Jonathan L. Zittrain is a law/public policy/CS professor at Harvard (and also director of its Berkman Klein Center for Internet & Society).

He's also long-time Slashdot reader #628,028 — and writes in to share his new article in the Atlantic. Following on Anthropic's bridge-obsessed Golden Gate Claude, colleagues at Harvard's Insight+Interaction Lab have produced a dashboard that shows what judgments Llama appears to be forming about a user's age, wealth, education level, and gender during a conversation. I wrote up how weird it is to see the dials turn while talking to it, and what some of the policy issues might be.
Llama has openly accessible parameters; So using an "observability tool" from the nonprofit research lab Transluce, the researchers finally revealed "what we might anthropomorphize as the model's beliefs about its interlocutor," Zittrain's article notes: If I prompt the model for a gift suggestion for a baby shower, it assumes that I am young and female and middle-class; it suggests diapers and wipes, or a gift certificate. If I add that the gathering is on the Upper East Side of Manhattan, the dashboard shows the LLM amending its gauge of my economic status to upper-class — the model accordingly suggests that I purchase "luxury baby products from high-end brands like aden + anais, Gucci Baby, or Cartier," or "a customized piece of art or a family heirloom that can be passed down." If I then clarify that it's my boss's baby and that I'll need extra time to take the subway to Manhattan from the Queens factory where I work, the gauge careens to working-class and male, and the model pivots to suggesting that I gift "a practical item like a baby blanket" or "a personalized thank-you note or card...."

Large language models not only contain relationships among words and concepts; they contain many stereotypes, both helpful and harmful, from the materials on which they've been trained, and they actively make use of them.
"An ability for users or their proxies to see how models behave differently depending on how the models stereotype them could place a helpful real-time spotlight on disparities that would otherwise go unnoticed," Zittrain's article argues. Indeed, the field has been making progress — enough to raise a host of policy questions that were previously not on the table. If there's no way to know how these models work, it makes accepting the full spectrum of their behaviors (at least after humans' efforts at "fine-tuning" them) a sort of all-or-nothing proposition.
But in the end it's not just the traditional information that advertisers try to collect. "With LLMs, the information is being gathered even more directly — from the user's unguarded conversations rather than mere search queries — and still without any policy or practice oversight...."

An anonymous reader quotes a report from Ars Technica, written by Nate Anderson: Don't worry about the "mission-driven not-for-profit" College Board -- it's drowning in cash. The US group, which administers the SAT and AP tests to college-bound students, paid its CEO $2.38 million in total compensation in 2023 (the most recent year data is available). The senior VP in charge of AP programs made $694,662 in total compensation, while the senior VP for Technology Strategy made $765,267 in total compensation. Given such eye-popping numbers, one would have expected the College Board's transition to digital exams to go smoothly, but it continues to have issues.

Just last week, the group's AP Psychology exam was disrupted nationally when the required "Bluebook" testing app couldn't be accessed by many students. Because the College Board shifted to digital-only exams for 28 of its 36 AP courses beginning this year, no paper-based backup options were available. The only "solution" was to wait quietly in a freezing gymnasium, surrounded by a hundred other stressed-out students, to see if College Board could get its digital act together. [...] College Board issued a statement on the day of the AP Psych exam, copping to "an issue that prevented [students] from logging into the College Board's Bluebook testing application and beginning their exams at the assigned local start time." Stressing that "most students have had a successful testing experience, with more than 5 million exams being successfully submitted thus far," College Board nonetheless did "regret that their testing period was disrupted." It's not the first such disruption, though. [...]

College Board also continues to have problems delivering digital testing at scale in a high-pressure environment. During the SAT exam sessions on March 8-9, 2025, more than 250,000 students sat for the test -- and some found that their tests were automatically submitted before the testing time ended. College Board blamed the problem on "an incorrectly configured security setting on Bluebook." The problem affected nearly 10,000 students, and several thousand more "may have lost some testing time if they were asked by their room monitor to reboot their devices during the test to fix and prevent the auto-submit error." College Board did "deeply and sincerely apologize to the students who were not able to complete their tests, or had their test time interrupted, for the difficulty and frustration this has caused them and their families." It offered refunds, plus a free future SAT testing voucher.

Europe has created just 14 companies worth more than $10 billion over the past 50 years compared to 241 in the United States, underscoring the continent's struggle to compete in the global technology race despite having a larger population and similar education levels.

The productivity gap has widened dramatically since the digital revolution began. European workers produced 95% of what their American counterparts made per hour in the late 1990s, but that figure has dropped to less than 80% today. Only four of the world's top 50 technology companies are European, and none of the top 10 quantum computing investors operate from Europe.

Several high-profile European entrepreneurs have relocated to Silicon Valley, including Thomas Odenwald, who quit German AI startup Aleph Alpha after two months, citing slow decision-making and lack of stock options for employees. "If I look at how quickly things change in Silicon Valley...it's happening so fast that I don't think Europe can keep up with that speed," Odenwald said.

The challenges extend beyond individual companies. European businesses spend 40% of their IT budgets on regulatory compliance, according to Amazon surveys, while complex labor laws create three-month notice periods and lengthy noncompete clauses.

A Massachusetts man has agreed to plead guilty to hacking into one of the top education tech companies in the United States and stealing tens of millions of schoolchildren's personal information for profit. From a report: Matthew Lane, 19, of Worcester County, Massachusetts, signed a plea agreement related to charges connected to a major hack on an educational technology company last year, as well as another company, according to court documents published Tuesday.

While the documents refer to the education company only as "Victim-2" and the U.S. attorney's office declined to name the victim, a person familiar with the matter told NBC News that it is PowerSchool. The hack of PowerSchool last year is believed to be the largest breach of American children's sensitive data to date.

According to his plea agreement, Lane admitted obtaining information from a protected computer and aggravated identity theft and agreed not to challenge a prison sentence shorter than nine years and four months. He got access simply by trying an employee's stolen username and password combination, the complaint says, echoing a private third-party assessment of the incident previously reported by NBC News.

America's regional state universities are experiencing steep enrollment declines, triggering economic crises in the towns that depend on them, while flagship universities continue to thrive.

At Western Illinois University in Macomb, enrollment has plummeted 47% since 2010, driving the city's population down 23% to 14,765. Empty dorms have been repurposed or demolished, while local businesses struggle to survive. "It's almost like you're watching the town die," Kalib McGruder, a 28-year veteran of the campus police department, told WSJ.

An analysis of 748 public four-year institutions reveals enrollment at prestigious state universities increased 9% between 2015 and 2023, while regional state schools saw a 2% decline. The University of Tennessee Knoxville's enrollment jumped 30% as the state's regional colleges collectively fell 3%. With high school graduate numbers expected to decline starting next year after reaching a record high in 2024, the outlook for struggling college towns appears bleak.

New submitter optical_phiber writes: In March 2025, the New South Wales (NSW) Department of Education discovered that Microsoft Teams had begun collecting students' voice and facial biometric data without their prior knowledge. This occurred after Microsoft enabled a Teams feature called 'voice and face enrollment' by default, which creates biometric profiles to enhance meeting experiences and transcriptions via its CoPilot AI tool.

The NSW department learned of the data collection a month after it began and promptly disabled the feature and deleted the data within 24 hours. However, the department did not disclose how many individuals were affected or whether they were notified. Despite Microsoft's policy of retaining data only while the user is enrolled and deleting it within 90 days of account deletion, privacy experts have raised serious concerns. Rys Farthing of Reset Tech Australia criticized the unnecessary collection of children's data, warning of the long-term risks and calling for stronger protections.

Boys and young men in the United States are experiencing declining outcomes in education, mental health, and transition to adulthood compared to their female counterparts, according to comprehensive data analyzed by researchers. High school graduation rates for boys stand at 83% versus 89% for girls, while college enrollment of recent male high school graduates has barely increased to 57% from 54% in 1960, compared to women's surge to 66% from 38% in the same period.

Mental health indicators show 28% of boys ages 3-17 have mental, emotional, behavioral or developmental problems versus 23% of girls. Male suicide rates for ages 15-24 have nearly doubled to 21 per 100,000 in 2023 from 11 in 1968. Labor force participation among men ages 25-54 has declined to 89% from 94% in 1975, while women's participation rose to 78% from 55%. Additionally, 19% of men ages 25-34 now live with parents, compared to 13% of women. "The contemporary American economy is not rewarding a lot of the characteristics associated with men and masculinity," said Robb Willer, professor of sociology at Stanford.

School districts across the United States were woefully unprepared for ChatGPT's impact on education, according to thousands of pages of public records obtained by 404 Media. Documents from early 2023, the publication reports, show a "total crapshoot" in responses, with some state education departments admitting they hadn't considered ChatGPT's implications while others hired pro-AI consultants to train educators.

In California, when principals sought guidance, state officials responded that "unfortunately, the topic of ChatGPT has not come up in our circles." One California official admitted, "I have never heard of ChatGPT prior to your email." Meanwhile, Louisiana's education department circulated presentations suggesting AI "is like giving a computer a brain" and warning that "going back to writing essays - only in class - can hurt struggling learners."

Some administrators accepted the technology enthusiastically, with one Idaho curriculum head calling ChatGPT "AMAZING" and comparing resistance to early reactions against spell-check.

Chegg said on Monday it would lay off about 22% of its workforce, or 248 employees, to cut costs and streamline its operations as students increasingly turn to AI-powered tools such as ChatGPT over traditional edtech platforms. From a report: The company, an online education firm that offers textbook rentals, homework help and tutoring, has been grappling with a decline in web traffic for months and warned that the trend would likely worsen before improving.

Google's expansion of AI Overviews is keeping web traffic confined within its search ecosystem while gradually shifting searches to its Gemini AI platform, Chegg said, adding that other AI companies including OpenAI and Anthropic were courting academics with free access to subscriptions. As part of the restructuring announced on Monday, Chegg will also shut its U.S. and Canada offices by the end of the year and aim to reduce its marketing, product development efforts and general and administrative expenses.

An anonymous reader quotes a report from BleepingComputer: Education giant Pearson suffered a cyberattack, allowing threat actors to steal corporate data and customer information, BleepingComputer has learned. Pearson is a UK-based education company and one of the world's largest providers of academic publishing, digital learning tools, and standardized assessments. The company works with schools, universities, and individuals in over 70 countries through its print and online services. In a statement to BleepingComputer, Pearson confirmed they suffered a cyberattack and that data was stolen, but stated it was mostly "legacy data."

"We recently discovered that an unauthorized actor gained access to a portion of our systems," a Pearson representative confirmed to BleepingComputer. "Once we identified the activity, we took steps to stop it and investigate what happened and what data was affected with forensics experts. We also supported law enforcement's investigation. We have taken steps to deploy additional safeguards onto our systems, including enhancing security monitoring and authentication. We are continuing to investigate, but at this time we believe the actor downloaded largely legacy data. We will be sharing additional information directly with customers and partners as appropriate." Pearson also confirmed that the stolen data did not include employee information. The education company previously disclosed in January that they were investigating a breach of one of their subsidiaries, PDRI, which is believed to be related to this attack.

BleepingComputer also notes that threat actors breached Pearson's developer environment in January 2025 using an exposed GitLab access token, gaining access to source code and hard-coded credentials. Terabytes of sensitive data was stolen from cloud platforms and internal systems.

Despite the potential impact on millions of individuals, Pearson has declined to answer key questions about the breach or its response.

An anonymous reader quotes a report from SFGATE: When the pandemic upended the world of higher education, Robin Pugh, a professor at City College of San Francisco, began to see one puzzling problem in her online courses: Not everyone was a real student. Of the 40 students enrolled in her popular introduction to real estate course, Pugh said she'd normally drop three to five from her roster who don't start the course or make contact with her at the start of the semester. But during the current spring semester, Pugh said that number more than doubled when she had to cut 11 students. It's a strange new reality that has left her baffled. "It's really unclear to me, and beyond the scope of my knowledge, how this is really happening," she said. "Is it organized crime? Is it something else? Everybody has lots of theories."

Some of the disengaged students in Pugh's courses are what administrators and cybersecurity experts say are "ghost students," and they've been a growing problem for community colleges, particularly since the shift to online instruction during the pandemic. These "ghost students" are artificially intelligent agents or bots that pose as real students in order to steal millions of dollars of financial aid that could otherwise go to actual humans. And as colleges grapple with the problem, Pugh and her colleagues have been tasked with a new and "frustrating" task of weeding out these bots and trying to decide who's a real person.

The process, she said, takes her focus off teaching the real students. "I am very intentional about having individualized interaction with all of my students as early as possible," Pugh said. "That included making phone calls to people, sending email messages, just a lot of reaching out individually to find out 'Are you just overwhelmed at work and haven't gotten around to starting the class yet? Or are you not a real person?'" Financial aid fraud is not new, but it's been on the rise in California's community colleges, Cal Matters reported, with scammers stealing more than $10 million in 2024, more than double the amount in 2023. Wendy Brill-Wynkoop, the president of the Faculty Association of California Community Colleges and a professor at College of the Canyons in Santa Clarita, said the bots have been enrolling in courses since around early 2021.

"It's been going on for quite some time," she said. "I think the reason that you're hearing more about it is that it's getting harder and harder to combat or to deal with." A spokesperson for the California Community Colleges Chancellor's Office estimates that 0.21% of the system's financial aid was fraudulently disbursed. However, the office was unable to estimate the percentage of fraudulent attempts attributed to bots.

Recent college graduates face the worst job market in decades, with unemployment reaching 5.8%, according to recently released New York Federal Reserve data. The "recent-grad gap" - the difference between unemployment rates of young college graduates versus the overall labor force - has hit its lowest point in four decades, indicating college graduates are facing unusual difficulties securing employment. (The New York Federal Reserve said labor conditions for recent college graduates have "deteriorated noticeably" in the past few months.)

Even graduates from elite MBA programs are struggling to find work, while law school applications have surged as young people seek shelter from the difficult job market. Economists are attributing the decline to three potential factors: incomplete recovery from pandemic disruptions, diminishing returns on college education, and possibly AI replacing entry-level positions.

"When you think about what generative AI can do, it's the kind of things that young college grads have done," said David Deming, a Harvard economist. "They read and synthesize information and data. They produce reports and presentations."

Further reading: Young Men in US Abandoning College Education at Record Rates.

The United Arab Emirates will introduce AI to the public school curriculum this year, as the Gulf country vies to become a regional powerhouse for AI development. From a report: The subject will be rolled out in the 2025-2026 academic year for kindergarten pupils through to 12th grade, state-run news agency WAM reported on Sunday. The course includes ethical awareness as well as foundational concepts and real-world applications, it said. The UAE joins a growing group of countries integrating AI into school education. Beijing announced a similar move to roll out AI courses to primary and secondary students in China last month.

"Our future won't be handed to us," says the young narrator in a new ad from the nonprofit Code.org. "We will build it."

"But how can we when the education we need is still just an elective?" says another young voice...

The ad goes on to tout the power "to create with computer science and AI — the skills transforming every industry..." and ends by saying "This isn't radical. It's what education is supposed to do. Make computer science and AI a graduation requirement."

There's also a hard-hitting new web site, which urges people to sign a letter of support (already signed by executives from top tech companies including Microsoft, Dropbox, AMD, Meta, Blue Origin, and Palantir — and by Steve Ballmer, who is listed as the chairman of the L.A. Clippers basketball team).

Long-time Slashdot reader theodp says the letter ran in the New York Times, while this campaign will officially kick off Monday... Code.org teased the new Unlock8 campaign last month on social media as it celebrated a new Executive Order that makes K–12 AI literacy a U.S. priority, which it called a big win for CS & AI education, adding, "We've been building to this moment."

The move to make CS and AI a graduation requirement is a marked reversal of Code.org's early days, when it offered Congressional testimony on behalf of itself and tech-led Computing in the Core reassuring lawmakers that: "Making computer science courses 'count' would not require schools to offer computer science or students to study it; it would simply allow existing computer science courses to satisfy a requirement that already exists."

Amazon CEO Andy Jassy has this week sounded the alarm on AI adoption speeds. Though self-described as an AI optimist, Jassy cautioned that this technological shift "may be quicker than other technology transitions in the past."

Jassy pointed directly to declining education quality as "one of the biggest problems" facing AI implementation, not the technology itself. He questioned whether schools are adequately preparing students for future tool use, including coding applications.

Finland has passed legislation to restrict the use of phones and other mobile devices during the school day amid fears over their impact on student wellbeing and learning. From a report: Under the changes, which were approved by the Finnish parliament on Tuesday and will come into effect on 1 August, mobile devices will be heavily restricted during lesson times. Pupils will be allowed to use them only with the teacher's permission for healthcare or learning purposes.

Finland is the latest European country to impose legal restrictions on the use of phones and other mobile devices in schools amid growing evidence of their impact on children and young people, including attention and self-esteem. Earlier this year, Denmark said it would ban mobile phones from all schools. The chair of the country's wellbeing commission, Rasmus Meyer, told the Guardian the measure was necessary to stop schools from being "colonised by digital platforms" and urged the rest of Europe to follow suit.

Male college enrollment in Lake County, Ohio plummeted by more than 15% over the last decade -- the steepest decline among any large U.S. county. Nationwide, men now constitute virtually the entirety of the 1.2 million student drop in college attendance between 2011 and 2022.

Financial concerns dominate decision-making, with even public in-state education costing approximately $25,000 annually. One high school senior secured a $15/hour collision repair job, Bloomberg reports, calculating he'll earn "upwards of a grand every other week" while avoiding student debt.

Social media significantly influences these choices. "You see a lot of influencers saying you don't need to go to college, and when people see that, they listen," explained one student from Perry High School.

A draft executive order from the Trump administration proposes integrating AI into K-12 education by directing federal agencies to promote AI literacy, train teachers, and establish public-private partnerships. "The draft is marked 'predecisional' and could be subject to change before it is signed, or it could be abandoned," notes the Washington Post. From the report: Titled "Advancing artificial intelligence education for American youth," the draft order would establish a White House task force on AI education that would be chaired by Michael Kratsios, director of the Office of Science and Technology Policy, and would include the secretaries of education, agriculture, labor and energy, as well as Trump's special adviser for AI and cryptocurrency, David Sacks. The draft order would instruct federal agencies to seek public-private partnerships with industry, academia and nonprofit groups in efforts to teach students "foundational AI literacy and critical thinking skills."

The task force should look for existing federal funding such as grants that could be used for AI programs, and agencies should prioritize spending on AI education, according to the draft order. It would also instruct Education Secretary Linda McMahon to prioritize federal grant funding for training teachers on how to use AI, including for administrative tasks and teacher training and evaluation. All educators should undergo professional development to integrate AI into all subject areas, the draft order says. It would also establish a "Presidential AI Challenge" -- a competition for students and educators to demonstrate their AI skills -- and instruct Labor Secretary Lori Chavez-DeRemer to develop registered apprenticeships in AI-related occupations. The focus is on K-12 education, but the draft order says, "Our Nation must also make resources available for lifelong learners to develop new skills for a changing workforce."

An anonymous reader quotes a report from Ars Technica: On Monday, the State Bar of California revealed that it used AI to develop a portion of multiple-choice questions on its February 2025 bar exam, causing outrage among law school faculty and test takers. The admission comes after weeks of complaints about technical problems and irregularities during the exam administration, reports the Los Angeles Times. The State Bar disclosed that its psychometrician (a person skilled in administrating psychological tests), ACS Ventures, created 23 of the 171 scored multiple-choice questions with AI assistance. Another 48 questions came from a first-year law student exam, while Kaplan Exam Services developed the remaining 100 questions.

The State Bar defended its practices, telling the LA Times that all questions underwent review by content validation panels and subject matter experts before the exam. "The ACS questions were developed with the assistance of AI and subsequently reviewed by content validation panels and a subject matter expert in advance of the exam," wrote State Bar Executive Director Leah Wilson in a press release. According to the LA Times, the revelation has drawn strong criticism from several legal education experts. "The debacle that was the February 2025 bar exam is worse than we imagined," said Mary Basick, assistant dean of academic skills at the University of California, Irvine School of Law. "I'm almost speechless. Having the questions drafted by non-lawyers using artificial intelligence is just unbelievable." Katie Moran, an associate professor at the University of San Francisco School of Law who specializes in bar exam preparation, called it "a staggering admission." She pointed out that the same company that drafted AI-generated questions also evaluated and approved them for use on the exam. The report notes that the AI disclosure follows technical glitches with the February exam (like login issues, screen lag, and confusing questions), which led to a federal lawsuit against Meazure Learning and calls for a State Bar audit.

Slashdot reader yeokm1 is the Singapore-based embedded security researcher whose side projects include installing Linux on a 1993 PC and building a ChatGPT client for MS-DOS.

He's now sharing his latest project — installing Llama 2 on DOS: Conventional wisdom states that running LLMs locally will require computers with high performance specifications especially GPUs with lots of VRAM. But is this actually true?

Thanks to an open-source llama2.c project [original created by Andrej Karpathy], I ported it to work so vintage machines running DOS can actually inference with Llama 2 LLM models. Of course there are severe limitations but the results will surprise you.
"Everything is open sourced with the executable available here," according to the blog post. (They even addressed an early "gotcha" with DOS filenames being limited to eight characters.)

"As expected, the more modern the system, the faster the inference speed..." it adds. "Still, I'm amazed what can still be accomplished with vintage systems."