InvestMacro

By Eve Bohnett, University of Florida 

Snow leopards are known as the “ghosts of the mountains” for a reason. Imagine waiting for months in the harsh, rugged mountains of Asia, hoping to catch even a glimpse of one. These elusive big cats move silently across rocky slopes, their pale coats blending so seamlessly with snow and stone that even the most seasoned biologists seldom spot them in the wild.

Travel writer Peter Matthiessen spent two months in 1973 searching the Tibetan plateau for them and wrote a 300-page book about the effort. He never saw one. Forty years later, Peter’s son Alex retraced his father’s steps – and didn’t see one either.

Researchers have struggled to come up with a figure for the global population. In 2017, the International Union for Conservation of Nature reclassified the snow leopard from endangered to vulnerable, citing estimates of between 2,500 and 10,000 adults in the wild. However, the group also warned that numbers continue to decline in many areas due to habitat loss, poaching and human-wildlife conflict. Those who study these animals want to help protect the species and their habitat – if only we can determine exactly where they live and how many there are.

Traditional tracking methods – searching for footprints, droppings and other signs – have their limits. Instead of waiting for a lucky face-to-face encounter, conservationists from the Wildlife Conservation Society, led by experts including Stéphane Ostrowski and Sorosh Poya Faryabi, began deploying automated camera traps in Afghanistan. These devices snap photos whenever movement is detected, capturing thousands of images over months, all in hopes of obtaining a rare glimpse of a snow leopard.

But capturing images is only half the battle. The next, even harder task is telling one snow leopard apart from another.

At first glance, it might sound simple: Each snow leopard has a unique pattern of black rosettes on its coat, like a fingerprint or a face in a crowd. Yet in practice, identifying individuals by these patterns is slow, subjective and prone to error. Photos may be taken at odd angles, under poor lighting, or with parts of the animal obscured – making matches tricky.

A common mistake happens when photos from different cameras are marked as depicting different animals when they actually show the same individual, inflating population estimates. Worse, camera trap images can get mixed up or misfiled, splitting encounters of one cat across multiple batches and identities.

I am a data analyst working with Wildlife Conservation Society and other partners at Wild Me. My work and others’ has found that even trained experts can misidentify animals, failing to recognize repeat visitors at locations monitored by motion-sensing cameras and counting the same animal more than once. One study found that the snow leopard population was overestimated by more than 30% because of these human errors.

To avoid these pitfalls, researchers follow camera sorting guidelines: At least three clear pattern differences or similarities must be confirmed between two images to declare them the same or different cats. Images too blurry, too dark or taken from difficult angles may have to be discarded. Identification efforts range from easy cases with clear, full-body shots to ambiguous ones needing collaboration and debate. Despite these efforts, variability remains, and more experienced observers tend to be more accurate.

Now people trying to count snow leopards are getting help from artificial intelligence systems, in two ways.

Spotting the spots

Modern AI tools are revolutionizing how we process these large photo libraries. First, AI can rapidly sort through thousands of images, flagging those that contain snow leopards and ignoring irrelevant ones such as those that depict blue sheep, gray-and-white mountain terrain, or shadows.

AI can identify individual snow leopards by analyzing their unique rosette patterns, even when poses or lighting vary. Each snow leopard encounter is compared with a catalog of previously identified photos and assigned a known ID if there is a match, or entered as a new individual if not.

In a recent study, several colleagues and I evaluated two AI algorithms, both separately and in tandem.

The first algorithm, called HotSpotter, identifies individual snow leopards by comparing key visual features such as coat patterns, highlighting distinctive “hot spots” with a yellow marker.

The second is a newer method called pose invariant embeddings, which operates similar to facial recognition technology: It recognizes layers of abstract features in the data, identifying the same animal regardless of how it is positioned in the photo or what kind of lighting there may be.

We trained these systems using a curated dataset of photos of snow leopards from zoos in the U.S., Europe and Tajikistan, and with images from the wild, including in Afghanistan.

Alone, each model worked about 74% of the time, correctly identifying the cat from a large photo library. But when combined, the two systems together were correct 85% of the time.

These algorithms were integrated into Wildbook, an open-source, web-based software platform developed by the nonprofit organization Wild Me and now adopted by ConservationX. We deployed the combined system on a free website, Whiskerbook.org, where researchers can upload images, seek matches using the algorithms, and confirm those matches with side-by-side comparisons. This site is among a growing family of AI-powered wildlife platforms that are helping conservation biologists work more efficiently and more effectively at protecting species and their habitats.

Humans still needed

These AI systems aren’t error-proof. AI quickly narrows down candidates and flags likely matches, but expert validation ensures accuracy, especially with tricky or ambiguous photos.

Another study we conducted pitted AI-assisted groups of experts and novices against each other. Each was given a set of three to 10 images of 34 known captive snow leopards and asked to use the Whiskerbook platform to identify them. They were also asked to estimate how many individual animals were in the set of photos.

The experts accurately matched about 90% of the images and delivered population estimates within about 3% of the true number. In contrast, the novices identified only 73% of the cats and underestimated the total number, sometimes by 25% or more, incorrectly merging two individuals into one.

Both sets of results were better than when experts or novices did not use any software.

The takeaway is clear: Human expertise remains important, and combining it with AI support leads to the most accurate results. My colleagues and I hope that by using tools like Whiskerbook and the AI systems embedded in them, researchers will be able to more quickly and more confidently study these elusive animals.

With AI tools like Whiskerbook illuminating the mysteries of these mountain ghosts, we have another way to safeguard snow leopards – but success depends on continued commitment to protecting their fragile mountain homes.

About the Author:

Eve Bohnett, Assistant Scholar, Center for Landscape Conservation Planning, University of Florida

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Whether you’re streaming a show, paying bills online or sending an email, each of these actions relies on computer programs that run behind the scenes. The process of writing computer programs is known as coding. Until recently, most computer code was written, at least originally, by human beings. But with the advent of generative artificial intelligence, that has begun to change.

Now, just as you can ask ChatGPT to spin up a recipe for a favorite dish or write a sonnet in the style of Lord Byron, you can now ask generative AI tools to write computer code for you. Andrej Karpathy, an OpenAI co-founder who previously led AI efforts at Tesla, recently termed this “vibe coding.”

For complete beginners or nontechnical dreamers, writing code based on vibes – feelings rather than explicitly defined information – could feel like a superpower. You don’t need to master programming languages or complex data structures. A simple natural language prompt will do the trick.

How it works

Vibe coding leans on standard patterns of technical language, which AI systems use to piece together original code from their training data. Any beginner can use an AI assistant such as GitHub Copilot or Cursor Chat, put in a few prompts, and let the system get to work. Here’s an example:

“Create a lively and interactive visual experience that reacts to music, user interaction or real-time data. Your animation should include smooth transitions and colorful and lively visuals with an engaging flow in the experience. The animation should feel organic and responsive to the music, user interaction or live data and facilitate an experience that is immersive and captivating. Complete this project using JavaScript or React, and allow for easy customization to set the mood for other experiences.”

But AI tools do this without any real grasp of specific rules, edge cases or security requirements for the software in question. This is a far cry from the processes behind developing production-grade software, which must balance trade-offs between product requirements, speed, scalability, sustainability and security. Skilled engineers write and review the code, run tests and establish safety barriers before going live.

But while the lack of a structured process saves time and lowers the skills required to code, there are trade-offs. With vibe coding, most of these stress-testing practices go out the window, leaving systems vulnerable to malicious attacks and leaks of personal data.

And there’s no easy fix: If you don’t understand every – or any – line of code that your AI agent writes, you can’t repair the code when it breaks. Or worse, as some experts have pointed out, you won’t notice when it’s silently failing.

The AI itself is not equipped to carry out this analysis either. It recognizes what “working” code usually looks like, but it cannot necessarily diagnose or fix deeper problems that the code might cause or exacerbate.

Why it matters

Vibe coding could be just a flash-in-the-pan phenomenon that will fizzle before long, but it may also find deeper applications with seasoned programmers. The practice could help skilled software engineers and developers more quickly turn an idea into a viable prototype. It could also enable novice programmers or even amateur coders to experience the power of AI, perhaps motivating them to pursue the discipline more deeply.

Vibe coding also may signal a shift that could make natural language a more viable tool for developing some computer programs. If so, it would echo early website editing systems known as WYSIWYG editors that promised designers “what you see is what you get,” or “drag-and-drop” website builders that made it easy for anyone with basic computer skills to launch a blog.

For now, I don’t believe that vibe coding will replace experienced software engineers, developers or computer scientists. The discipline and the art are much more nuanced than what AI can handle, and the risks of passing off “vibe code” as legitimate software are too great.

But as AI models improve and become more adept at incorporating context and accounting for risk, practices like vibe coding might cause the boundary between AI and human programmer to blur further.

About the Author:

Chetan Jaiswal, Associate Professor of Computer Science, Quinnipiac University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

By Daniel S. Schiff, Purdue University; Arne Bewersdorff, Technical University of Munich, and Marie Hornberger, Technical University of Munich 

It is “the policy of the United States to promote AI literacy and proficiency among Americans,” reads an executive order President Donald Trump issued on April 23, 2025. The executive order, titled Advancing Artificial Intelligence Education for American Youth, signals that advancing AI literacy is now an official national priority.

This raises a series of important questions: What exactly is AI literacy, who needs it, and how do you go about building it thoughtfully and responsibly?

The implications of AI literacy, or lack thereof, are far-reaching. They extend beyond national ambitions to remain “a global leader in this technological revolution” or even prepare an “AI-skilled workforce,” as the executive order states. Without basic literacy, citizens and consumers are not well equipped to understand the algorithmic platforms and decisions that affect so many domains of their lives: government services, privacy, lending, health care, news recommendations and more. And the lack of AI literacy risks ceding important aspects of society’s future to a handful of multinational companies.

How, then, can institutions help people understand and use – or resist – AI as individuals, workers, parents, innovators, job seekers, students, employers and citizens? We are a policy scientist and two educational researchers who study AI literacy, and we explore these issues in our research.

What AI literacy is and isn’t

At its foundation, AI literacy includes a mix of knowledge, skills and attitudes that are technical, social and ethical in nature. According to one prominent definition, AI literacy refers to “a set of competencies that enables individuals to critically evaluate AI technologies; communicate and collaborate effectively with AI; and use AI as a tool online, at home, and in the workplace.”

AI literacy is not simply programming or the mechanics of neural networks, and it is certainly not just prompt engineering – that is, the act of carefully writing prompts for chatbots. Vibe coding, or using AI to write software code, might be fun and important, but restricting the definition of literacy to the newest trend or the latest need of employers won’t cover the bases in the long term. And while a single master definition may not be needed, or even desirable, too much variation makes it tricky to decide on organizational, educational or policy strategies.

Who needs AI literacy? Everyone, including the employees and students using it, and the citizens grappling with its growing impacts. Every sector and sphere of society is now involved with AI, even if this isn’t always easy for people to see.

Exactly how much literacy everyone needs and how to get there is a much tougher question. Are a few quick HR training sessions enough, or do we need to embed AI across K-12 curricula and deliver university micro credentials and hands-on workshops? There is much that researchers don’t know, which leads to the need to measure AI literacy and the effectiveness of different training approaches.

Measuring AI literacy

While there is a growing and bipartisan consensus that AI literacy matters, there’s much less consensus on how to actually understand people’s AI literacy levels. Researchers have focused on different aspects, such as technical or ethical skills, or on different populations – for example, business managers and students – or even on subdomains like generative AI.

A recent review study identified more than a dozen questionnaires designed to measure AI literacy, the vast majority of which rely on self-reported responses to questions and statements such as “I feel confident about using AI.” There’s also a lack of testing to see whether these questionnaires work well for people from different cultural backgrounds.

Moreover, the rise of generative AI has exposed gaps and challenges: Is it possible to create a stable way to measure AI literacy when AI is itself so dynamic?

In our research collaboration, we’ve tried to help address some of these problems. In particular, we’ve focused on creating objective knowledge assessments, such as multiple-choice surveys tested with thorough statistical analyses to ensure that they accurately measure AI literacy. We’ve so far tested a multiple-choice survey in the U.S., U.K. and Germany and found that it works consistently and fairly across these three countries.

There’s a lot more work to do to create reliable and feasible testing approaches. But going forward, just asking people to self-report their AI literacy probably isn’t enough to understand where different groups of people are and what supports they need.

Approaches to building AI literacy

Governments, universities and industry are trying to advance AI literacy.

Finland launched the Elements of AI series in 2018 with the hope of educating its general public on AI. Estonia’s AI Leap initiative partners with Anthropic and OpenAI to provide access to AI tools for tens of thousands of students and thousands of teachers. And China is now requiring at least eight hours of AI education annually as early as elementary school, which goes a step beyond the new U.S. executive order. On the university level, Purdue University and the University of Pennsylvania have launched new master’s in AI programs, targeting future AI leaders.

Despite these efforts, these initiatives face an unclear and evolving understanding of AI literacy. They also face challenges to measuring effectiveness and minimal knowledge on what teaching approaches actually work. And there are long-standing issues with respect to equity − for example, reaching schools, communities, segments of the population and businesses that are stretched or under-resourced.

Next moves on AI literacy

Based on our research, experience as educators and collaboration with policymakers and technology companies, we think a few steps might be prudent.

Building AI literacy starts with recognizing it’s not just about tech: People also need to grasp the social and ethical sides of the technology. To see whether we’re getting there, we researchers and educators should use clear, reliable tests that track progress for different age groups and communities. Universities and companies can try out new teaching ideas first, then share what works through an independent hub. Educators, meanwhile, need proper training and resources, not just additional curricula, to bring AI into the classroom. And because opportunity isn’t spread evenly, partnerships that reach under-resourced schools and neighborhoods are essential so everyone can benefit.

Critically, achieving widespread AI literacy may be even harder than building digital and media literacy, so getting there will require serious investment – not cuts – to education and research.

There is widespread consensus that AI literacy is important, whether to boost AI trust and adoption or to empower citizens to challenge AI or shape its future. As with AI itself, we believe it’s important to approach AI literacy carefully, avoiding hype or an overly technical focus. The right approach can prepare students to become “active and responsible participants in the workforce of the future” and empower Americans to “thrive in an increasingly digital society,” as the AI literacy executive order calls for.

The Conversation will be hosting a free webinar on practical and safe use of AI with our tech editor and an AI expert on June 24 at 2pm ET/11am PT. Sign up to get your questions answered.

About the Author:

Daniel S. Schiff, Assistant Professor of Political Science, Purdue University; Arne Bewersdorff, Post Doctoral Researcher in Educational Sciences, Technical University of Munich, and Marie Hornberger, Research Associate at the School of Social Sciences and Technology, Technical University of Munich

This article is republished from The Conversation under a Creative Commons license. Read the original article.

By Bruce Schneier, Harvard Kennedy School and Nathan Sanders, Harvard University 

If you’ve worried that AI might take your job, deprive you of your livelihood, or maybe even replace your role in society, it probably feels good to see the latest AI tools fail spectacularly. If AI recommends glue as a pizza topping, then you’re safe for another day.

But the fact remains that AI already has definite advantages over even the most skilled humans, and knowing where these advantages arise — and where they don’t — will be key to adapting to the AI-infused workforce.

AI will often not be as effective as a human doing the same job. It won’t always know more or be more accurate. And it definitely won’t always be fairer or more reliable. But it may still be used whenever it has an advantage over humans in one of four dimensions: speed, scale, scope and sophistication. Understanding these dimensions is the key to understanding AI-human replacement.

Speed

First, speed. There are tasks that humans are perfectly good at but are not nearly as fast as AI. One example is restoring or upscaling images: taking pixelated, noisy or blurry images and making a crisper and higher-resolution version. Humans are good at this; given the right digital tools and enough time, they can fill in fine details. But they are too slow to efficiently process large images or videos.

AI models can do the job blazingly fast, a capability with important industrial applications. AI-based software is used to enhance satellite and remote sensing data, to compress video files, to make video games run better with cheaper hardware and less energy, to help robots make the right movements, and to model turbulence to help build better internal combustion engines.

Real-time performance matters in these cases, and the speed of AI is necessary to enable them.

Scale

The second dimension of AI’s advantage over humans is scale. AI will increasingly be used in tasks that humans can do well in one place at a time, but that AI can do in millions of places simultaneously. A familiar example is ad targeting and personalization. Human marketers can collect data and predict what types of people will respond to certain advertisements. This capability is important commercially; advertising is a trillion-dollar market globally.

AI models can do this for every single product, TV show, website and internet user. This is how the modern ad-tech industry works. Real-time bidding markets price the display ads that appear alongside the websites you visit, and advertisers use AI models to decide when they want to pay that price – thousands of times per second.

Scope

Next, scope. AI can be advantageous when it does more things than any one person could, even when a human might do better at any one of those tasks. Generative AI systems such as ChatGPT can engage in conversation on any topic, write an essay espousing any position, create poetry in any style and language, write computer code in any programming language, and more. These models may not be superior to skilled humans at any one of these things, but no single human could outperform top-tier generative models across them all.

It’s the combination of these competencies that generates value. Employers often struggle to find people with talents in disciplines such as software development and data science who also have strong prior knowledge of the employer’s domain. Organizations are likely to continue to rely on human specialists to write the best code and the best persuasive text, but they will increasingly be satisfied with AI when they just need a passable version of either.

Sophistication

Finally, sophistication. AIs can consider more factors in their decisions than humans can, and this can endow them with superhuman performance on specialized tasks. Computers have long been used to keep track of a multiplicity of factors that compound and interact in ways more complex than a human could trace. The 1990s chess-playing computer systems such as Deep Blue succeeded by thinking a dozen or more moves ahead.

Modern AI systems use a radically different approach: Deep learning systems built from many-layered neural networks take account of complex interactions – often many billions – among many factors. Neural networks now power the best chess-playing models and most other AI systems.

Chess is not the only domain where eschewing conventional rules and formal logic in favor of highly sophisticated and inscrutable systems has generated progress. The stunning advance of AlphaFold2, the AI model of structural biology whose creators Demis Hassabis and John Jumper were recognized with the Nobel Prize in chemistry in 2024, is another example.

This breakthrough replaced traditional physics-based systems for predicting how sequences of amino acids would fold into three-dimensional shapes with a 93 million-parameter model, even though it doesn’t account for physical laws. That lack of real-world grounding is not desirable: No one likes the enigmatic nature of these AI systems, and scientists are eager to understand better how they work.

But the sophistication of AI is providing value to scientists, and its use across scientific fields has grown exponentially in recent years.

Context matters

Those are the four dimensions where AI can excel over humans. Accuracy still matters. You wouldn’t want to use an AI that makes graphics look glitchy or targets ads randomly – yet accuracy isn’t the differentiator. The AI doesn’t need superhuman accuracy. It’s enough for AI to be merely good and fast, or adequate and scalable. Increasing scope often comes with an accuracy penalty, because AI can generalize poorly to truly novel tasks. The 4 S’s are sometimes at odds. With a given amount of computing power, you generally have to trade off scale for sophistication.

Even more interestingly, when an AI takes over a human task, the task can change. Sometimes the AI is just doing things differently. Other times, AI starts doing different things. These changes bring new opportunities and new risks.

For example, high-frequency trading isn’t just computers trading stocks faster; it’s a fundamentally different kind of trading that enables entirely new strategies, tactics and associated risks. Likewise, AI has developed more sophisticated strategies for the games of chess and Go. And the scale of AI chatbots has changed the nature of propaganda by allowing artificial voices to overwhelm human speech.

It is this “phase shift,” when changes in degree may transform into changes in kind, where AI’s impacts to society are likely to be most keenly felt. All of this points to the places that AI can have a positive impact. When a system has a bottleneck related to speed, scale, scope or sophistication, or when one of these factors poses a real barrier to being able to accomplish a goal, it makes sense to think about how AI could help.

Equally, when speed, scale, scope and sophistication are not primary barriers, it makes less sense to use AI. This is why AI auto-suggest features for short communications such as text messages can feel so annoying. They offer little speed advantage and no benefit from sophistication, while sacrificing the sincerity of human communication.

Many deployments of customer service chatbots also fail this test, which may explain their unpopularity. Companies invest in them because of their scalability, and yet the bots often become a barrier to support rather than a speedy or sophisticated problem solver.

Where the advantage lies

Keep this in mind when you encounter a new application for AI or consider AI as a replacement for or an augmentation to a human process. Looking for bottlenecks in speed, scale, scope and sophistication provides a framework for understanding where AI provides value, and equally where the unique capabilities of the human species give us an enduring advantage.

About the Author:

Bruce Schneier, Adjunct Lecturer in Public Policy, Harvard Kennedy School and Nathan Sanders, Affiliate, Berkman Klein Center for Internet & Society, Harvard University

This article is republished from The Conversation under a Creative Commons license. Read the original article.