconducting quantum bits, and observed a dynamic phenomenon in which the time translation symmetry was only destroyed at the system boundary (the two ends of the chain) during the evolution of about 240 layers of quantum circuits. This is the core feature of topological time crystals that is different from conventional time crystals reported in the past. The combination of topology and time crystals has constructed a new non-equilibrium phase of matter, enriched the types of time crystals, and expanded people's understanding of the quantum world. The superconducting quantum chip used in the experiment adopts an easily scalable nearest neighbor connectivity architecture with a high degree of programming flexibility, which can be used to explore more exotic quantum phenomena.

This research result has realized a new form of non-equilibrium matter, innovatively introduced the concept of topology into time crystals, enriched the types of time crystals, and demonstrated the great potential of using medium-scale quantum chips to explore exotic quantum states.

Source: Tsinghua University

Paper link:
https://doi.org/10.1038/s41586-022-04854-3

Scientists make important progress in content-addressable memory

Content addressable memory (CAM) is a memory that is addressed by content. It compares the input data with all the data stored in the array at the same time and outputs the corresponding matching information. It can perform search and similarity detection operations inside the memory with high parallelism and high energy efficiency, reducing data transmission. In recent years, it has shown great potential in completing tasks such as edge intelligent computing.

In addition to data storage, traditional SRAM-based CAMs also require additional hardware overhead to implement data comparison operations. In order to reduce the hardware cost required to perform CAM operations, the international community has successively explored CAM implementation solutions that use new non-volatile memories (such as resistive memory, phase change memory, ferroelectric transistors, etc.), but all require dual-branch complementary data storage units and comparison circuits to implement the core linear inseparable comparison operations, and further implementation of multi-valued CAM requires additional search circuits, and there are still problems such as large hardware overhead and high computing energy consumption.

To address this key issue, the research team led by Academician Huang Ru and Research Fellow Huang Qianqian from the School of Integrated Circuits/Advanced Innovation Center of Integrated Circuits of Peking University was the first in the world to realize CAM operation with a single branch and a single device. They proposed and experimentally prepared a new ferroelectric fin tunneling field-effect transistor (Fe-FinTFET) with bipolar characteristics, which can simultaneously realize three-state CAM and multi-value CAM operations without the need for additional devices or circuits, greatly reducing hardware overhead and energy consumption.

Based on the designed new content-addressable memory, the research team further demonstrated its potential in implementing small sample learning tasks such as Hamming distance and Manhattan distance calculations, significantly improving area efficiency and energy efficiency (compared to traditional GPU-based computing methods, the search speed is increased by 3 orders of magnitude and the search energy is reduced by 2 orders of magnitude), laying an important foundation for highly integrated and energy-efficient edge machine learning computing chips.

Source: Peking University

Paper link:
https://ieeexplore.ieee.org/document/9830413