 |
|
|
|||||
Outdoor lighting for buildings, roads and advertising can help people see in the dark of night, but many astronomers are growing increasingly concerned that these lights could be blinding us to the rest of the universe. An estimate from 2023 showed that the rate of human-produced light is increasing in the night sky by as much as 10% per year. Im an astronomer who has chaired a standing commission on astronomical site protection for the International Astronomical Union-sponsored working groups studying ground-based light pollution. My work with these groups has centered around the idea that lights from human activities are now affecting astronomical observatories on what used to be distant mountaintops. Map of North Americas artificial sky brightness, as a ratio to the natural sky brightness [Image: Falchi et al., Science Advances (2016), CC BY-NC] Hot science in the cold, dark night While orbiting telescopes like the Hubble Space Telescope or the James Webb Space Telescope give researchers a unique view of the cosmosparticularly because they can see light blocked by the Earths atmosphereground-based telescopes also continue to drive cutting-edge discovery. Telescopes on the ground capture light with gigantic and precise focusing mirrors that can be 20 to 35 feet wide. Moving all astronomical observations to space to escape light pollution would not be possible, because space missions have a much greater cost and so many large ground-based telescopes are already in operation or under construction. Around the world, there are 17 ground-based telescopes with primary mirrors as big or bigger than Webbs 20-foot mirror, and three more under construction with mirrors planned to span 80 to 130 feet. The newest telescope starting its scientific mission right now, the Vera Rubin Observatory in Chile, has a mirror with a 28-foot diameter and a 3-gigapixel camera. One of its missions is to map the distribution of dark matter in the universe. To do that, it will collect a sample of 2.6 billion galaxies. The typical galaxy in that sample is 100 times fainter than the natural glow in the nighttime air in the Earths atmosphere, so this Rubin Observatory program depends on near-total natural darkness. The more light pollution there is, the fewer stars a person can see when looking at the same part of the night sky. The image on the left depicts the constellation Orion in a dark sky; the image on the right is taken near the city of Orem, Utah, a city of about 100,000 people. [Photo: jpstanley/Flickr, CC BY] Any light scattered at nightroad lighting, building illumination, billboardswould add glare and noise to the scene, greatly reducing the number of galaxies Rubin can reliably measure in the same time, or greatly increasing the total exposure time required to get the same result. The LED revolution Astronomers care specifically about artificial light in the blue-green range of the electromagnetic spectrum, as that used to be the darkest part of the night sky. A decade ago, the most common outdoor lighting was from sodium vapor discharge lamps. They produced an orange-pink glow, which meant that they put out very little blue and green light. Even observatories relatively close to growing urban areas had skies that were naturally dark in the blue and green part of the spectrum, enabling all kinds of new observations. Then came the solid-state LED lighting revolution. Those lights put out a broad rainbow of color with very high efficiency, meaning they produce lots of light per watt of electricity. The earliest versions of LEDs put out a large fraction of their energy in the blue and green, but advancing technology now gets the same efficiency with warmer lights that have much less blue and green. Nevertheless, the formerly pristine darkness of the night sky now has much more light, particularly in the blue and green, from LEDs in cities and towns, lighting roads, public spaces, and advertising. The broad output of color from LEDs affects the whole spectrum, from ultraviolet through deep red. The U.S. Department of Energy commissioned a study in 2019 which predicted that the higher energy efficiency of LEDs would mean that the amount of power used for lights at night would go down, with the amount of light emitted staying roughly the same. But satellites looking down at the Earth reveal that just isnt the case. The amount of light is going steadily up, meaning that cities and businesses were willing to keep their electricity bills about the same as energy efficiency improved, and just get more light. Natural darkness in retreat As human activity spreads out over time, many of the remote areas that host observatories are becoming less remote. Light domes from large urban areas slightly brighten the dark sky at mountaintop observatories up to 200 miles away. When these urban areas are adjacent to an observatory, the addition to the skyglow is much stronger, making detection of the faintest galaxies and stars that much harder. The Mount Wilson Observatory in the Angeles National Forest may look remote, but urban sprawl from Los Angeles means that it is much closer to dense human activity today than it was when it was established in 1904. [Photo: USDA/USFS, CC BY] When the Mount Wilson Observatory was constructed in the Angeles National Forest near Pasadena, California, in the early 1900s, it was a very dark site, considerably far from the 500,000 people living in Greater Los Angeles. Today, 18.6 million people live in the L.A. area, and urban sprawl has brought civilization much closer to Mount Wilson. When Kitt Peak National Observatory was first under construction in the late 1950s, it was far from metro Tucson, Arizona, with its population of 230,000. Today, that area houses 1 million people, and Kitt Peak faces much more light pollution. Even telescopes in darker, more secluded regionslike northern Chile or western Texasexperience light pollution from industrial activities like open-pit mining or oil and gas facilities. European Southern Observatorys Very Large Telescope at the Paranal site in the sparsely populated Atacama Desert in northern Chile [Photo: J.L. Dauvergne & G. Hüdepohl/ESO, CC BY-ND] The case of the European Southern Observatory An interesting modern challenge is facing the European Southern Observatory, which operates four of the worlds largest optical telescopes. Their site in northern Chile is very remote, and it is nominally covered by strict national regulations protecting the dark sky. AES Chile, an energy provider with strong U.S. investor backing, announced a plan in December 2024 for the development of a large industrial plant and transport hub close to the observatory. The plant would produce liquid hydrogen and ammonia for green energy. Even though formally compliant with the national lighting norm, the fully built operation could scatter enough artificial light into the night sky to turn the current observatorys pristine darkness into a state similar to some of the legacy observatories now near large urban areas. The location of AES Chiles planned project in relation to the European Southern Observatorys telescope sites [Image: European Southern Observatory, CC BY-ND] This light pollution could mean the facility wont have the same ability to detect and measure the faintest galaxies and stars. Light pollution doesnt only affect observatories. Today, around 80% of the worlds population cannot see the Milky Way at night. Some Asian cities are so bright that the eyes of people walking outdoors cannot become visually dark-adapted. In 2009, the International Astronomical Union declared that there is a universal right to starlight. The dark night sky belongs to all peopleits awe-inspiring beauty is something that you dont have to be an astronomer to appreciate. Richard Green is an astronomer emeritus at Steward Observatory at the University of Arizona. This article is republished from The Conversation under a Creative Commons license. Read the original article.
Category:
E-Commerce
New technologies usually follow the technology adoption life cycle. Innovators and early adopters rush to embrace new technologies, while laggards and skeptics jump in much later. At first glance, it looks like artificial intelligence is following the same pattern, but a new crop of studies suggests that AI might follow a different course, one with significant implications for business, education, and society. This general phenomenon has often been described as AI hesitancy or AI reluctance. The typical adoption curve assumes a person who is hesitant or reluctant to embrace a technology will eventually do so anyway. This pattern has repeated over and over. Why would AI be any different? Emerging research on the reasons behind AI hesitancy, however, suggests there are different dynamics at play that might alter the traditional adoption cycle. For example, a recent study found that while some causes of this hesitation closely mirror those regarding previous technologies, others are unique to AI. In many ways, as someone who closely watches the spread of AI, there may be a better analogy: veganism. AI veganism The idea of an AI vegan is someone who abstains from using AI, the same way a vegan is someone who abstains from eating products derived from animals. Generally, the reasons people choose veganism do not fade automatically over time. They might be reasons that can be addressed, but theyre not just about getting more comfortable eating animals and animal products. Thats why the analogy in the case of AI is appealing. Unlike many other technologies, its important not to assume that skeptics and laggards will eventually become adopters. Many of those refusing to embrace AI actually fit the traditional archetype of an early adopter. The study on AI hesitation focused on college students who are often among the first demographics to adopt new technologies. There is some historical precedent for this analogy. Under the hood, AI is just a set of algorithms. Algorithmic aversion is a well-known phenomenon where humans are biased against algorithmic decision-makingeven if it is shown to be more effective. For example, people prefer dating advice from humans over advice from algorithms, even when the algorithms perform better. But the analogy to veganism applies in other ways, providing insights into what to expect in the future. In fact, studies show that three of the main reasons people choose veganism each have a parallel in AI avoidance. Ethical concerns One motivation for veganism is concern over the ethical sourcing of animal by-products. Similarly, studies have found that when users are aware that many content creators did not knowingly opt into letting their work be used to train AI, they are more likely to avoid using AI. Many vegans have ethical concerns about the treatment of animals. Some people who avoid using AI have ethical concerns about the treatment of content creators. [Photo: Vuk Valcic/SOPA Images/LightRocket via Getty Images] These concerns were at the center of the Writers Guild of America and Screen Actors Guild-American Federation of Television and Radio Artists strikes in 2023, where the two unions argued for legal protections against companies using creatives works to train AI without consent or compensation. While some creators may be protected by such trade agreements, lots of models are instead trained on the work of amateur, independent, or freelance creators without these systematic protections. Environmental concerns A second motivation for veganism is concern over the environmental impacts of intensive animal agriculture, from deforestation to methane production. Research has shown that the computing resources needed to support AI are growing exponentially, dramatically increasing demand for electricity and water, and that efficiency improvements are unlikely to lower the overall power usage due to a rebound effect, which is when efficiency gains spur new technologies that consume more energy. Both AI and meat production spark concerns about environmental impact. [Photo: Kichul Shin/NurPhoto via Getty Images] One preliminary study found that increasing users awareness of the power demands of AI can affect how they use these systems. Another survey found that concern about ater usage to cool AI systems was a factor in students refusal to use the technology at Cambridge University. Personal wellness A third motivation for veganism is concern for possible negative health effects of eating animals and animal products. A potential parallel concern could be at work in AI veganism. A Microsoft Research study found that people who were more confident in using generative AI showed diminished critical thinking. The 2025 Cambridge University survey found some students avoiding AI out of concern that using it could make them lazy. It is not hard to imagine that the possible negative mental health effects of using AI could drive some AI abstinence in the same way the possible negative physical health effects of an omnivorous diet may drive some to veganism. How society reacts Veganism has led to a dedicated industry catering to that diet. Some restaurants feature vegan entrees. Some manufacturers specialize in vegan foods. Could it be the case that some companies will try to use the absence of AI as a selling point for their products and services? If so, it would be similar to how companies such as DuckDuckGo and the Mozilla Foundation provide alternative search engines and web browsers with enhanced privacy as their main feature. There are few vegans compared to nonvegans in the U.S. Estimates range as high as 4% of the population. But the persistence of veganism has enabled a niche market to serve them. Time will tell if AI veganism takes hold. David Joyner is an associate dean and senior research associate at the College of Computing at the Georgia Institute of Technology. This article is republished from The Conversation under a Creative Commons license. Read the original article.
Category:
E-Commerce
In early March, Volkan Çinar, a chemistry postdoc at MIT, received an email recruiting him to train AI models. Çinar studies carbon-carbon bonds formation in graphene. Given the stiff competition for jobs in academia, Çinar was no longer sure if his dream of working in academia made sense. So he was receptive to the emails pitch. The email came from Handshake, the job search platform which connects 18 million students from 1,600 higher ed institutions to career opportunities, introducing its new MOVE (Model Validation Expert) Fellowship. The new program gives Handshake an entrée into the high end of AI model training, the hot sector thats seen Meta acquire a 49% stake in Scale for more than $14 billion and Surge bootstrap itself to $1 billion in revenue. For talent like Çinar, MOVE offers better money than teaching and comes with AI training. Id never considered working in AI, Çinar says. But given that Im exploring other positions, I thought Id give it a try, even if it meant the risk of paving the road for AI models to take over his field. Paid subscribers will learn: What to expect from the program The Fellowships acceptance rate and pay range How to make yourself competitive for an AI gig A better way to source expert talent for AI labs Handshake, which started in 2014 and is valued at $3.5 billion according to PitchBook, saw an opportunity to expand into AI training last year. Founder Garret Lord said he came up with the idea after talking to a researcher from an AI lab. At the time Handshake had already been approached by major AI labs directly and middlemen to source experts to train AI models. Many of the PhDs and master’s students were saying they were frustrated by the experience as related to getting trained effectively and not getting paid on time, Lord said on the Kleiner Perkins Grit podcast. Lord realized he was sitting on a major asset: a huge talent network. So Handshake created MOVE, which launched on June 10. We wanted to make sure that as the world of work continues to evolve into more of an AI economy, says Christine Cruzvergara, chief education strategy officer at Handshake, that we were doing everything we can to equip our students with the skills and the experiences that they might need to be really competitive in that job market. The MOVE fellowship pairs graduate students and postdocs like Çinar with short-term gigs of about 10 to 20 hours a week with generative AI labs. Fellows go through Handshakes training program, which takes approximately two to five hours, though some projects may require more training. Fellows learn about how large language models work, as well as how to create good prompts. Depending on their level of expertise they can earn between $40 and $130 an hour. Fellows are under strict NDAs when it comes to discussing their actual work, but Rachel Mitchell, a doctoral candidate in education at University of Miami described the general workflow. Mitchell works in pods with other experts to develop prompts or scenarios for the AI models to chew on. After running the prompt through the AI model the pod determines which parts the model successfully answered and which parts it didnt. The pod will discuss what went wrong, and rewrite the prompt and run it again to make sure that the result wasnt created because of an unclear prompt. The feedback goes to the AI lab which then uses it to train and improve the model. I want to make sure that we’re actually developing the knowledge of the AI and testing its limits, says Mitchell, and that it’s not failing due to human error. Unlike data labeling company Surge, which uses an algorithm to match experts to projects, Handshake uses human review. Fellows are matched to the projects based on their résumés, interviews, and the training they do at Handshake. An expert will help evaluate applicants for domain knowledge. The expert typically also ends up as the project lead and can provide feedback on the fellows work as well. Today, the MOVE program has more than 1,000 fellows and works with alumni from any university in America. Democratizing access to information Neither Mitchell nor Çinar feel as if they are cannibalizing jobs in their industry. AI, at its best, is a thought partner, says Mitchell, who hopes to become a professor. My contribution is helping develop a tool that’s reliable enough that someone who is skilled or has some background knowledge can use it effectively. Mitchell sees generative AI as an essential link for helping disseminate her work to the people who will apply it in classroom settings and dont have the time or resources to do a literature search. She points out teachers are overworked and burnt out, and many may be trying to lighten the load by using AI to generate their lesson plans. However, AI often produces flawed results, generating a recommendation with vague learning objectives instead of specific goals that can be measured in students. But if generative AI draws from the latest research on best practices in education, it could generate useful plans that teachers can use. Similarly, Çinar says that in his field, researchers conduct literature searches to make sure the techniques they are testing are research backed. Wading through scientific journals can be arduous, and using AI to help can slice down the work. Çinar doesnt know what his next step will be: research or industry. What he is sure of is that itll probably ask for AI expertise. He hopes his time spent writing prompts and navigating AI models will help him compete in the job market. He also noted that the program has come with other side benefits. First, hes getting to meet and network with other experts and learn from them. Second, he gets to learn how to operate in a corporate setting. Handling these projects, delivering them on time, and doing quality work as a whole package, he says, I think will be very valuable for any job I end up at. A lifeline in the entry-level job drought In a world where AI skills are in high demand, Handshake offers an attractive prospect: getting paid to get hands-on AI experience writing prompts in your field. According to a Brookings Institute report, job postings asking for AI skills have grown from roughly 3,000 in 2010 to over 80,000 this year. Meanwhile, companies like Duolingo and Shopify are mandating that employees use AI on the job. However, 63% of Americans have never used AI at work, and only 24% reported receiving AI training according to an October 2024 survey from Pew. Emory University sees the MOVE fellowship as an opportunity to equip students for the AI uture. Branden Grimmett, associate dean of Emory College and vice provost of career and professional development at Emory University, said the school immediately signed on after a presentation from Handshake in March. In particular, Grimmett and his colleagues appreciated how the fellowship offers training to students in the humanities. Since launching nearly a month ago, 110 Emory students have applied to join the program, which Grimmett says is a high level of interest compared with other job and training offerings. Handshake says approximately 25% of applicants are accepted into the program, even though the program is still so young as of yet no one has reported being able to parlay their MOVE experience into a job. No generalists, experts only Handshake says applicants who want to stand out should be very clear on their subject-matter expertise. Reference advanced research publications, patents, or teaching that demonstrates your knowledge, Cruzvegara notes. She also points out its important to highlight problem-solving and critical thinking skills and experience with designing assessments and rubrics, or tasks that require deep reasoning. In addition, since fellows work in pods together, demonstrating a history of collaboration such as working in research groups or interdisciplinary environments is crucial. Handshake looks for applicants who combine domain expertise with the desire and intellectual acumen to challenge AI models. Lord envisions a future where MOVE directly leads to a job, where fellows have badges on their profiles and there are leaderboards by school. Going forward, Handshake is considering opening the program to young professionals. Cruzvegara points out that AI is here and the choices are to resist and get left behind or figure out how to adapt. I don’t believe that AI alone is going to take your job, but I do think that people who know AI certainly will, Cruzvegara says. We’re doing our best to make sure that as much of the early talent population as possible is actually exposed to and aware of how they can be using AI in their future roles.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||