Research results published in the journal,external pageNaturecall_made,show that realistic global forest carbon potential is approximately 226 Gigatonnes(Gt)of carbon.The study,which involved hundreds of scientists around the world,highlights the critical importance of forest conservation,restoration,and sustainable management in moving towards international climate and biodiversity targets.The researchers stress that this potential can be achieved by incentivizing community-?driven efforts to promote biodiversity.The forest carbon potential has been ahighly controversial topic.Four years ago,a study published in the journal Science found that the restoration of forests could capture over 200 Gt of carbon-which could draw down approximately 30 percent of excess anthropogenic carbon.While this study elevated adiscussion about the role of nature in fighting climate change,it also raised concerns around the adverse environmental impacts of mass tree plantations,carbon offsetting schemes,and greenwashing.While some scientific studies have supported the scale of this finding,others argued that this forest carbon estimate could be up to 4or 5times too high.To address this controversial topic an international team of hundreds of researchers led by the Crowther Lab at ETH Zurich joined forces to build an integrated assessment using acomprehensive range of approaches,including vast ground-?sourced data and satellite datasets.Achieving forest carbon potential Due to ongoing deforestation,the total amount of carbon stored in forests is~328 Gt below its natural state.Of course,much of this land is used for extensive human development including urban and agricultural land.However,outside of those areas,researchers found that forests could capture approximately 226 Gt Cin regions with alow human footprint if they were allowed to recover.Approximately 61 percent of this potential can be achieved by protecting existing forests,so that they can recover to maturity.The remaining 39 percent can be achieved by reconnecting fragmented forest landscapes through sustainable ecosystem management and restoration.“Most of the world’s forests are highly degraded.In fact,many people have never been in one of the few old growth forests that remain on Earth,”said Lidong Mo,a lead author of the study.“To restore global biodiversity,ending deforestation must be atop priority.”The dataset revealed that biodiversity accounts for approximately half of the global forest productivity.As such,the researchers highlighted that,to achieve the full carbon potential,restoration efforts should include anatural diversity of species.In addition,sustainable agricultural,forestry,and restoration practices that promote biodiversity have the greatest potential for carbon capture.Redefining restoration The authors stress that responsible restoration is afundamentally social endeavour.It includes countless actions such as conservation,natural regeneration,rewilding,silviculture,agroforestry,and all other community-?driven efforts to promote biodiversity.It requires equitable development,driven by policies that prioritize the rights of local communities and Indigenous people.“We need to redefine what restoration means to many people,”said Thomas Crowther,the senior author of the paper and aprofessor at ETH Zurich.“Restoration is not about mass tree plantations to offset carbon emissions.Restoration means directing the flow of wealth towards millions of local communities,Indigenous populations,and farmers that promote biodiversity across the globe.Only when healthy biodiversity is the preferred choice for local communities will we get long-?term carbon capture as abiproduct.” 查看详细>>

Despite their huge success,the inner workings of large language models such as OpenAI’s GPT model family and Google Bard remain amystery,even to their developers.Researchers at ETH and Google have uncovered apotential key mechanism behind their ability to learn on-?the-fly and fine-?tune their answers based on interactions with their users.Johannes von Oswald is adoctoral student in the group headed by Angelika Steger,ETH Professor for Theoretical Computer Science,andresearches learning algorithms for neural networks.His new external pagepapercall_made,which has not yet been peer-?reviewed,will be presented at the International Conference on Machine Learning(ICML)in late July.The Tin GPT stands for transformers.What are transformers and why did they become so prevalent in modern AI?Johannes von Oswald:Transformers are aparticular artificial neural network architecture.It is for example used by large language models such as ChatGPT,but was put on the map in 2017 by researchers at Google,where it led to state-?of-the-art performance in language translation.Intriguingly,a slightly modified version of this architecture was already developed by the AI-?Pioneer Jürgen Schmidhuber back in 1991.And what distinguishes this architecture?Before the recent breakthrough of Transformers,different tasks,e.g.image classification and language translation,had used different model architectures that were each specialised on these specific domains.A crucial aspect that sets transformers apart from these previous AI models is that they seem to work extremely well on any kind of task.Because of their widespread use,it is important to understand how they work.What did your research reveal?While neural networks are generally regarded ablack box that spit out output when provided with input,we showed that transformers can learn on their own to implement algorithms within their architecture.We were able to show that they can implement aclassic and powerful machine learning algorithm that learns from the recent information it receives.Can you give an example when this type of learning can occur?You might,for instance,provide the language model with several texts and the sentiment–either positive or negative–associated with each of them.You can go on to present the model with atext it hasn’t seen before,and it will predict whether it is positive or negative based on the examples you have provided.So,you’re saying that the model teaches itself atechnique to learn new things?Yes,it’s surprising but true.Driven simply by the pressure to improve on its training objective,namely to predict the immediate future,it develops atechnique that enables it to learn from the conversations it has with its users,for example.This type of learning is what we refer to as in-?context learning.All these models get is text input.Can you describe how transformers use this minimal information to optimise their output?One way to achieve this–and our paper shows that it’s alikely possibility–is to learn what you might call aworld model that allows you to make predictions.What is interesting is that this learning takes place inside the transformer that has already been trained.Learning would normally involve changing the connections in the model’s neural network.We showed that the transformer model is somehow able to simulate the same learning process within its fixed neural architecture instead. 查看详细>>

The fast switching and modulation of light is at the heart,among other things,of modern data transfer,in which information is sent through fibre optic cables in the shape of modulated light beams.It has been possible for several years now to miniaturise light modulators and to integrate them into chips,but the light sources themselves–light emitting diodes(LEDs)or lasers–still pose problems to engineers.A team of researchers at ETH Zurich led by Prof.Lukas Novotny,together with colleagues at EMPA in Dübendorf and at ICFO in Barcelona,have now found anew mechanism by which tiny but efficient light sources could be produced in the future.The results of their research have recently been published in the scientific journalexternal pageNature Materialscall_made.Trying the unexpected“To achieve this,we first had to try the unexpected”,says Novotny.For several years he and his coworkers have been working on miniature light sources that are based on the tunnel effect.Between two electrodes(made of gold and graphene in this case)separated by an insulating material,electrons can tunnel according to the rules of quantum mechanics.Under particular circumstances–that is,if the tunnel process is inelastic,meaning that the energy of the electrons is not conserved–light can be created.“Unfortunately,the yield of those light sources is rather poor because the radiative emission is very inefficient”,explains postdoc Sotirios Papadopoulos.This emission problem is well-?known in other areas of technology.In mobile phones,for instance,the chips that create the microwaves needed for transmission are only afew millimetres in size.By contrast,the microwaves themselves have awavelength of around 20 centimetres,which makes them ahundred times larger than the chip.To overcome this difference in size an antenna is needed(which,in modern phones,is actually no longer visible from the outside).Likewise,in the experiments of the Zurich researchers the wavelength of the light is much larger than the light source.Semiconductor outside the tunnel junction“One might think,then,that we were consciously looking for an antenna solution–but in reality we weren’t”,says Papadopoulos.Like other groups before them,the researchers were investigating layers of semiconductor materials such as tungsten disulfide with athickness of asingle atom sandwiched between the electrodes of the tunnel junction in order to create light in this way.In principle one would assume that the optimal position should be somewhere between the two electrodes,maybe alittle closer to one than to the other.Instead,the researchers tried something completely different by putting the semiconductor on top of the graphene electrode–completely outside the tunnel junction.Reference Wang,L,Papadopoulos,S,Iyikanat,F,Zhang,J,Huang,J,Taniguchi,T,Watanabe,K,Calame,M,Perrin,ML,García de Abajo,FJ,Novotny,L.Exciton-?assisted electron tunnelling in van der Waals heterostructures.Nat.Mater.(2023).external pagehttps://doi.org/10.1038/s41563-?023-01556-7call_made 查看详细>>

Despite approved treatments being available,multiple myeloma remains incurable.But researchers at ETH Zurich and University Hospital Zurich set out to improve treatment outcomes by testing hundreds of existing therapeutics outside the body to predict their effectiveness.Multiple myeloma is arare blood cancer caused by the uncontrolled multiplication of abnormal plasma cells.These plasma cells are aspecial type of white blood cells that play an important role in the immune system by producing essential antibodies in the bone marrow and lymph nodes.Despite an increasing number of approved drugs and treatment approaches such as immunotherapy becoming available,the disease is still not curable.The average life expectancy of patients after diagnosis is only five years.One of the main challenges is the cancer’s tendency to return even after treatment.This is because treatment makes the cancer cells more resistant to the drugs used,until eventually,after several rounds of treatment,no effective options remain.To address this issue,ETH researchers have adapted their screening platform to look for ways out of this problem,and thereby offering new hope for more successful treatment outcomes.Biopsies under the microscope The researchers are using ahigh-throughput screening method called pharmacoscopy,developed by Professor Berend Snijder,to test the effectiveness of various treatments on the patient’s cancer cells.This state-of-the-art method allows for several hundred different treatment combinations to be tested simultaneously outside the body.By analysing the reactions of the cells to each treatment,they can determine which method holds the most promise for each patient.To do this,the cells from the biopsies are placed in special plates with 384 small wells,each containing adifferent combination of potential treatment substances.After 24 hours,the cells are stained using different antibodies,and their reactions are evaluated using images generated by automated microscopy.A deep-learning algorithm is then used to identify and classify the cell types.The process is largely automated,allowing for efficient and accurate analysis of the results.138 biopsies individually tested The researchers used pharmacoscopy to closely examine 138 bone marrow biopsies from 89 myeloma patients with different stages of multiple myeloma,including newly diagnosed and untreated,as well as patients that underwent multiple treatment rounds.The goal was to observe the behaviour of cancer cells in response to various approved drugs and drug combinations in each biopsy.Based on the cells’appearance,the researchers could determine the best treatment option for each patient.Although the Snijder lab had previously used pharmacoscopy in similar studies on other types of blood cancer,such as lymphomas and leukaemias,the platform had to be adapted for this myeloma study.Hope for more effective treatments The groundbreaking work done by the researchers at ETH and University Hospital Zurich offers hope for more effective treatments for multiple myeloma.By using Pharmacoscopy to test hundreds of treatments,the researchers were able to identify new,more effective treatment options for each patient.This new personalized medicine approach is transferable to the clinic and can thus help doctors find the best option for their patients at an early stage."First,however,we will have to validate the method further in clinical trials,"says Snijder.Now,the Snijder lab wants to develop the platform further to expand its use to solid tumours.Unlike blood cancers,solid tumours must first be dissociated to acertain degree before they can be tested in the 384-well plate format.They are currently adapting the screening platform for brain tumours,among others. 查看详细>>

A team led by ETH Zurich chemical engineers Chih-?Jen Shih and Andrew deMello have developed arapid test system made of smart graphene paper.It only costs afew cents per test strip,is easy to use but is as accurate as lab measurements.The approach will impact more than just disease monitoring.Rapid pregnancy and COVID-?19 tests have agreat advantage over other medical analyses:they are so simple that anyone can perform the test themselves,virtually anywhere.This is due to the robust principle behind these microfluidic methods,whereby aqueous solutions diffuse through apaper test strip with the aid of capillary forces.During this process,antibodies capture the target substances,such as virus particles or pregnancy hormones,and concentrate them at adesired location.A staining system then slowly makes the increasingly concentrated target substance visible as astripe.As simple and reliable as this basic principle is,visual assessment of the results can be difficult.The question of whether we really can see aline or whether we’re just imagining it has probably been on all of our minds at least once since the outbreak of the COVID-?19 pandemic.This is exactly where the invention of the ETH team comes in.They have found away to form conductive electrodes directly inside the test strip paper.As the target substance is captured,it triggers an electronic signal.This makes measurements much faster,more sensitive and more accurate.Making low-?cost paper technology better Chih-??Jen Shih and Andrew deMelloshare apassion:“Our biggest incentive is to improve basic chemical and biological experiments in ways that create new scientific opportunities.”This is exactly what their research groups have now succeeded in doing with the rapid tests.A host of analytical applications benefit from the researchers’combining simple and cost-?effective paper-?based microfluidics with the sensitivity and accuracy of electronic measurement techniques.From patients being able to monitor blood biomarkers themselves,to soil,air and water sampling in the field,to disease testing in remote parts of the world in amatter of minutes–the potential range of applications covers virtually all chemical,biological and medical analyses that can be performed in aqueous solutions.A combination of skills is the key Previous attempts to equip low-?cost paper chemistry with detection electrodes have been hampered by afundamental property of conductive materials:in principle,electrical conductors hardly interact with water and act as abarrier when it comes to the flow of sample and reaction mixtures in apaper strip.Overcoming this hurdle and developing the technology into areliable process that also works in less developed areas of the world called for acombination of skills,deMello says.For the new rapid test,the ETH researchers therefore pooled their expertise.From Shih’s group came the know-?how to generate conductivity directly in the paper,while deMello’s group brought their knowledge of microfluidic systems to the table.A laser decomposes cellulose to pure carbon The basis for this invention is the use of alaser to convert the sugar polymers that make up the cellulose in the paper into graphene.This special form of carbon is conductive and is considered an electronic material of the future.During laser conversion,the cellulose molecules are decomposed into their elements,carbon,oxygen and hydrogen,in aprocess comparable to the caramelisation of household sugar.However,while heating sugar in apan for along time leaves behind only ordinary carbon,which does not conduct electric current,the ETH Zurich scientists were able to use the laser to reorganise the carbon atoms of the cellulose into conductive graphene.Clever tuning of the laser energy makes the difference Producing graphene in cellulose paper alone wouldn’t do the trick,though,because–like virtually all other conductive materials–this miracle material is hydrophobic:it repels water,which means water cannot simply flow through it.However,by cleverly tuning the laser energy,the researchers can control the decomposition of the cellulose to graphene such that not only is the original porosity of the cellulose preserved,but individual oxygen groups of the cellulose remain intact on the surface of the graphene areas. 查看详细>>

For years,ETH Zurich has ranked among the top ten universities in the world according to the QS World University Rankings;in summer 2022,ETH came in at ninth place.Quacquarelli Symonds(QS)has now published the rankings broken down by subject,and ETH performed brilliantly here as well:16 disciplines ranked in the top ten.Three earth sciences departments(Earth&Marine Sciences,Geology and Geophysics)once again took the number one spot.ETH also maintained or improved its rankings in fields where it is traditionally strong:in architecture,for example,it rose to third place.ETH ranked higher in all the engineering sciences than in the previous year–from eighth to fifth place in Engineering–Mechanical,Aeronautical&Manufacturing,for example.And it put in astrong showing once again in Chemistry,moving up three positions to seventh place.In Mathematics,as well as Physics and Astronomy,ETH moved up one place,with both disciplines landing in eighth place.“We can be proud of the fact that ETH so consistently achieves such excellent results across the disciplines.This ultimately reflects well on all of Switzerland and cannot be taken for granted,”says Joël Mesot,President of ETH Zurich.“My appeal goes out to politicians to ensure that the appropriate framework conditions are in place so that we can continue to keep up with the best in the world in the future.”Best university system in the world The QS rankings by discipline not only awarded high marks to ETH Zurich as an institution but also to Switzerland overall as acentre of education:According to QS,Switzerland has the best higher education system in the world.This is because,according to their ranking,Switzerland has the highest concentration of first-?ranked disciplines in the world(four,three of them at ETH Zurich)as well as the third highest concentration of top-?ten disciplines(15%of all programmes in Switzerland). 查看详细>>

People are already using the artificial intelligence(AI)tool ChatGPT to write essays for university,texts for Instagram posts and even parliamentary speeches.What are the next major developments we can expect to see from these large language models?Florian von Wangenheim:Whenever anything new happens in the field of AI,we always feel like we’re on the brink of amajor development.But actually,we’re right in the middle of it.Currently,theAI tools that deal with language are getting alot of attention–speech recognition,automated text creation,even adaptation of texts to specific writing styles,etc.I believe the next step will be personalisation.Bots will communicate with us,using our gestures and facial expressions to try to figure out how we learn best and what messages we are most receptive to.Besides ChatGPT,AI is in the headlines because of TikTok’s‘Bold Glamour‘filter,use of which is nearly impossible to detect.Is 2023 going to be apivotal year for AI in digital communications?This development isn’t new;it’s been going on for several decades now and will continue into the future.What we’re seeing today is the result of years of development work.But it’s safe to say that in 2023,no one will be able to avoid this progress any longer.It’s become clear that many things can no longer be dismissed as unrealistic or far-?off fantasies.How will AI change our digital communication?Will we be using AI filters in Zoom meetings in the future?It’s hard to say.During the coronavirus,for example,my observation was that after experimenting for awhile,users kept the backgrounds in virtual meetings fairly constant.As long as we all feel like we have alot to do,I don’t think anyone wants to be suspected of only being concerned with their appearance while others are hard at work.I find it much more likely that during ameeting,we’ll get real-?time help from tools that work much the same way ChatGPT does,but are better at solving aparticular problem–they might provide arelevant solution or some particularly clever wording that helps me impress my boss.Whether my boss would be happy if Igot help from an AI tool is one question.Here’s another:Will AI steadily reduce the need for people to know things?Einstein said that you don’t need to know things if you know where you can look them up.If we can speak of anew dimension to this phenomenon,it’s certainly in terms of how knowledge has become incredibly widely available:first with the internet and search engines,and now with knowledge being dialogue-?based via AI tools such as ChatGPT. 查看详细>>

ETH spin-??off Versantis is dedicated to the fight against liver diseases.Their lead drug candidate gives hope to arising number of affected patients.Co-?founder Vincent Forster talks about agreat opportunity,which will change their business as well as boost the further development of their drug pipeline.Vincent,you surprised us with apress release:GENFIT acquires Swiss biotech Versantis.Congratulations from ETH Zürich for this great achievement.How does it feel?My team and Iare very proud and at the same time overwhelmed.It needs both luck and perseverance to get where we are.As is often the case in biotech,it felt like we were walking athin line between failure and success,and now the resilience of our team paid out.What happens now?Versantis becomes part of GENFIT:our research,the underlying intellectual property assets,and our team.It’s an excellent fit.Our liver drugs complement GENFIT’s pipeline and our expertise in advanced liver diseases,such as Acute-?on-Chronic Liver Failure,bolsters their scientific and medical troops.GENFIT on the other hand provides decades of experience in chronic liver conditions and solid managerial and clinical development teams.Together we have an ideal environment and key resources to push the development of life-?saving drugs forward.Can you take us back to the beginning of the Versantis story?The idea was born when Iapplied for my first patent.At the time Iwas aPhD student in the group of Prof.Leroux.We invented adialysis method for the treatment of ammonia and drug intoxication.Soon afterwards Iwon aPioneer Fellowship from ETH Zürich.This incubator programme provided me with the means to carry out the proof-?of-concept,develop abusiness plan and prioritise the most relevant applications and market spaces.You have identified liver diseases as your main target.How can Versantis help fighting the disease?Applying our method to liver diseases will hopefully contribute to an urgently needed improvement in patents’lives.Despite the liver playing acentral role in amyriad of vital functions,liver disease has been neglected to the point that it is now one of the fastest-?growing cause of death.Every fourth person today is affected by chronic liver disease,many of them undiagnosed.A sick liver can no longer eliminate the multiple toxins and wastes our body produces,which triggers awhole cascade of health issues leading to multiple organ failure and Acute-?on-Chronic Liver Failure,a severe condition associated with up to 75%mortality within amonth.Getting swiftly rid of the toxins will stop that cascade,allow the body time to activate its own healing process,and reverse the fatal course of the acute event.Our method is based on injecting the drug into the patient’s abdomen where it captures and traps the toxins.It can then be removed from the abdomen with the excess toxins.Thus,the drug achieves its action without relying on the liver-?or kidney metabolic pathway which is defective in these patients.Will we still see you in Zurich’s start-?up ecosystem?I have received alot of valuable support during the early steps of my career.I would be happy to give some of this support back to the next generation of entrepreneurs–as acoach or mentor.Therefore,yes,I will be around. 查看详细>>

If only it were less prone to error,quantum physics might already be giving us instant solutions to seemingly unsolvable problems.ETH researchers are therefore working to develop systems that are more robust.In crude terms,our digitally driven information society is based on asimple binary opposition:0 or 1.But what happens when other alternatives exist alongside these polar opposites?Might this give rise to awhole raft of different states and enable us to process complex information much faster?It is precisely the prospect of going beyond conventional methods of data processing that has inspired such high hopes in the field of quantum physics–not only on the part of scientists in basic and theoretical research,but also among the CEOs of major corporations.Were this vision to materialise,and computers behave in accordance with the laws of quantum mechanics,it would open the door to awhole new world of applications.For example,such apowerful system would be able to determine the mechanism of proteins at aradically faster rate than aconventional computer could ever hope to achieve.This,in turn,would massively accelerate the development of new medicines.A rocky road Given such prospects,it is little wonder that quantum physics should exercise afascination far beyond its immediate circle.Yet the road that will take us to aquantum computer capable of answering everyday questions is arocky one–and much longer than many are prepared to admit.“We’re talking about decades,not years,before we reach that point,”says Jonathan Home,Professor of Experimental Quantum Optics and Photonics at ETH Zurich.And Professor Home is one of those working in afield in which quantum research is relatively far along.He uses individual atoms as qubits.These are the basic units of information used by aquantum computer to perform calculations.Home uses beryllium and calcium atoms held in special electrical ion traps.These are then manipulated with alaser according to the laws of quantum mechanics.“Atoms are great systems for information processing because they can be isolated–and because,provided they remain isolated,they can store quantum information for acouple of seconds or even minutes,”he explains.In order to be able to use this information,however,these fragile quantum objects have to be reconnected with the everyday physical world.During this step,even the slightest anomalies can corrupt the entire system.The question is,therefore,how to reduce this susceptibility to error and,at the same time,increase the number of qubits. 查看详细>>

Roe deer are among the few mammals whose embryos go into aparticularly long period of dormancy.Using modern molecular methods,ETH Zurich researchers have shown for the first time exactly what happens to the embryo during this phase.They have identified signals that control the embryo`s awakening.Everyone is familiar with the roe deer,either from crossword puzzles or from real-?life encounters during ajog or ahike in the forest:majestic creatures with elegant big black eyes.As common as roe deer may seem in Swiss forests,one of their characteristics is unique among deer species.After mating and fertilisation of the egg in midsummer,the pinhead-?sized embryo does not implant in the uterus,but enters into aperiod of dormancy,called embryonic diapause.This period lasts for over four months until December.Only then does the embryo continue its development at normal pace and implants in the uterus.In May,after four and ahalf months of“real”gestation,the doe gives birth to one to three fawns.Although the phenomenon has been known for more than 150 years,it still puzzles.Various forms of embryonic diapause are known to occur in over 130 mammalian species.However,they rarely last as long as observed in the roe deer.And,most importantly,almost no other species shows such apronounced,continuous deceleration instead of acomplete halt.In mice,scientists can artificially induce diapause.However,in roe deer it is still unclear which factors control diapause while keeping the embryo alive.The research group led by Susanne Ulbrich,Professor of Animal Physiology at ETH Zurich,has been investigating the mystery of roe deer diapause for some time.In anew study,the researchers show which molecular processes take place in the embryo while it is dormant:embryonic cells continue to divide during diapause,albeit very slowly.The number of cells,including embryonic stem cells,doubles only every two to three weeks.The study,which has just been published in the journal PNAS,involved not only the ETH group,but also researchers from the Universities of Zurich and Bern,as well as German and French research institutions. 查看详细>>