AI Can Design Cities, But Can It Understand What Matters To People? 10 Ways To Keep Humans In Control

Cottonbro Studio/Pexels, CC BY

BY ABEER ELSHATER AND HISHAM ABUSAADA

Generative AI (GenAI) is a type of artificial intelligence that creates new content – like text, images, or ideas – by learning patterns from existing data. GenAI, particularly through large language models (LLMs) such as ChatGPT and DeepSeek, is rapidly becoming part of everyday urban design research and practice.

The models can summarise literature in seconds, generate policy scenarios, and help draft complex narratives.

For urban designers and researchers working under pressure, this feels like a breakthrough. But beneath this efficiency lies a deeper question: are we enhancing urban design knowledge, or quietly reshaping it in ways we do not fully understand?

Urban design is an academic and professional field concerned with shaping the physical form and experience of cities. It looks at the relationships between buildings, spaces, people and activities within broader urban systems.

The field has evolved differently across regions, reflecting diverse historical, political and spatial contexts. For example, in Europe, urban design has often been shaped by post-war reconstruction and rehabilitation of the destroyed urban forms, while in the United States it has been influenced by urban renewal policies and large-scale redevelopment. Urban design is not a fixed set of principles, but a context-dependent theory and practice that responds to specific local challenges and conditions.

GenAI is now widely used in urban design to help with analysis and decision-making. For instance, researchers use machine learning to study pedestrian movement and traffic patterns from video data, which helps planners create safer and more efficient streets.

Some studies use GenAI to create and test different urban design options, such as changing land use, building density, or access to green spaces, so designers can quickly compare choices. In environmental planning, GenAI models can simulate urban heat or air quality, helping with climate-sensitive decisions. These examples show that GenAI provides ways to test ideas and handle complex challenges, rather than replacing designers.

Our work as urban designers and researchers has always depended on interpretation, context and ethical judgment. Cities are not just datasets; they are lived environments shaped by history, culture and power. When LLMs enter this space, they influence how problems are framed and how solutions are imagined. Their use therefore should not be just technical, but should be managed critically. Each theory developed for a particular city or place evolved to address the needs of specific groups of people within a distinct context and for a particular purpose. LLMs need to be developed faster to have this sensitivity about people and place history.

Our recent research was motivated by the rapid and often uncritical integration of LLMs into planning research and practice. The work asks a central question: how do these tools reshape the way urban knowledge is produced, interpreted and validated in a discipline that depends heavily on context, judgment and field-based understanding?

Our key finding is that LLMs can be very helpful; they can speed up writing, support analysis and help explore ideas. However, they also carry important risks, especially when their outputs are treated as fully correct or used without considering context.

We propose some cornerstones for responsible use. These are not strict rules, but practical guides to keep human judgment central, ensure ideas stay grounded in context, and maintain responsibility in planning research and practice.

10 cornerstones

Research sovereignty should remain with the human. The direction of inquiry must always come from the researcher. If planners begin by asking the model what to study or how to frame a problem, they risk producing inconsistency and generic outputs.

Engagement with GenAI is critical, not passive. LLMs generate plausible text based on patterns, not verified truth. This means every output should be tested and refined. Accepting it at face value risks embedding hidden biases and weak assumptions.

Knowledge should be grounded in context. Cities are deeply specific. A recommendation that works in one place may fail in another due to social, political, or cultural differences. LLMs tend to produce generalised solutions without understanding local realities. Planners must anchor these suggestions in field knowledge and community insight.

Everyone should be careful. They should not trust GenAI too quickly. In planning debates such as zoning or rent control, LLMs can sound very confident, even when they are wrong. Sometimes they may even give references that do not exist. This can spread incorrect information and weaken trust in research.

While any of the LLMs can assist in identifying and organising sources, they cannot replace the critical judgment required to assess accuracy, context and fit. The responsibility for validating references remains with the researcher.

Planners must recognise that LLMs do not “remember” in the way humans do. They lack continuity across conversations and can lose track of earlier assumptions. AI forgets things. Maintaining coherence in long-term research, therefore, depends on the researcher, not the tool.

A subtler issue is rigidity. LLMs often repeat dominant ideas or default solutions, even when the context differs. For example, when asked how to improve a congested street, an LLM may suggest widening roads or adding car lanes, even where such interventions could harm walkability and heritage value. Breaking out of these patterns requires active intervention.

We can understand GenAI as a partner in thinking, but not an equal one. The planner must decide what matters, whose voices are included, and what ethical priorities guide the work.

Effective use of GenAI requires strategic manoeuvring. This means combining AI-generated insights with collected data, community engagement and professional judgment. The value of LLMs lies not in replacing urban design processes, but in enriching them, if used carefully.

Academic integrity is non-negotiable. Urban design research is not just about producing text; it is about engaging intellectually with people, places and consequences.

Why this matters

GenAI in urban design is like fire – powerful, but dangerous without human control.

Used well, GenAI can help urban designers think more broadly and act more effectively. Used poorly, it risks reducing urban design to automated generalisation, detached from the lived experience of cities.

The future of urban design is not about choosing between humans and machines, but about designing thoughtful collaboration between them. The challenge is not whether machines can think, but how we think with them.

READ RIGINAL STORY HERE

The Greatest Risk Of AI In Higher Education Isn’t Cheating – It’s The Erosion Of Learning Itself

Growing sophistication and autonomy of technology systems means that scientific research can increasingly be automated, potentially leaving people with fewer opportunities to gain skills practicing research methods. NurPhoto/Getty Images

BY NIR EISIKOVITS AND JACOB BURLEY

Public debate about artificial intelligence in higher education has largely orbited a familiar worry: cheating. Will students use chatbots to write essays? Can instructors tell? Should universities ban the tech? Embrace it?

These concerns are understandable. But focusing so much on cheating misses the larger transformation already underway, one that extends far beyond student misconduct and even the classroom.

Universities are adopting AI across many areas of institutional life. Some uses are largely invisible, like systems that help allocate resources, flag “at-risk” students, optimize course scheduling or automate routine administrative decisions. Other uses are more noticeable. Students use AI tools to summarize and study, instructors use them to build assignments and syllabuses and researchers use them to write code, scan literature and compress hours of tedious work into minutes.

People may use AI to cheat or skip out on work assignments. But the many uses of AI in higher education, and the changes they portend, beg a much deeper question: As machines become more capable of doing the labor of research and learning, what happens to higher education? What purpose does the university serve?

Over the past eight years, we’ve been studying the moral implications of pervasive engagement with AI as part of a joint research project between the Applied Ethics Center at UMass Boston and the Institute for Ethics and Emerging Technologies. In a recent white paper, we argue that as AI systems become more autonomous, the ethical stakes of AI use in higher ed rise, as do its potential consequences.

As these technologies become better at producing knowledge work – designing classes, writing papers, suggesting experiments and summarizing difficult texts – they don’t just make universities more productive. They risk hollowing out the ecosystem of learning and mentorship upon which these institutions are built, and on which they depend.

Nonautonomous AI

Consider three kinds of AI systems and their respective impacts on university life:

AI-powered software is already being used throughout higher education in admissions review, purchasing, academic advising and institutional risk assessment. These are considered “nonautonomous” systems because they automate tasks, but a person is “in the loop” and using these systems as tools.

These technologies can pose a risk to students’ privacy and data security. They also can be biased. And they often lack sufficient transparency to determine the sources of these problems. Who has access to student data? How are “risk scores” generated? How do we prevent systems from reproducing inequities or treating certain students as problems to be managed?

These questions are serious, but they are not conceptually new, at least within the field of computer science. Universities typically have compliance offices, institutional review boards and governance mechanisms that are designed to help address or mitigate these risks, even if they sometimes fall short of these objectives.

Hybrid AI

Hybrid systems encompass a range of tools, including AI-assisted tutoring chatbots, personalized feedback tools and automated writing support. They often rely on generative AI technologies, especially large language models. While human users set the overall goals, the intermediate steps the system takes to meet them are often not specified.

Hybrid systems are increasingly shaping day-to-day academic work. Students use them as writing companions, tutors, brainstorming partners and on-demand explainers. Faculty use them to generate rubrics, draft lectures and design syllabuses. Researchers use them to summarize papers, comment on drafts, design experiments and generate code.

This is where the “cheating” conversation belongs. With students and faculty alike increasingly leaning on technology for help, it is reasonable to wonder what kinds of learning might get lost along the way. But hybrid systems also raise more complex ethical questions.

One has to do with transparency. AI chatbots offer natural-language interfaces that make it hard to tell when you’re interacting with a human and when you’re interacting with an automated agent. That can be alienating and distracting for those who interact with them. A student reviewing material for a test should be able to tell if they are talking with their teaching assistant or with a robot. A student reading feedback on a term paper needs to know whether it was written by their instructor. Anything less than complete transparency in such cases will be alienating to everyone involved and will shift the focus of academic interactions from learning to the means or the technology of learning. University of Pittsburgh researchers have shown that these dynamics bring forth feelings of uncertainty, anxiety and distrust for students. These are problematic outcomes.

A second ethical question relates to accountability and intellectual credit. If an instructor uses AI to draft an assignment and a student uses AI to draft a response, who is doing the evaluating, and what exactly is being evaluated? If feedback is partly machine-generated, who is responsible when it misleads, discourages or embeds hidden assumptions? And when AI contributes substantially to research synthesis or writing, universities will need clearer norms around authorship and responsibility – not only for students, but also for faculty.

Finally, there is the critical question of cognitive offloading. AI can reduce drudgery, and that’s not inherently bad. But it can also shift users away from the parts of learning that build competence, such as generating ideas, struggling through confusion, revising a clumsy draft and learning to spot one’s own mistakes.

Autonomous agents

The most consequential changes may come with systems that look less like assistants and more like agents. While truly autonomous technologies remain aspirational, the dream of a researcher “in a box” – an agentic AI system that can perform studies on its own – is becoming increasingly realistic.

Agentic tools are anticipated to “free up time” for work that focuses on more human capacities like empathy and problem-solving. In teaching, this may mean that faculty may still teach in the headline sense, but more of the day-to-day labor of instruction can be handed off to systems optimized for efficiency and scale. Similarly, in research, the trajectory points toward systems that can increasingly automate the research cycle. In some domains, that already looks like robotic laboratories that run continuously, automate large portions of experimentation and even select new tests based on prior results.

At first glance, this may sound like a welcome boost to productivity. But universities are not information factories; they are systems of practice. They rely on a pipeline of graduate students and early-career academics who learn to teach and research by participating in that same work. If autonomous agents absorb more of the “routine” responsibilities that historically served as on-ramps into academic life, the university may keep producing courses and publications while quietly thinning the opportunity structures that sustain expertise over time.

The same dynamic applies to undergraduates, albeit in a different register. When AI systems can supply explanations, drafts, solutions and study plans on demand, the temptation is to offload the most challenging parts of learning. To the industry that is pushing AI into universities, it may seem as if this type of work is “inefficient” and that students will be better off letting a machine handle it. But it is the very nature of that struggle that builds durable understanding. Cognitive psychology has shown that students grow intellectually through doing the work of drafting, revising, failing, trying again, grappling with confusion and revising weak arguments. This is the work of learning how to learn.

Taken together, these developments suggest that the greatest risk posed by automation in higher education is not simply the replacement of particular tasks by machines, but the erosion of the broader ecosystem of practice that has long sustained teaching, research and learning.

An uncomfortable inflection point

So what purpose do universities serve in a world in which knowledge work is increasingly automated?

One possible answer treats the university primarily as an engine for producing credentials and knowledge. There, the core question is output: Are students graduating with degrees? Are papers and discoveries being generated? If autonomous systems can deliver those outputs more efficiently, then the institution has every reason to adopt them.

But another answer treats the university as something more than an output machine, acknowledging that the value of higher education lies partly in the ecosystem itself. This model assigns intrinsic value to the pipeline of opportunities through which novices become experts, the mentorship structures through which judgment and responsibility are cultivated, and the educational design that encourages productive struggle rather than optimizing it away. Here, what matters is not only whether knowledge and degrees are produced, but how they are produced and what kinds of people, capacities and communities are formed in the process. In this version, the university is meant to serve as no less than an ecosystem that reliably forms human expertise and judgment.

In a world where knowledge work itself is increasingly automated, we think universities must ask what higher education owes its students, its early-career scholars and the society it serves. The answers will determine not only how AI is adopted, but also what the modern university becomes.

READ ORIGINAL STORY HERE

AI Is About To Radically Alter Military Command Structures That Haven’t Changed Much Since Napoleon’s Army

This U.S. Army command post, seen from a drone, is loaded with modern technology but uses a centuries-old structure. Col. Scott Woodward, U.S. Army

BY BENJAMIN JENSEN

PROFESSOR OF STRATEGIC STUDIES AT

THE MARINE CORPS UNIVERSITY

SCHOOL OF ADVANCED WARFIGHTING,

SCHOLAR-IN-RESIDENCE, AMERICAN

UNIVERSITY SCHOOL OF INTERNATIONAL

SERVICE

Despite two centuries of evolution, the structure of a modern military staff would be recognizable to Napoleon. At the same time, military organizations have struggled to incorporate new technologies as they adapt to new domains – air, space and information – in modern war.

The sizes of military headquarters have grown to accommodate the expanded information flows and decision points of these new facets of warfare. The result is diminishing marginal returns and a coordination nightmare – too many cooks in the kitchen – that risks jeopardizing mission command.

AI agents – autonomous, goal-oriented software powered by large language models – can automate routine staff tasks, compress decision timelines and enable smaller, more resilient command posts. They can shrink the staff while also making it more effective.

As an international relations scholar and reserve officer in the U.S. Army who studies military strategy, I see both the opportunity afforded by the technology and the acute need for change.

That need stems from the reality that today’s command structures still mirror Napoleon’s field headquarters in both form and function – industrial-age architectures built for massed armies. Over time, these staffs have ballooned in size, making coordination cumbersome. They also result in sprawling command posts that modern precision artillery, missiles and drones can target effectively and electronic warfare can readily disrupt.

Russia’s so-called “Graveyard of Command Posts” in Ukraine vividly illustrates how static headquarters where opponents can mass precision artillery, missiles and drones become liabilities on a modern battlefield.

The role of AI agents

Military planners now see a world in which AI agents – autonomous, goal-oriented software that can perceive, decide and act on their own initiative – are mature enough to deploy in command systems. These agents promise to automate the fusion of multiple sources of intelligence, threat-modeling, and even limited decision cycles in support of a commander’s goals. There is still a human in the loop, but the humans will be able to issue commands faster and receive more timely and contextual updates from the battlefield.

These AI agents can parse doctrinal manuals, draft operational plans and generate courses of action, which helps accelerate the tempo of military operations. Experiments – including efforts I ran at Marine Corps University – have demonstrated how even basic large language models can accelerate staff estimates and inject creative, data-driven options into the planning process. These efforts point to the end of traditional staff roles.

There will still be people – war is a human endeavor – and ethics will still factor into streams of algorithms making decisions. But the people who remain deployed are likely to gain the ability to navigate mass volumes of information with the help of AI agents.

These teams are likely to be smaller than modern staffs. AI agents will allow teams to manage multiple planning groups simultaneously.

For example, they will be able to use more dynamic red teaming techniques – role-playing the opposition – and vary key assumptions to create a wider menu of options than traditional plans. The time saved not having to build PowerPoint slides and updating staff estimates will be shifted to contingency analysis – asking “what if” questions – and building operational assessment frameworks – conceptual maps of how a plan is likely to play out in a particular situation – that provide more flexibility to commanders.

Designing the next military staff

To explore the optimal design of this AI agent-augmented staff, I led a team of researchers at the bipartisan think tank Center for Strategic & International Studies’ Futures Lab to explore alternatives. The team developed three baseline scenarios reflecting what most military analysts are seeing as the key operational problems in modern great power competition: joint blockades, firepower strikes and joint island campaigns. Joint refers to an action coordinated among multiple branches of a military.

In the example of China and Taiwan, joint blockades describe how China could isolate the island nation and either starve it or set conditions for an invasion. Firepower strikes describe how Beijing could fire salvos of missiles – similar to what Russia is doing in Ukraine – to destroy key military centers and even critical infrastructure. Last, in Chinese doctrine, a Joint Island Landing Campaign describes the cross-strait invasion their military has spent decades refining.

Any AI agent-augmented staff should be able to manage warfighting functions across these three operational scenarios.

The research team found that the best model kept humans in the loop and focused on feedback loops. This approach – called the Adaptive Staff Model and based on pioneering work by sociologist Andrew Abbott – embeds AI agents within continuous human-machine feedback loops, drawing on doctrine, history and real-time data to evolve plans on the fly.

In this model, military planning is ongoing and never complete, and focused more on generating a menu of options for the commander to consider, refine and enact. The research team tested the approach with multiple AI models and found that it outperformed alternatives in each case.

AI agents are not without risk. First, they can be overly generalized, if not biased. Foundation models – AI models trained on extremely large datasets and adaptable to a wide range of tasks – know more about pop culture than war and require refinement. This makes it important to benchmark agents to understand their strengths and limitations.

Second, absent training in AI fundamentals and advanced analytical reasoning, many users tend to use models as a substitute for critical thinking. No smart model can make up for a dumb, or worse, lazy user.

Seizing the ‘agentic’ moment

To take advantage of AI agents, the U.S. military will need to institutionalize building and adapting agents, include adaptive agents in war games, and overhaul doctrine and training to account for human-machine teams. This will require a number of changes.

First, the military will need to invest in additional computational power to build the infrastructure required to run AI agents across military formations. Second, they will need to develop additional cybersecurity measures and conduct stress tests to ensure the agent-augmented staff isn’t vulnerable when attacked across multiple domains, including cyberspace and the electromagnetic spectrum.

Third, and most important, the military will need to dramatically change how it educates its officers. Officers will have to learn how AI agents work, including how to build them, and start using the classroom as a lab to develop new approaches to the age-old art of military command and decision-making. This could include revamping some military schools to focus on AI, a concept floated in the White House’s AI Action Plan released on July 23, 2025.

Absent these reforms, the military is likely to remain stuck in the Napoleonic staff trap: adding more people to solve ever more complex problems.

READ ORIGINAL STORY HERE

AI Tools Collect And Store Data About You From All Your Devices – Here’s How To Be Aware Of What You’re Revealing

ChatGPT stores and analyzes everything you type into a prompt screen. Screenshot by Christopher Ramezan, CC BY-ND

BY CHRISTOPHER RAMEZAN

ASSISTANT PROFESSOR OF 

CYBERSECURITY,

WEST VIRGINIA UNIVERSITY

Like it or not, artificial intelligence has become part of daily life. Many devices – including electric razors and toothbrushes – have become “AI-powered,” using machine learning algorithms to track how a person uses the device, how the device is working in real time, and provide feedback. From asking questions to an AI assistant like ChatGPT or Microsoft Copilot to monitoring a daily fitness routine with a smartwatch, many people use an AI system or tool every day.

While AI tools and technologies can make life easier, they also raise important questions about data privacy. These systems often collect large amounts of data, sometimes without people even realizing their data is being collected. The information can then be used to identify personal habits and preferences, and even predict future behaviors by drawing inferences from the aggregated data.

As an assistant professor of cybersecurity at West Virginia University, I study how emerging technologies and various types of AI systems manage personal data and how we can build more secure, privacy-preserving systems for the future.

Generative AI software uses large amounts of training data to create new content such as text or images. Predictive AI uses data to forecast outcomes based on past behavior, such as how likely you are to hit your daily step goal, or what movies you may want to watch. Both types can be used to gather information about you.

How AI tools collect data

Generative AI assistants such as ChatGPT and Google Gemini collect all the information users type into a chat box. Every question, response and prompt that users enter is recorded, stored and analyzed to improve the AI model.

OpenAI’s privacy policy informs users that “we may use content you provide us to improve our Services, for example to train the models that power ChatGPT.” Even though OpenAI allows you to opt out of content use for model training, it still collects and retains your personal data. Although some companies promise that they anonymize this data, meaning they store it without naming the person who provided it, there is always a risk of data being reidentified.

Predictive AI

Beyond generative AI assistants, social media platforms like Facebook, Instagram and TikTok continuously gather data on their users to train predictive AI models. Every post, photo, video, like, share and comment, including the amount of time people spend looking at each of these, is collected as data points that are used to build digital data profiles for each person who uses the service.

The profiles can be used to refine the social media platform’s AI recommender systems. They can also be sold to data brokers, who sell a person’s data to other companies to, for instance, help develop targeted advertisements that align with that person’s interests.

Many social media companies also track users across websites and applications by putting cookies and embedded tracking pixels on their computers. Cookies are small files that store information about who you are and what you clicked on while browsing a website.

One of the most common uses of cookies is in digital shopping carts: When you place an item in your cart, leave the website and return later, the item will still be in your cart because the cookie stored that information. Tracking pixels are invisible images or snippets of code embedded in websites that notify companies of your activity when you visit their page. This helps them track your behavior across the internet.

This is why users often see or hear advertisements that are related to their browsing and shopping habits on many of the unrelated websites they browse, and even when they are using different devices, including computers, phones and smart speakers. One study found that some websites can store over 300 tracking cookies on your computer or mobile phone.

Data privacy controls – and limitations

Like generative AI platforms, social media platforms offer privacy settings and opt-outs, but these give people limited control over how their personal data is aggregated and monetized. As media theorist Douglas Rushkoff argued in 2011, if the service is free, you are the product.

Many tools that include AI don’t require a person to take any direct action for the tool to collect data about that person. Smart devices such as home speakers, fitness trackers and watches continually gather information through biometric sensors, voice recognition and location tracking. Smart home speakers continually listen for the command to activate or “wake up” the device. As the device is listening for this word, it picks up all the conversations happening around it, even though it does not seem to be active.

Some companies claim that voice data is only stored when the wake word – what you say to wake up the device – is detected. However, people have raised concerns about accidental recordings, especially because these devices are often connected to cloud services, which allow voice data to be stored, synced and shared across multiple devices such as your phone, smart speaker and tablet.

If the company allows, it’s also possible for this data to be accessed by third parties, such as advertisers, data analytics firms or a law enforcement agency with a warrant.

Privacy rollbacks

This potential for third-party access also applies to smartwatches and fitness trackers, which monitor health metrics and user activity patterns. Companies that produce wearable fitness devices are not considered “covered entities” and so are not bound by the Health Information Portability and Accountability Act. This means that they are legally allowed to sell health- and location-related data collected from their users.

Concerns about HIPAA data arose in 2018, when Strava, a fitness company released a global heat map of user’s exercise routes. In doing so, it accidentally revealed sensitive military locations across the globe through highlighting the exercise routes of military personnel.

The Trump administration has tapped Palantir, a company that specializes in using AI for data analytics, to collate and analyze data about Americans. Meanwhile, Palantir has announced a partnership with a company that runs self-checkout systems.

Such partnerships can expand corporate and government reach into everyday consumer behavior. This one could be used to create detailed personal profiles on Americans by linking their consumer habits with other personal data. This raises concerns about increased surveillance and loss of anonymity. It could allow citizens to be tracked and analyzed across multiple aspects of their lives without their knowledge or consent.

Some smart device companies are also rolling back privacy protections instead of strengthening them. Amazon recently announced that starting on March 28, 2025, all voice recordings from Amazon Echo devices would be sent to Amazon’s cloud by default, and users will no longer have the option to turn this function off. This is different from previous settings, which allowed users to limit private data collection.

Changes like these raise concerns about how much control consumers have over their own data when using smart devices. Many privacy experts consider cloud storage of voice recordings a form of data collection, especially when used to improve algorithms or build user profiles, which has implications for data privacy laws designed to protect online privacy.

Implications for data privacy

All of this brings up serious privacy concerns for people and governments on how AI tools collect, store, use and transmit data. The biggest concern is transparency. People don’t know what data is being collected, how the data is being used, and who has access to that data.

Companies tend to use complicated privacy policies filled with technical jargon to make it difficult for people to understand the terms of a service that they agree to. People also tend not to read terms of service documents. One study found that people averaged 73 seconds reading a terms of service document that had an average read time of 29-32 minutes.

Data collected by AI tools may initially reside with a company that you trust, but can easily be sold and given to a company that you don’t trust.

AI tools, the companies in charge of them and the companies that have access to the data they collect can also be subject to cyberattacks and data breaches that can reveal sensitive personal information. These attacks can by carried out by cybercriminals who are in it for the money, or by so-called advanced persistent threats, which are typically nation/state- sponsored attackers who gain access to networks and systems and remain there undetected, collecting information and personal data to eventually cause disruption or harm.

While laws and regulations such as the General Data Protection Regulation in the European Union and the California Consumer Privacy Act aim to safeguard user data, AI development and use have often outpaced the legislative process. The laws are still catching up on AI and data privacy. For now, you should assume any AI-powered device or platform is collecting data on your inputs, behaviors and patterns.

Using AI tools

Although AI tools collect people’s data, and the way this accumulation of data affects people’s data privacy is concerning, the tools can also be useful. AI-powered applications can streamline workflows, automate repetitive tasks and provide valuable insights.

But it’s crucial to approach these tools with awareness and caution.

When using a generative AI platform that gives you answers to questions you type in a prompt, don’t include any personally identifiable information, including names, birth dates, Social Security numbers or home addresses. At the workplace, don’t include trade secrets or classified information. In general, don’t put anything into a prompt that you wouldn’t feel comfortable revealing to the public or seeing on a billboard. Remember, once you hit enter on the prompt, you’ve lost control of that information.

Remember that devices which are turned on are always listening – even if they’re asleep. If you use smart home or embedded devices, turn them off when you need to have a private conversation. A device that’s asleep looks inactive, but it is still powered on and listening for a wake word or signal. Unplugging a device or removing its batteries is a good way of making sure the device is truly off.

Finally, be aware of the terms of service and data collection policies of the devices and platforms that you are using. You might be surprised by what you’ve already agreed to.

This article is part of a series on data privacy that explores who collects your data, what and how they collect, who sells and buys your data, what they all do with it, and what you can do about it.

READ ORIGINAL STORY HERE

AI Literacy: What It Is, What It Isn’t, Who Needs It And Why It’s Hard To Define

Ethics is an important aspect of AI literacy.

BY DANIEL S, SCHIFF, ARNE BEWERSDORF AND MARIE HORNBERGER

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.

READ ORIGINAL STORY HERE

An AI System Has Reached Human Level On A Test For ‘General Intelligence’. Here’s What That Means

AUTHORS:

MICHAEL TOMOTHY BENNETT

PH.D STUDENT, SCHOOL OF 

COMPUTING, AUSTRALIAN 

NATIONAL UNIVERSITY

ELIJA PERRIER

RESEARCH FELLOW, 

STANFORD CENTER FOR 

RESPONSIBLE QUANTUM,

STANFORD UNIVERSITY

TECHNOLOGY

A new artificial intelligence (AI) model has just achieved human-level results on a test designed to measure “general intelligence”.

On December 20, OpenAI’s o3 system scored 85% on the ARC-AGI benchmark, well above the previous AI best score of 55% and on par with the average human score. It also scored well on a very difficult mathematics test.

Creating artificial general intelligence, or AGI, is the stated goal of all the major AI research labs. At first glance, OpenAI appears to have at least made a significant step towards this goal.

While scepticism remains, many AI researchers and developers feel something just changed. For many, the prospect of AGI now seems more real, urgent and closer than anticipated. Are they right?

Generalisation and intelligence

To understand what the o3 result means, you need to understand what the ARC-AGI test is all about. In technical terms, it’s a test of an AI system’s “sample efficiency” in adapting to something new – how many examples of a novel situation the system needs to see to figure out how it works.

An AI system like ChatGPT (GPT-4) is not very sample efficient. It was “trained” on millions of examples of human text, constructing probabilistic “rules” about which combinations of words are most likely.

The result is pretty good at common tasks. It is bad at uncommon tasks, because it has less data (fewer samples) about those tasks.

Until AI systems can learn from small numbers of examples and adapt with more sample efficiency, they will only be used for very repetitive jobs and ones where the occasional failure is tolerable.

The ability to accurately solve previously unknown or novel problems from limited samples of data is known as the capacity to generalise. It is widely considered a necessary, even fundamental, element of intelligence.

Grids and patterns

The ARC-AGI benchmark tests for sample efficient adaptation using little grid square problems like the one below. The AI needs to figure out the pattern that turns the grid on the left into the grid on the right.

Each question gives three examples to learn from. The AI system then needs to figure out the rules that “generalise” from the three examples to the fourth.

These are a lot like the IQ tests sometimes you might remember from school.

Weak rules and adaptation

We don’t know exactly how OpenAI has done it, but the results suggest the o3 model is highly adaptable. From just a few examples, it finds rules that can be generalised.

To figure out a pattern, we shouldn’t make any unnecessary assumptions, or be more specific than we really have to be. In theory, if you can identify the “weakest” rules that do what you want, then you have maximised your ability to adapt to new situations.

What do we mean by the weakest rules? The technical definition is complicated, but weaker rules are usually ones that can be described in simpler statements.

In the example above, a plain English expression of the rule might be something like: “Any shape with a protruding line will move to the end of that line and ‘cover up’ any other shapes it overlaps with.”

Searching chains of thought?

While we don’t know how OpenAI achieved this result just yet, it seems unlikely they deliberately optimised the o3 system to find weak rules. However, to succeed at the ARC-AGI tasks it must be finding them.

We do know that OpenAI started with a general-purpose version of the o3 model (which differs from most other models, because it can spend more time “thinking” about difficult questions) and then trained it specifically for the ARC-AGI test.

French AI researcher Francois Chollet, who designed the benchmark, believes o3 searches through different “chains of thought” describing steps to solve the task. It would then choose the “best” according to some loosely defined rule, or “heuristic”.

This would be “not dissimilar” to how Google’s AlphaGo system searched through different possible sequences of moves to beat the world Go champion.

You can think of these chains of thought like programs that fit the examples. Of course, if it is like the Go-playing AI, then it needs a heuristic, or loose rule, to decide which program is best.

There could be thousands of different seemingly equally valid programs generated. That heuristic could be “choose the weakest” or “choose the simplest”.

However, if it is like AlphaGo then they simply had an AI create a heuristic. This was the process for AlphaGo. Google trained a model to rate different sequences of moves as better or worse than others.

What we still don’t know

The question then is, is this really closer to AGI? If that is how o3 works, then the underlying model might not be much better than previous models.

The concepts the model learns from language might not be any more suitable for generalisation than before. Instead, we may just be seeing a more generalisable “chain of thought” found through the extra steps of training a heuristic specialised to this test. The proof, as always, will be in the pudding.

Almost everything about o3 remains unknown. OpenAI has limited disclosure to a few media presentations and early testing to a handful of researchers, laboratories and AI safety institutions.

Truly understanding the potential of o3 will require extensive work, including evaluations, an understanding of the distribution of its capacities, how often it fails and how often it succeeds.

When o3 is finally released, we’ll have a much better idea of whether it is approximately as adaptable as an average human.

If so, it could have a huge, revolutionary, economic impact, ushering in a new era of self-improving accelerated intelligence. We will require new benchmarks for AGI itself and serious consideration of how it ought to be governed.

If not, then this will still be an impressive result. However, everyday life will remain much the same.

READ ORIGINAL STORY HERE

4 Ways AI Can Be Used And Abused In The 2024 Election, From Deepfakes To Foreign Interference

Special counsel Robert Mueller’s investigation into the 2016 U.S. election concluded that Russia had worked to get President Donald Trump elected. Jonathan Ernst/Pool via AP

BY BARBARA A. TRISH

PROFESSOR OF POLITICAL SCIENCE

GRINNELL COLLEGE

The American public is on alert about artificial intelligence and the 2024 election.

A September 2024 poll by the Pew Research Center found that well over half of Americans worry that artificial intelligence – or AI, computer technology mimicking the processes and products of human intelligence – will be used to generate and spread false and misleading information in the campaign.

My academic research on AI may help quell some concerns. While this innovative technology certainly has the potential to manipulate voters or spread lies at scale, most uses of AI in the current election cycle are, so far, not novel at all.

I’ve identified four roles AI is playing or could play in the 2024 campaign – all arguably updated versions of familiar election activities.

1. Voter information

The 2022 launch of ChatGPT brought the promise and peril of generative AI into public consciousness. This technology is called “generative” because it produces text responses to user prompts: It can write poetry, answer history questions – and provide information about the 2024 election.

Rather than search Google for voting information, people may instead ask generative AI a question. “How much has inflation changed since 2020?” for example. Or, “Who’s running for U.S. Senate in Texas?”

Some generative AI platforms such as Google’s AI chatbot Gemini, decline to answer questions about candidates and voting. Some, such as Facebook’s AI tool Llama, respond – and respond accurately.

But generative AI can also produce misinformation. In the most extreme cases, AI can have “hallucinations,” offering up wildly inaccurate results.

A CBS news account from June 2024 reported that ChatGPT had given incorrect or incomplete responses to some prompts asking how to vote in battleground states. And ChatGPT didn’t consistently follow the policy of its owner, OpenAI, and refer users to CanIVote.org, a respected site for voting information.

As with the web, people should verify the results of AI searches. And beware: Google’s Gemini now automatically returns answers to Google search queries at the top of every results page. You might inadvertently stumble into AI tools when you think you’re searching the internet.

2. Deepfakes

Deepfakes are fabricated images, audio and video produced by generative AI and designed to replicate reality. Essentially, these are highly convincing versions of what are now called “cheapfakes” – altered images made using basic tools such as Photoshop and video-editing software.

The potential of deepfakes to deceive voters became clear when an AI-generated robocall impersonating Joe Biden before the January 2024 New Hampshire primary advised Democrats to save their votes for November.

After that, the Federal Communication Commission ruled that AI-generated robocalls are subject to the same regulations as all robocalls. They cannot be auto-dialed or delivered to cellphones or landlines without prior consent.

The agency also slapped a US$6 million fine on the consultant who created the fake Biden call – but not for tricking voters. He was fined for transmitting inaccurate caller-ID information.

While synthetic media can be used to spread disinformation, deepfakes are now part of the creative toolbox of political advertisers.

One early deepfake aimed more at persuasion than overt deception was an AI-generated ad from a 2022 mayoral race contest portraying the then-incumbent mayor of Shreveport, Louisiana, as a failing student summoned to the principal’s office.

The ad included a quick disclaimer that it was a deepfake, a warning not required by the federal government, but it was easy to miss.

Wired magazine’s AI Elections Project, which is tracking uses of AI in the 2024 cycle, shows that deepfakes haven’t overwhelmed the ads voters see. But they have been used by candidates across the political spectrum, up and down the ballot, for many purposes – including deception.

Former President Donald Trump hints at a Democratic deepfake when he questions the crowd size at Vice President Kamala Harris’ campaign events. In lobbing such allegations, Trump is attempting to reap the “liar’s dividend” – the opportunity to plant the idea that truthful content is fake.

Discrediting a political opponent this way is nothing new. Trump has been claiming that the truth is really just “fake news” since at least the “birther” conspiracy of 2008, when he helped to spread rumors that presidential candidate Barack Obama’s birth certificate was fake.

3. Strategic distraction

Some are concerned that AI might be used by election deniers in this cycle to distract election administrators by burying them in frivolous public records requests.

For example, the group True the Vote has lodged hundreds of thousands of voter challenges over the past decade working with just volunteers and a web-based app. Imagine its reach if armed with AI to automate their work.

Such widespread, rapid-fire challenges to the voter rolls could divert election administrators from other critical tasks, disenfranchise legitimate voters and disrupt the election.

As of now, there’s no evidence that this is happening.

4. Foreign election interference

Confirmed Russian interference in the 2016 election underscored that the threat of foreign meddling in U.S. politics, whether by Russia or another country invested in discrediting Western democracy, remains a pressing concern.

In July, the Department of Justice seized two domain names and searched close to 1,000 accounts that Russian actors had used for what it called a “social media bot farm,” similar to those Russia used to influence the opinions of hundreds of millions of Facebook users in the 2020 campaign. Artificial intelligence could give these efforts a real boost.

There’s also evidence that China is using AI this cycle to spread malicious information about the U.S. One such social media post transcribed a Biden speech inaccurately to suggest he made sexual references.

AI may help election interferers do their dirty work, but new technology is hardly necessary for foreign meddling in U.S. politics.

In 1940, the United Kingdom – an American ally – was so focused on getting the U.S. to enter World War II that British intelligence officers worked to help congressional candidates committed to intervention and to discredit isolationists.

One target was the prominent Republican isolationist U.S. Rep. Hamilton Fish. Circulating a photo of Fish and the leader of an American pro-Nazi group taken out of context, the British sought to falsely paint Fish as a supporter of Nazi elements abroad and in the U.S.
Can AI be controlled?

Acknowledging that it doesn’t take new technology to do harm, bad actors can leverage the efficiencies embedded in AI to create a formidable challenge to election operations and integrity.

Federal efforts to regulate AI’s use in electoral politics face the same uphill battle as most proposals to regulate political campaigns. States have been more active: 19 now ban or restrict deepfakes in political campaigns.

Some platforms engage in light self-moderation. Google’s Gemini responds to prompts asking for basic election information by saying, “I can’t help with responses on elections and political figures right now.”

Campaign professionals may employ a little self-regulation, too. Several speakers at a May 2024 conference on campaign tech expressed concern about pushback from voters if they learn that a campaign is using AI technology. In this sense, the public concern over AI might be productive, creating a guardrail of sorts.

But the flip side of that public concern – what Stanford University’s Nate Persily calls “AI panic” – is that it can further erode trust in elections.

READ ORIGINAL STORY HERE

The Hidden Risk Of Letting AI Decide – Losing The Skills To Choose For Ourselves

BY JOE ARVAI

DANA AND DAVID DORNSIFE

PROFESSOR OF PSYCHOLOGY AND

DIRECTOR OF THE WRIGLEY INSTITUTE FOR

ENVIRONMENTAL AND SUSTAINABILITY,

USC DORNSIFE COLLEGE OF LETTERS,

ARTS AND SCIENCES

As artificial intelligence creeps further into people’s daily lives, so do worries about it. At the most alarmist are concerns about AI going rogue and terminating its human masters.

But behind the calls for a pause on the development of AI is a suite of more tangible social ills. Among them are the risks AI poses to people’s privacy and dignity and the inevitable fact that, because the algorithms under AI’s hood are programmed by humans, it is just as biased and discriminatory as many of us. Throw in the lack of transparency about how AI is designed, and by whom, and it’s easy to understand why so much time these days is devoted to debating its risks as much as its potential.

But my own research as a psychologist who studies how people make decisions leads me to believe that all these risks are overshadowed by an even more corrupting, though largely invisible, threat. That is, AI is mere keystrokes away from making people even less disciplined and skilled when it comes to thoughtful decisions.

Making thoughtful decisions

The process of making thoughtful decisions involves three common sense steps that begin with taking time to understand the task or problem you’re confronted with. Ask yourself, what is it that you need to know, and what do you need to do in order to make a decision that you’ll be able to credibly and confidently defend later?

The answers to these questions hinge on actively seeking out information that both fills gaps in your knowledge and challenges your prior beliefs and assumptions. In fact, it’s this counterfactual information – alternative possibilities that emerge when people unburden themselves of certain assumptions – that ultimately equips you to defend your decisions when they are criticized.

The second step is seeking out and considering more than one option at a time. Want to improve your quality of life? Whether it’s who you vote for, the jobs you accept or the things you buy, there’s always more than one road that will get you there. Expending the effort to actively consider and rate at least a few plausible options, and in a manner that is honest about the trade-offs you are willing to make across their pros and cons, is a hallmark of a thoughtful and defensible choice.

The third step is being willing to delay closure on a decision until after you’ve done all the necessary heavy mental lifting. It’s no secret: Closure feels good because it means you’ve put a difficult or important decision behind you. But the cost of moving on prematurely can be much higher than taking the time to do your homework. If you don’t believe me, just think about all those times you let your feelings guide you, only to experience regret because you didn’t take the time to think a little harder.

Dangers of outsourcing decisions to AI

None of these three steps are terribly difficult to take. But, for most, they’re not intuitive either. Making thoughtful and defensible decisions requires practice and self-discipline. And this is where the hidden harm that AI exposes people to comes in: AI does most of its “thinking” behind the scenes and presents users with answers that are stripped of context and deliberation. Worse, AI robs people of the opportunity to practice the process of making thoughtful and defensible decisions on their own.

Consider how people approach many important decisions today. Humans are well known for being prone to a wide range of biases because we tend to be frugal when it comes to expending mental energy. This frugality leads people to like it when seemingly good or trustworthy decisions are made for them. And we are social animals who tend to value the security and acceptance of their communities more than they might value their own autonomy.

Add AI to the mix and the result is a dangerous feedback loop: The data that AI is mining to fuel its algorithms is made up of people’s biased decisions that also reflect the pressure of conformity instead of the wisdom of critical reasoning. But because people like having decisions made for them, they tend to accept these bad decisions and move on to the next one. In the end, neither we nor AI end up the wiser.

Being thoughtful in the age of AI

It would be wrongheaded to argue that AI won’t offer any benefits to society. It most likely will, especially in fields like cybersecurity, health care and finance, where complex models and massive amounts of data need to be analyzed routinely and quickly. However, most of our day-to-day decisions don’t require this kind of analytic horsepower.

But whether we asked for it or not, many of us have already received advice from – and work performed by – AI in settings ranging from entertainment and travel to schoolwork, health care and finance. And designers are hard at work on next-generation AI that will be able to automate even more of our daily decisions. And this, in my view, is dangerous.

In a world where what and how people think is already under siege thanks to the algorithms of social media, we risk putting ourselves in an even more perilous position if we allow AI to reach a level of sophistication where it can make all kinds of decisions on our behalf. Indeed, we owe it to ourselves to resist the siren’s call of AI and take back ownership of the true privilege – and responsibility – of being human: being able to think and choose for ourselves. We’ll feel better and, importantly, be better if we do.

Silicon Valley And Shenzhen, China, Will Get All The Growth From AI If Other Regions Don’t Invest Now To Compete

BY AMITRAJEET A. BATABYAL

DISTINGUISHED PROFESSOR, ARTHUR J

GOSNELL PROFESSOR OF ECONOMICS,

& INTERIM HEAD, DEPT. OF SUSTAINABILITY

ROCHESTER INSTITUTE OF TECHNOLOGY 

The 21st century has witnessed an unprecedented surge in technological advancements, with artificial intelligence emerging as a worldwide transformative force across the economy. The integration of AI-based technologies into regional economies through the manufacturing and design of goods such as smartphones and smart speakers has sparked significant changes, leading to increased efficiency, innovation and economic growth.

So far, analyses by the urban theorist Richard Florida and others have shown that the economic development driven by AI, like other waves of high technology, are tending to concentrate in specific areas, such as the San Francisco Bay Area and the Washington-to-Boston Northeast Corridor, as well as Shenzhen, often referred to as China’s Silicon Valley.

All are centers of innovation with vibrant tech ecosystems and home to leading global tech companies, such as Google, Apple and many AI startups in the case of Silicon Valley, and Huawei and Tencent in the case of Shenzhen. The ability of AI-based technologies to augment rather than replace human capabilities in such hubs has led to the creation of new job opportunities. This suggests that regions that actively support the development of these technologies are likely to witness a positive relationship between workforce transformation and economic growth.

Technology and creativity

There are two significant points Florida makes about regional growth dynamics related to AI-based technologies and regional growth. First, regions that want to thrive economically need to attract what he has coined the creative class: professionals, including but not limited to university professors, scientists and engineers.

Second, attracting these individuals is important because they possess creative capital, or the ability to create new ideas, technologies, business models, cultural forms and whole new industries that can improve regional economies and lives. This means that creative class members are a basic driver of regional economic growth and development.

How does AI play into this established dynamic of tech-led regional development that produces winners and losers?

As an expert on regional economics, I and my colleagues studied the use of AI-based technologies and regional economic growth. Our analysis sheds light on this critical question by examining just how AI-infused technologies benefit regional economies and those that produce creative goods in the short and long term.

AI and economic growth

We examined a hypothetical region reflective of creative hubs like Silicon Valley, Shenzhen and the Toronto-Waterloo Corridor, focusing on individuals using an AI-based technology to create products such as smartphones, autonomous vehicles and smart speakers. These technologies enhance smartphones with features such as facial recognition, assist in the production of autonomous vehicles through AI-driven design and simulations, and enable smart speakers and personal assistants to understand and respond to user commands via natural language processing and machine learning algorithms.

The use of an AI-based technology permits creative people in a region to enhance the impact that their own creative capital, knowledge and skills have on the production of these goods. Our research shows that an AI-powered regional economy will reach a balanced growth path, or the point at which the productivity of each creative person is positive and stable.

So how do initial differences between creative regions in the use of AI-based technologies affect long-term economic growth? What effects do initial differences in the use of AI-based technologies between, for instance, San Francisco and Seattle have on long-term economic growth in these same cities?

Long-term growth

Consider two regions, A and B – think of A as the San Francisco Bay Area and B as Seattle. A is able to save double the amount that B does investing in an advanced AI-based technology, and A also invests twice as much as B in improving the skills of its creative workforce.

Our research indicates that while A saved twice as much as B on AI and skill development, this small initial gap results in a 32-fold difference in long-term output per creative worker between the two regions. In simple terms, even small differences in savings rates early on can lead to significant gaps in economic output per creative individual over time.

Similarly, our research also demonstrates that even though creative region A saves twice the amount that creative region B does to create a more powerful AI-based technology and skills, this twofold initial difference between the two regions leads to a 64-fold difference in the long-run accumulation of skills per creative person between these same two regions. Once again, the relatively minor initial differences in the two savings rates translate into a greatly magnified impact on the long-run values of skills per creative person.

Some policy lessons

For a given creative region such as the Toronto-Waterloo Corridor in Canada, taking steps now to generate more powerful AI-based technologies is likely to result in substantially magnified benefits in terms of increased output and skills per creative person in the long run.

Second, consider a creative region that is lagging another creative region in terms of output and skills per creative person. For such a region to get ahead, it will need to increase its investment in AI-based technology and skills.

Research shows that AI assets and capabilities in the U.S. are concentrated in San Francisco, San Jose, New York, Los Angeles, Boston and Seattle. Without targeted investments in building and improving AI-based technologies, the present highly skewed nature of AI activity in the U.S. is likely to continue to create large pools of high-skilled workers in some regions while other regions suffer a “brain drain” that leaves lower-skilled workers behind.

This influence is noteworthy, but it is also a double-edged sword. It promises to raise productivity and growth but also to increase the gap between creative regions that make initial investments to improve AI-based technologies and skills – currently coastal regions in the U.S. – and those that do not in the vast American heartland.

READ ORIGINAL STORY HERE