Jul 16 2025
AI’s giants want to take over the classroom

School’s out and it’s high summer, but a bunch of teachers are plotting how they’re going to use AI this upcoming school year. God help them. 

On July 8, OpenAI, Microsoft, and Anthropic announced a $23 million partnership with one of the largest teachers’ unions in the United States to bring more AI into K–12 classrooms. Called the National Academy for AI Instruction, the initiative will train teachers at a New York City headquarters on how to use AI both for teaching and for tasks like planning lessons and writing reports, starting this fall

The companies could face an uphill battle. Right now, most of the public perceives AI’s use in the classroom as nothing short of ruinous—a surefire way to dampen critical thinking and hasten the decline of our collective attention span (a viral story from New York magazine, for example, described how easy it now is to coast through college thanks to constant access to ChatGPT). 

Amid that onslaught, AI companies insist that AI promises more individualized learning, faster and more creative lesson planning, and quicker grading. The companies sponsoring this initiative are, of course, not doing it out of the goodness of their hearts.

No—as they hunt for profits, their goal is to make users out of teachers and students. Anthropic is pitching its AI models to universities, and OpenAI offers free courses for teachers. In an initial training session for teachers by the new National Academy for AI Instruction, representatives from Microsoft showed teachers how to use the company’s AI tools for lesson planning and emails, according to the New York Times. 

It’s early days, but what does the evidence actually say about whether AI is helping or hurting students? There’s at least some data to support the case made by tech companies: A recent survey of 1,500 teens conducted by Harvard’s Graduate School of Education showed that kids are using AI to brainstorm and answer questions they’re afraid to ask in the classroom. Studies examining settings ranging from math classes in Nigeria to colleges physics courses at Harvard have suggested that AI tutors can lead students to become more engaged. 

And yet there’s more to the story. The same Harvard survey revealed that kids are also frequently using AI for cheating and shortcuts. And an oft-cited paper from Microsoft found that relying on AI can reduce critical thinking. Not to mention the fact that “hallucinations” of incorrect information are an inevitable part of how large language models work.

There’s a lack of clear evidence that AI can be a net benefit for students, and it’s hard to trust that the AI companies funding this initiative will give honest advice on when not to use AI in the classroom.

Despite the fanfare around the academy’s launch, and the fact the first teacher training is scheduled to take place in just a few months, OpenAI and Anthropic told me they couldn’t share any specifics. 

It’s not as if teachers themselves aren’t already grappling with how to approach AI. One such teacher, Christopher Harris, who leads a library system covering 22 rural school districts in New York, has created a curriculum aimed at AI literacy. Topics range from privacy when using smart speakers (a lesson for second graders) to misinformation and deepfakes (instruction for high schoolers). I asked him what he’d like to see in the curriculum used by the new National Academy for AI Instruction.

“The real outcome should be teachers that are confident enough in their understanding of how AI works and how it can be used as a tool that they can teach students about the technology as well,” he says. The thing to avoid would be overfocusing on tools and pre-built prompts that teachers are instructed to use without knowing how they work. 

But all this will be for naught without an adjustment to how schools evaluate students in the age of AI, Harris says: “The bigger issue will be shifting the fundamental approaches to how we assign and assess student work in the face of AI cheating.”

The new initiative is led by the American Federation of Teachers, which represents 1.8 million members, as well as the United Federation of Teachers, which represents 200,000 members in New York. If they win over these groups, the tech companies will have significant influence over how millions of teachers learn about AI. But some educators are resisting the use of AI entirely, including several hundred who signed an open letter last week.

Helen Choi is one of them. “I think it is incumbent upon educators to scrutinize the tools that they use in the classroom to look past hype,” says Choi, an associate professor at the University of Southern California, where she teaches writing. “Until we know that something is useful, safe, and ethical, we have a duty to resist mass adoption of tools like large language models that are not designed by educators with education in mind.”

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here.

Ecommerce MGMT 0 Comments
Jul 16 2025
AI text-to-speech programs could “unlearn” how to imitate certain people

A technique known as “machine unlearning” could teach AI models to forget specific voices—an important step in stopping the rise of audio deepfakes, where someone’s voice is copied to carry out fraud or scams.

Recent advances in artificial intelligence have revolutionized the quality of text-to-speech technology so that people can convincingly re-create a piece of text in any voice, complete with natural speaking patterns and intonations, instead of having to settle for a robotic voice reading it out word by word. “Anyone’s voice can be reproduced or copied with just a few seconds of their voice,” says Jong Hwan Ko, a professor at Sungkyunkwan University in Korea and the coauthor of a new paper that demonstrates one of the first applications of machine unlearning to speech generation.

Copied voices have been used in scams, disinformation, and harassment. Ko, who researches audio processing, and his collaborators wanted to prevent this kind of identity fraud. “People are starting to demand ways to opt out of the unknown generation of their voices without consent,” he says. 

AI companies generally keep a tight grip on their models to discourage misuse. For example, if you ask ChatGPT to give you someone’s phone number or instructions for doing something illegal, it will likely just tell you it cannot help. However, as many examples over time have shown, clever prompt engineering or model fine-tuning can sometimes get these models to say things they otherwise wouldn’t. The unwanted information may still be hiding somewhere inside the model so that it can be accessed with the right techniques. 

At present, companies tend to deal with this issue by applying guardrails; the idea is to check whether the prompts or the AI’s responses contain disallowed material. Machine unlearning instead asks whether an AI can be made to forget a piece of information that the company doesn’t want it to know. The technique takes a leaky model and the specific training data to be redacted and uses them to create a new model—essentially, a version of the original that never learned that piece of data. While machine unlearning has ties to older techniques in AI research, it’s only in the past couple of years that it’s been applied to large language models.

Jinju Kim, a master’s student at Sungkyunkwan University who worked on the paper with Ko and others, sees guardrails as fences around the bad data put in place to keep people away from it. “You can’t get through the fence, but some people will still try to go under the fence or over the fence,” says Kim. But unlearning, she says, attempts to remove the bad data altogether, so there is nothing behind the fence at all. 

The way current text-to-speech systems are designed complicates this a little more, though. These so-called “zero-shot” models use examples of people’s speech to learn to re-create any voice, including those not in the training set—with enough data, it can be a good mimic when supplied with even a small sample of someone’s voice. So “unlearning” means a model not only needs to “forget” voices it was trained on but also has to learn not to mimic specific voices it wasn’t trained on. All the while, it still needs to perform well for other voices. 

To demonstrate how to get those results, Kim taught a recreation of VoiceBox, a speech generation model from Meta, that when it was prompted to produce a text sample in one of the voices to be redacted, it should instead respond with a random voice. To make these voices realistic, the model “teaches” itself using random voices of its own creation. 

According to the team’s results, which are to be presented this week at the International Conference on Machine Learning, prompting the model to imitate a voice it has “unlearned” gives back a result that—according to state-of-the-art tools that measure voice similarity—mimics the forgotten voice more than 75% less effectively than the model did before. In practice, this makes the new voice unmistakably different. But the forgetfulness comes at a cost: The model is about 2.8% worse at mimicking permitted voices. While these percentages are a bit hard to interpret, the demo the researchers released online offers very convincing results, both for how well redacted speakers are forgotten and how well the rest are remembered. A sample from the demo is given below. 

A voice sample of a speaker to be forgotten by the model.
The generated text-to-speech audio from the original model using the above as a prompt.
The generated text-to-speech audio using the same prompt, but now from the model where the speaker was forgotten.

Ko says the unlearning process can take “several days,” depending on how many speakers the researchers want the model to forget. Their method also requires an audio clip about five minutes long for each speaker whose voice is to be forgotten.

In machine unlearning, pieces of data are often replaced with randomness so that they can’t be reverse-engineered back to the original. In this paper, the randomness for the forgotten speakers is very high—a sign, the authors claim, that they are truly forgotten by the model. 

 “I have seen people optimizing for randomness in other contexts,” says Vaidehi Patil, a PhD student at the University of North Carolina at Chapel Hill who researches machine unlearning. “This is one of the first works I’ve seen for speech.” Patil is organizing a machine unlearning workshop affiliated with the conference, and the voice unlearning research will also be presented there. 

She points out that unlearning itself involves inherent trade-offs between efficiency and forgetfulness because the process can take time, and can degrade the usability of the final model. “There’s no free lunch. You have to compromise something,” she says.

Machine unlearning may still be at too early a stage for, say, Meta to introduce Ko and Kim’s methods into VoiceBox. But there is likely to be industry interest. Patil is researching unlearning for Google DeepMind this summer, and while Meta did not respond with a comment, it has hesitated for a long time to release VoiceBox to the wider public because it is so vulnerable to misuse. 

The voice unlearning team seems optimistic that its work could someday get good enough for real-life deployment. “In real applications, we would need faster and more scalable solutions,” says Ko. “We are trying to find those.”

Ecommerce MGMT 0 Comments
Jul 16 2025
Google’s generative video model Veo 3 has a subtitles problem

As soon as Google launched its latest video-generating AI model at the end of May, creatives rushed to put it through its paces. Released just months after its predecessor, Veo 3 allows users to generate sounds and dialogue for the first time, sparking a flurry of hyperrealistic eight-second clips stitched together into ads, ASMR videos, imagined film trailers, and humorous street interviews. Academy Award–nominated director Darren Aronofsky used the tool to create a short film called Ancestra. During a press briefing, Demis Hassabis, Google DeepMind’s CEO, likened the leap forward to “emerging from the silent era of video generation.” 

But others quickly found that in some ways the tool wasn’t behaving as expected. When it generates clips that include dialogue, Veo 3 often adds nonsensical, garbled subtitles, even when the prompts it’s been given explicitly ask for no captions or subtitles to be added. 

Getting rid of them isn’t straightforward—or cheap. Users have been forced to resort to regenerating clips (which costs them more money), using external subtitle-removing tools, or cropping their videos to get rid of the subtitles altogether.

Josh Woodward, vice president of Google Labs and Gemini, posted on X on June 9 that Google had developed fixes to reduce the gibberish text. But over a month later, users are still logging issues with it in Google Labs’ Discord channel, demonstrating how difficult it can be to correct issues in major AI models.

Like its predecessors, Veo 3 is available to paying members of Google’s subscription tiers, which start at $249.99 a month. To generate an eight-second clip, users enter a text prompt describing the scene they’d like to create into Google’s AI filmmaking tool Flow, Gemini, or other Google platforms. Each Veo 3 generation costs a minimum of 20 AI credits, and the account can be topped up at a cost of $25 per 2,500 credits.

Mona Weiss, an advertising creative director, says that regenerating her scenes in a bid to get rid of the random captions is becoming expensive. “If you’re creating a scene with dialogue, up to 40% of its output has gibberish subtitles that make it unusable,” she says. “You’re burning through money trying to get a scene you like, but then you can’t even use it.”

When Weiss reported the problem to Google Labs through its Discord channel in the hopes of getting a refund for her wasted credits, its team pointed her to the company’s official support team. They offered her a refund for the cost of Veo 3, but not for the credits. Weiss declined, as accepting would have meant losing access to the model altogether. The Google Labs’ Discord support team has been telling users that subtitles can be triggered by speech, saying that they’re aware of the problem and are working to fix it. 

So why does Veo 3 insist on adding these subtitles, and why does it appear to be so difficult to solve the problem? It probably comes down to what the model has been trained on.  

Although Google hasn’t made this information public, that training data is likely to include YouTube videos, clips from vlogs and gaming channels, and TikTok edits, many of which come with subtitles. These embedded subtitles are part of the video frames rather than separate text tracks layered on top, meaning it’s difficult to remove them before they’re used for training, says Shuo Niu, an assistant professor at Clark University in Massachusetts who studies video sharing platforms and AI.

“The text-to-video model is trained using reinforcement learning to produce content that mimics human-created videos, and if such videos include subtitles, the model may ‘learn’ that incorporating subtitles enhances similarity with human-generated content,” he says.

“We’re continuously working to improve video creation, especially with text, speech that sounds natural, and audio that syncs perfectly,” a Google spokesperson says. “We encourage users to try their prompt again if they notice an inconsistency and give us feedback using the thumbs up/down option.”

As for why the model ignores instructions such as “No subtitles,” negative prompts (telling a generative AI model not to do something) are usually less effective than positive ones, says Tuhin Chakrabarty, an assistant professor at Stony Brook University who studies AI systems. 

To fix the problem, Google would have to check every frame of each video Veo 3 has been trained on, and either get rid of or relabel those with captions before retraining the model—an endeavor that would take weeks, he says. 

Katerina Cizek, a documentary maker and artistic director at the MIT Open Documentary Lab, believes the problem exemplifies Google’s willingness to launch products before they’re fully ready. 

“Google needed a win,” she says. “They needed to be the first to pump out a tool that generates lip-synched audio. And so that was more important than fixing their subtitle issue.”  

Ecommerce MGMT 0 Comments
Jul 15 2025
California is set to become the first US state to manage power outages with AI

California’s statewide power grid operator is poised to become the first in North America to deploy artificial intelligence to manage outages, MIT Technology Review has learned. 

“We wanted to modernize our grid operations. This fits in perfectly with that,” says Gopakumar Gopinathan, a senior advisor on power system technologies at the California Independent System Operator—known as the CAISO and pronounced KAI-so. “AI is already transforming different industries. But we haven’t seen many examples of it being used in our industry.” 

At the DTECH Midwest utility industry summit in Minneapolis on July 15, CAISO is set to announce a deal to run a pilot program using new AI software called Genie, from the energy-services giant OATI. The software uses generative AI to analyze and carry out real-time analyses for grid operators and comes with the potential to autonomously make decisions about key functions on the grid, a switch that might resemble going from uniformed traffic officers to sensor-equipped stoplights. 

But while CAISO may deliver electrons to cutting-edge Silicon Valley companies and laboratories, the actual task of managing the state’s electrical system is surprisingly analog. 

Today, CAISO engineers scan outage reports for keywords about maintenance that’s planned or in the works, read through the notes, and then load each item into the grid software system to run calculations on how a downed line or transformer might affect power supply.

“Even if it takes you less than a minute to scan one on average, when you amplify that over 200 or 300 outages, it adds up,” says Abhimanyu Thakur, OATI’s vice president of platforms, visualization, and analytics. “Then different departments are doing it for their own respective keywords. Now we consolidate all of that into a single dictionary of keywords and AI can do this scan and generate a report proactively.” 

If CAISO finds that Genie produces reliable, more efficient data analyses for managing outages, Gopinathan says, the operator may consider automating more functions on the grid. “After a few rounds of testing, I think we’ll have an idea about what is the right time to call it successful or not,” he says. 

Regardless of the outcome, the experiment marks a significant shift. Most grid operators are using the same systems that utilities have used “for decades,” says Richard Doying, who spent more than 20 years as a top executive at the Midcontinent Independent System Operator, the grid operator for an area encompassing 15 states from the upper Midwest down to Louisiana. 

“These organizations are carved up for people working on very specific, specialized tasks and using their own proprietary tools that they’ve developed over time,” says Doying, now a vice president at the consultancy Grid Strategies. “To the extent that some of these new AI tools are able to draw from data across different areas of an organization and conduct more sophisticated analysis, that’s only helpful for grid operators.”

Last year, a Department of Energy report found that AI had potential to speed up studies on grid capacity and transmission, improve weather forecasting to help predict how much energy wind and solar plants would produce at a given time, and optimize planning for electric-vehicle charging networks. Another report by the energy department’s Loan Programs Office concluded that adding more “advanced” technology such as sensors to various pieces of equipment will generate data that can enable AI to do much more over time. 

In April, the PJM Interconnection—the nation’s largest grid system, spanning 13 states along the densely populated mid-Atlantic and Eastern Seaboard—took a big step toward embracing AI by inking a deal with Google to use its Tapestry software to improve regional planning and speed up grid connections for new power generators. 

ERCOT, the Texas grid system, is considering adopting technology similar to what CAISO is now set to use, according to a source with knowledge of the plans who requested anonymity because they were not authorized to speak publicly. ERCOT did not respond to a request for comment. 

Australia offers an example of what the future may look like. In New South Wales, where grid sensors and smart technology are more widely deployed, AI software rolled out in February is now predicting the production and flow of electricity from rooftop solar units across the state and automatically adjusting how much power from those panels can enter the grid. 

Until now, much of the discussion around AI and energy has focused on the electricity demands of AI data centers (check out MIT Technology Review’s Power Hungry series for more on this).

“We’ve been talking a lot about what the grid can do for AI and not nearly as much about what AI can do for the grid,” says Charles Hua, a coauthor of one of last year’s Energy Department reports who now serves executive director of PowerLines, a nonprofit that advocates for improving the affordability and reliability of US grids. “In general, there’s a huge opportunity for grid operators, regulators, and other stakeholders in the utility regulatory system to use AI effectively and harness it for a more resilient, modernized, and strengthened grid.” 

For now, Gopinathan says, he’s remaining cautiously optimistic. 

“I don’t want to overhype it,” he says. 

Still, he adds, “it’s a first step for bigger automation.”

“Right now, this is more limited to our outage management system. Genie isn’t talking to our other parts yet,” he says. “But I see a world where AI agents are able to do a lot more.”

Ecommerce MGMT 0 Comments
Jul 12 2025
Cybersecurity’s global alarm system is breaking down

Every day, billions of people trust digital systems to run everything from communication to commerce to critical infrastructure. But the global early warning system that alerts security teams to dangerous software flaws is showing critical gaps in coverage—and most users have no idea their digital lives are likely becoming more vulnerable.

Over the past 18 months, two pillars of global cybersecurity have flirted with apparent collapse. In February 2024, the US-backed National Vulnerability Database (NVD)—relied on globally for its free analysis of security threats—abruptly stopped publishing new entries, citing a cryptic “change in interagency support.” Then, in April of this year, the Common Vulnerabilities and Exposures (CVE) program, the fundamental numbering system for tracking software flaws, seemed at similar risk: A leaked letter warned of an imminent contract expiration.

Cybersecurity practitioners have since flooded Discord channels and LinkedIn feeds with emergency posts and memes of “NVD” and “CVE” engraved on tombstones. Unpatched vulnerabilities are the second most common way cyberattackers break in, and they have led to fatal hospital outages and critical infrastructure failures. In a social media post, Jen Easterly, a US cybersecurity expert, said: “Losing [CVE] would be like tearing out the card catalog from every library at once—leaving defenders to sort through chaos while attackers take full advantage.” If CVEs identify each vulnerability like a book in a card catalogue, NVD entries provide the detailed review with context around severity, scope, and exploitability. 

In the end, the Cybersecurity and Infrastructure Security Agency (CISA) extended funding for CVE another year, attributing the incident to a “contract administration issue.” But the NVD’s story has proved more complicated. Its parent organization, the National Institute of Standards and Technology (NIST), reportedly saw its budget cut roughly 12% in 2024, right around the time that CISA pulled its $3.7 million in annual funding for the NVD. Shortly after, as the backlog grew, CISA launched its own “Vulnrichment” program to help address the analysis gap, while promoting a more distributed approach that allows multiple authorized partners to publish enriched data. 

“CISA continuously assesses how to most effectively allocate limited resources to help organizations reduce the risk of newly disclosed vulnerabilities,” says Sandy Radesky, the agency’s associate director for vulnerability management. Rather than just filling the gap, she emphasizes, Vulnrichment was established to provide unique additional information, like recommended actions for specific stakeholders, and to “reduce dependency of the federal government’s role to be the sole provider of vulnerability enrichment.”

Meanwhile, NIST has scrambled to hire contractors to help clear the backlog. Despite a return to pre-crisis processing levels, a boom in vulnerabilities newly disclosed to the NVD has outpaced these efforts. Currently, over 25,000 vulnerabilities await processing—nearly 10 times the previous high in 2017, according to data from the software company Anchore. Before that, the NVD largely kept pace with CVE publications, maintaining a minimal backlog.

“Things have been disruptive, and we’ve been going through times of change across the board,” Matthew Scholl, then chief of the computer security division in NIST’s Information Technology Laboratory, said at an industry event in April. “Leadership has assured me and everyone that NVD is and will continue to be a mission priority for NIST, both in resourcing and capabilities.” Scholl left NIST in May after 20 years at the agency, and NIST declined to comment on the backlog. 

The situation has now prompted multiple government actions, with the Department of Commerce launching an audit of the NVD in May and House Democrats calling for a broader probe of both programs in June. But the damage to trust is already transforming geopolitics and supply chains as security teams prepare for a new era of cyber risk. “It’s left a bad taste, and people are realizing they can’t rely on this,” says Rose Gupta, who builds and runs enterprise vulnerability management programs. “Even if they get everything together tomorrow with a bigger budget, I don’t know that this won’t happen again. So I have to make sure I have other controls in place.”

As these public resources falter, organizations and governments are confronting a critical weakness in our digital infrastructure: Essential global cybersecurity services depend on a complex web of US agency interests and government funding that can be cut or redirected at any time.

Security haves and have-nots

What began as a trickle of software vulnerabilities in the early Internet era has become an unstoppable avalanche, and the free databases that have tracked them for decades have struggled to keep up. In early July, the CVE database crossed over 300,000 catalogued vulnerabilities. Numbers jump unpredictably each year, sometimes by 10% or much more. Even before its latest crisis, the NVD was notorious for delayed publication of new vulnerability analyses, often trailing private security software and vendor advisories by weeks or months.

Gupta has watched organizations increasingly adopt commercial vulnerability management (VM) software that includes its own threat intelligence services. “We’ve definitely become over-reliant on our VM tools,” she says, describing security teams’ growing dependence on vendors like Qualys, Rapid7, and Tenable to supplement or replace unreliable public databases. These platforms combine their own research with various data sources to create proprietary risk scores that help teams prioritize fixes. But not all organizations can afford to fill the NVD’s gap with premium security tools. “Smaller companies and startups, already at a disadvantage, are going to be more at risk,” she explains. 

Komal Rawat, a security engineer in New Delhi whose mid-stage cloud startup has a limited budget, describes the impact in stark terms: “If NVD goes, there will be a crisis in the market. Other databases are not that popular, and to the extent they are adopted, they are not free. If you don’t have recent data, you’re exposed to attackers who do.”

The growing backlog means new devices could be more likely to have vulnerability blind spots—whether that’s a Ring doorbell at home or an office building’s “smart” access control system. The biggest risk may be “one-off” security flaws that fly under the radar. “There are thousands of vulnerabilities that will not affect the majority of enterprises,” says Gupta. “Those are the ones that we’re not getting analysis on, which would leave us at risk.”

NIST acknowledges it has limited visibility into which organizations are most affected by the backlog. “We don’t track which industries use which products and therefore cannot measure impact to specific industries,” a spokesperson says. Instead, the team prioritizes vulnerabilities on the basis of CISA’s known exploits list and those included in vendor advisories like Microsoft Patch Tuesday.

The biggest vulnerability

Brian Martin has watched this system evolve—and deteriorate—from the inside. A former CVE board member and an original project leader behind the Open Source Vulnerability Database, he has built a combative reputation over the decades as a leading historian and practitioner. Martin says his current project, VulnDB (part of Flashpoint Security), outperforms the official databases he once helped oversee. “Our team processes more vulnerabilities, at a much faster turnaround, and we do it for a fraction of the cost,” he says, referring to the tens of millions in government contracts that support the current system. 

When we spoke in May, Martin said his database contains more than 112,000 vulnerabilities with no CVE identifiers—security flaws that exist in the wild but remain invisible to organizations that rely solely on public channels. “If you gave me the money to triple my team, that non-CVE number would be in the 500,000 range,” he said.

In the US, official vulnerability management duties are split between a web of contractors, agencies, and nonprofit centers like the Mitre Corporation. Critics like Martin say that creates potential for redundancy, confusion, and inefficiency, with layers of middle management and relatively few actual vulnerability experts. Others defend the value of this fragmentation. “These programs build on or complement each other to create a more comprehensive, supportive, and diverse community,” CISA said in a statement. “That increases the resilience and usefulness of the entire ecosystem.”

As American leadership wavers, other nations are stepping up. China now operates multiple vulnerability databases, some surprisingly robust but tainted by the possibility that they are subject to state control. In May, the European Union accelerated the launch of its own database, as well as a decentralized “Global CVE” architecture. Following social media and cloud services, vulnerability intelligence has become another front in the contest for technological independence. 

That leaves security professionals to navigate multiple potentially conflicting sources of data. “It’s going to be a mess, but I would rather have too much information than none at all,” says Gupta, describing how her team monitors multiple databases despite the added complexity. 

Resetting software liability

As defenders adapt to the fragmenting landscape, the tech industry faces another reckoning: Why don’t software vendors carry more responsibility for protecting their customers from security issues? Major vendors routinely disclose—but don’t necessarily patch—thousands of new vulnerabilities each year. A single exposure could crash critical systems or increase the risks of fraud and data misuse. 

For decades, the industry has hidden behind legal shields. “Shrink-wrap licenses” once forced consumers to broadly waive their right to hold software vendors liable for defects. Today’s end-user license agreements (EULAs), often delivered in pop-up browser windows, have evolved into incomprehensibly long documents. Last November, a lab project called “EULAS of Despair” used the length of War and Peace (587,287 words) to measure these sprawling contracts. The worst offender? Twitter, at 15.83 novels’ worth of fine print.

“This is a legal fiction that we’ve created around this whole ecosystem, and it’s just not sustainable,” says Andrea Matwyshyn, a US special advisor and technology law professor at Penn State University, where she directs the Policy Innovation Lab of Tomorrow. “Some people point to the fact that software can contain a mix of products and services, creating more complex facts. But just like in engineering or financial litigation, even the most messy scenarios can be resolved with the assistance of experts.”

This liability shield is finally beginning to crack. In July 2024, a faulty security update in CrowdStrike’s popular endpoint detection software crashed millions of Windows computers worldwide and caused outages at everything from airlines to hospitals to 911 systems. The incident led to billions in estimated damages, and the city of Portland, Oregon, even declared a “state of emergency.” Now, affected companies like Delta Airlines have hired high-priced attorneys to pursue major damages—a signal opening of the floodgates to litigation.

Despite the soaring number of vulnerabilities, many fall into long-established categories, such as SQL injections that interfere with database queries and buffer memory overflows that enable code to be executed remotely. Matwyshyn advocates for a mandatory “software bill of materials,” or S-BOM—an ingredients list that would let organizations understand what components and potential vulnerabilities exist throughout their software supply chains. One recent report found 30% of data breaches stemmed from the vulnerabilities of third-party software vendors or cloud service providers.

She adds: “When you can’t tell the difference between the companies that are cutting corners and a company that has really invested in doing right by their customers, that results in a market where everyone loses.”

CISA leadership shares this sentiment, with a spokesperson emphasizing its “secure-by-design principles,” such as “making essential security features available without additional cost, eliminating classes of vulnerabilities, and building products in a way that reduces the cybersecurity burden on customers.”

Avoiding a digital ‘dark age’

It will likely come as no surprise that practitioners are looking to AI to help fill the gap, while at the same time preparing for a coming swarm of cyberattacks by AI agents. Security researchers have used an OpenAI model to discover new “zero-day” vulnerabilities. And both the NVD and CVE teams are developing “AI-powered tools” to help streamline data collection, identification, and processing. NIST says that “up to 65% of our analysis time has been spent generating CPEs”—product information codes that pinpoint affected software. If AI can solve even part of this tedious process, it could dramatically speed up the analysis pipeline.

But Martin cautions against optimism around AI, noting that the technology remains unproven and often riddled with inaccuracies—which, in security, can be fatal. “Rather than AI or ML [machine learning], there are ways to strategically automate bits of the processing of that vulnerability data while ensuring 99.5% accuracy,” he says. 

AI also fails to address more fundamental challenges in governance. The CVE Foundation, launched in April 2025 by breakaway board members, proposes a globally funded nonprofit model similar to that of the internet’s addressing system, which transitioned from US government control to international governance. Other security leaders are pushing to revitalize open-source alternatives like Google’s OSV Project or the NVD++ (maintained by VulnCheck), which are accessible to the public but currently have limited resources.

As these various reform efforts gain momentum, the world is waking up to the fact that vulnerability intelligence—like disease surveillance or aviation safety—requires sustained cooperation and public investment. Without it, a patchwork of paid databases will be all that remains, threatening to leave all but the richest organizations and nations permanently exposed.

Matthew King is a technology and environmental journalist based in New York. He previously worked for cybersecurity firm Tenable.

Ecommerce MGMT 0 Comments
Jul 12 2025
The first babies have been born following “simplified” IVF in a mobile lab

This week I’m sending congratulations to two sets of parents in South Africa. Babies Milayah and Rossouw arrived a few weeks ago. All babies are special, but these two set a new precedent. They’re the first to be born following “simplified” IVF performed in a mobile lab.

This new mobile lab is essentially a trailer crammed with everything an embryologist needs to perform IVF on a shoestring. It was designed to deliver reproductive treatments to people who live in rural parts of low-income countries, where IVF can be prohibitively expensive or even nonexistent. And it seems to work!

While IVF is increasingly commonplace in wealthy countries—around 12% of all births in Spain result from such procedures—it remains expensive and isn’t always covered by insurance or national health providers. And it’s even less accessible in low-income countries—especially for people who live in rural areas.

People often assume that countries with high birth rates don’t need access to fertility treatments, says Gerhard Boshoff, an embryologist at the University of Pretoria in South Africa. Sub-Saharan African countries like Niger, Angola, and Benin all have birth rates above 40 per 1,000 people, which is over four times the rates in Italy and Japan, for example.

But that doesn’t mean people in Sub-Saharan Africa don’t need IVF. Globally, around one in six adults experience infertility at some point in their lives, according to the World Health Organization. Research by the organization suggests that infertility rates are similar in high-income and low-income countries. As the WHO’s director general Tedros Adhanom Ghebreyesus puts it: “Infertility does not discriminate.”

For many people in rural areas of low-income countries, IVF clinics simply don’t exist. South Africa is considered a “reproductive hub” of the African continent, but even in that country there are fewer than 30 clinics for a population of over 60 million. A recent study found there were no such clinics in Angola or Malawi.  

Willem Ombelet, a retired gynecologist, first noticed these disparities back in the 1980s, while he was working at an IVF lab in Pretoria. “I witnessed that infertility was [more prevalent] in the black population than the white population—but they couldn’t access IVF because of apartheid,” he says. The experience spurred him to find ways to make IVF accessible for everyone. In the 1990s, he launched The Walking Egg—a science and art project with that goal.

In 2008, Ombelet met Jonathan Van Blerkom, a reproductive biologist and embryologist who had already been experimenting with a simplified version of IVF. Typically, embryos are cultured in an incubator that provides a sterile mix of gases. Van Blerkom’s approach was to preload tubes with the required gases and seal them with a rubber stopper. “We don’t need a fancy lab,” says Ombelet.

a sleeping infant in a hat and fuzzy sweater
Milayah was born on June 18.
COURTESY OF THE WALKING EGG

Eggs and sperm can be injected into the tubes through the stoppers, and the resulting embryos can be grown inside. All you really need is a good microscope and a way to keep the tube warm, says Ombelet. Once the embryos are around five days old, they can be transferred to a person’s uterus or frozen. “The cost is one tenth or one twentieth of a normal lab,” says Ombelet.

Ombelet, Van Blerkom, and their colleagues found that this approach appeared to work as well as regular IVF. The team ran their first pilot trial at a clinic in Belgium in 2012. The first babies conceived with the simplified IVF process were born later that year.

More recently, Boshoff wondered if the team could take the show on the road. Making IVF simpler and cheaper is one thing, but getting it to people who don’t have access to IVF care is another. What if the team could pack the simplified IVF lab into a trailer and drive it around rural South Africa?

“We just needed to figure out how to have everything in a very confined space,” says Boshoff. As part of the Walking Egg project, he and his colleagues found a way to organize the lab equipment and squeeze in air filters. He then designed a “fold-out system” that allowed the team to create a second room when the trailer was parked. This provides some privacy for people who are having embryos transferred, he says.

People who want to use the mobile IVF lab will first have to undergo treatment at a local medical facility, where they will take drugs that stimulate their ovaries to release eggs, and then have those eggs collected. The rest of the process can be done in the mobile lab, says Boshoff, who presented his work at the European Society of Human Reproduction and Embryology’s annual meeting in Paris earlier this month.

The first trial started last year. The team partnered with one of the few existing fertility clinics in rural South Africa, which put them in touch with 10 willing volunteers. Five of the 10 women got pregnant following their simplified IVF in the mobile lab. One miscarried, but four pregnancies continued. On June 18, baby Milayah arrived. Two days later, another mother welcomed baby Rossouw. The other babies could come any day now.

“We’ve proven that a very cheap and easy [IVF] method can be used even in a mobile unit and have comparable results to regular IVF,” says Ombelet, who says his team is planning similar trials in Egypt and Indonesia. “The next step is to roll it out all over the world.”

This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.

Ecommerce MGMT 0 Comments
Jul 11 2025
This tool strips away anti-AI protections from digital art

A new technique called LightShed will make it harder for artists to use existing protective tools to stop their work from being ingested for AI training. It’s the next step in a cat-and-mouse game—across technology, law, and culture—that has been going on between artists and AI proponents for years. 

Generative AI models that create images need to be trained on a wide variety of visual material, and data sets that are used for this training allegedly include copyrighted art without permission. This has worried artists, who are concerned that the models will learn their style, mimic their work, and put them out of a job.

These artists got some potential defenses in 2023, when researchers created tools like Glaze and Nightshade to protect artwork by “poisoning” it against AI training (Shawn Shan was even named MIT Technology Review’s Innovator of the Year last year for his work on these). LightShed, however, claims to be able to subvert these tools and others like them, making it easy for the artwork to be used for training once again.

To be clear, the researchers behind LightShed aren’t trying to steal artists’ work. They just don’t want people to get a false sense of security. “You will not be sure if companies have methods to delete these poisons but will never tell you,” says Hanna Foerster, a PhD student at the University of Cambridge and the lead author of a paper on the work. And if they do, it may be too late to fix the problem.

AI models work, in part, by implicitly creating boundaries between what they perceive as different categories of images. Glaze and Nightshade change enough pixels to push a given piece of art over this boundary without affecting the image’s quality, causing the model to see it as something it’s not. These almost imperceptible changes are called perturbations, and they mess up the AI model’s ability to understand the artwork.

Glaze makes models misunderstand style (e.g., interpreting a photorealistic painting as a cartoon). Nightshade instead makes the model see the subject incorrectly (e.g., interpreting a cat in a drawing as a dog). Glaze is used to defend an artist’s individual style, whereas Nightshade is used to attack AI models that crawl the internet for art.

Foerster worked with a team of researchers from the Technical University of Darmstadt and the University of Texas at San Antonio to develop LightShed, which learns how to see where tools like Glaze and Nightshade splash this sort of digital poison onto art so that it can effectively clean it off. The group will present its findings at the Usenix Security Symposium, a leading global cybersecurity conference, in August. 

The researchers trained LightShed by feeding it pieces of art with and without Nightshade, Glaze, and other similar programs applied. Foerster describes the process as teaching LightShed to reconstruct “just the poison on poisoned images.” Identifying a cutoff for how much poison will actually confuse an AI makes it easier to “wash” just the poison off. 

LightShed is incredibly effective at this. While other researchers have found simple ways to subvert poisoning, LightShed appears to be more adaptable. It can even apply what it’s learned from one anti-AI tool—say, Nightshade—to others like Mist or MetaCloak without ever seeing them ahead of time. While it has some trouble performing against small doses of poison, those are less likely to kill the AI models’ abilities to understand the underlying art, making it a win-win for the AI—or a lose-lose for the artists using these tools.

Around 7.5 million people, many of them artists with small and medium-size followings and fewer resources, have downloaded Glaze to protect their art. Those using tools like Glaze see it as an important technical line of defense, especially when the state of regulation around AI training and copyright is still up in the air. The LightShed authors see their work as a warning that tools like Glaze are not permanent solutions. “It might need a few more rounds of trying to come up with better ideas for protection,” says Foerster.

The creators of Glaze and Nightshade seem to agree with that sentiment: The website for Nightshade warned the tool wasn’t future-proof before work on LightShed ever began. And Shan, who led research on both tools, still believes defenses like his have meaning even if there are ways around them. 

“It’s a deterrent,” says Shan—a way to warn AI companies that artists are serious about their concerns. The goal, as he puts it, is to put up as many roadblocks as possible so that AI companies find it easier to just work with artists. He believes that “most artists kind of understand this is a temporary solution,” but that creating those obstacles against the unwanted use of their work is still valuable.

Foerster hopes to use what she learned through LightShed to build new defenses for artists, including clever watermarks that somehow persist with the artwork even after it’s gone through an AI model. While she doesn’t believe this will protect a work against AI forever, she thinks this could help tip the scales back in the artist’s favor once again.

Ecommerce MGMT 0 Comments
Jul 11 2025
China’s energy dominance in three charts

China is the dominant force in next-generation energy technologies today. It’s pouring hundreds of billions of dollars into putting renewable sources like wind and solar on its grid, manufacturing millions of electric vehicles, and building out capacity for energy storage, nuclear power, and more. This investment has been transformational for the country’s economy and has contributed to establishing China as a major player in global politics. 

Meanwhile, in the US, a massive new tax and spending bill just cut hundreds of billions in credits, grants, and loans for clean energy technologies. It’s a stark reversal from previous policies, and it could have massive effects at a time when it feels as if everyone is chasing China on energy.

So while we all try to get our heads around what’s next for climate tech in the US and beyond, let’s look at just how dominant China is when it comes to clean energy, as documented in three charts.

China is on an absolute tear installing wind and solar power. The country reached nearly 900 gigawatts of installed capacity for solar at the end of 2024, and the rapid pace of building has continued into this year. An additional 198 GW was installed between January and May, with 93 GW coming in May alone. 

For context, those additions over the first five months of the year account for more than double the capacity of the grid in California. Not the renewables capacity of that state—the entire grid. 

Meanwhile, the policy shift in the US is projected to slow down new solar and wind additions. With tax credits and other support stripped away, much of the new capacity that was expected to come online by the end of the decade will now face delays or cancellations. 

That’s significant because of all the new electricity generation capacity that’s come online in the US recently, renewables make up the vast majority. Solar and battery storage alone are expected to make up over 80% of capacity additions in 2025. So slowing down wind and solar basically means slowing down adding new electricity capacity, at a time when demand is very much set to rise. (Hello, AI?)

China’s EV market is also booming—the country is currently flirting with a big symbolic milestone, nearing the point where over half of all new vehicles sold in the country are electric. (It already passed that mark for a single month and could do so on a yearly basis in the next couple of years.)

It’s not just selling those vehicles within China, either: the country exports them globally, with customers including established markets like Europe and growing ones like India and Brazil. As of 2024, more than 70% of electric and plug-in hybrid vehicles on roads around the world were built in ChinaSome leaders in legacy automakers are taking notice. Ford CEO Jim Farley shared some striking comments at the Aspen Ideas Festival last month about how far ahead China is on vehicle technology and price. “They have far superior in-vehicle technology,” Farley said. “We are in a global competition with China, and it’s not just EVs. And if we lose this, we do not have a future Ford.” 

Looking ahead, China is still pouring money into renewables, storage, grids, and energy efficiency technologies. It’s also outspending the rest of the world on nuclear power. The country tripled its investment in renewable power from 2015 to 2025.

The situation isn’t set in stone, though: The US actually very briefly overtook China on battery investments over the past year, as Cat Clifford at Cipher reported last week. But changes resulting from the new bill could very quickly reverse that progress, cementing China as the place for battery manufacturing and innovation.

In a story earlier this week, the MIT economist David Autor laid out the high stakes for this race. Advanced manufacturing and technology are beneficial for US prosperity, and putting public support and trade protections in place for key industries could be crucial to keeping them going, he says.  

I’d add that this whole discussion shouldn’t be about a zero-sum competition between the US and China. But many experts argue that the US, where I and many readers live, is surrendering its leadership and ability to develop key energy technologies of the future.  

Ultimately, the numbers don’t lie: By a lot of measures, China is the world’s leader in energy. The question is, will that change anytime soon?  

This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.

Ecommerce MGMT 0 Comments
Jul 10 2025
Why the AI moratorium’s defeat may signal a new political era

The “Big, Beautiful Bill” that President Donald Trump signed into law on July 4 was chock full of controversial policies—Medicaid work requirements, increased funding for ICE, and an end to tax credits for clean energy and vehicles, to name just a few. But one highly contested provision was missing. Just days earlier, during a late-night voting session, the Senate had killed the bill’s 10-year moratorium on state-level AI regulation. 

“We really dodged a bullet,” says Scott Wiener, a California state senator and the author of SB 1047, a bill that would have made companies liable for harms caused by large AI models. It was vetoed by Governor Gavin Newsom last year, but Wiener is now working to pass SB 53, which establishes whistleblower protections for employees of AI companies. Had the federal AI regulation moratorium passed, he says, that bill likely would have been dead.

The moratorium could also have killed laws that have already been adopted around the country, including a Colorado law that targets algorithmic discrimination, laws in Utah and California aimed at making AI-generated content more identifiable, and other legislation focused on preserving data privacy and keeping children safe online. Proponents of the moratorium, such OpenAI and Senator Ted Cruz, have said that a “patchwork” of state-level regulations would place an undue burden on technology companies and stymie innovation. Federal regulation, they argue, is a better approach—but there is currently no federal AI regulation in place.

Wiener and other state lawmakers can now get back to work writing and passing AI policy, at least for the time being—with the tailwind of a major moral victory at their backs. The movement to defeat the moratorium was impressively bipartisan: 40 state attorneys general signed a letter to Congress opposing the measure, as did a group of over 250 Republican and Democratic state lawmakers. And while congressional Democrats were united against the moratorium, the final nail in its coffin was hammered in by Senator Marsha Blackburn of Tennessee, a Tea Party conservative and Trump ally who backed out of a compromise with Cruz at the eleventh hour.

The moratorium fight may have signaled a bigger political shift. “In the last few months, we’ve seen a much broader and more diverse coalition form in support of AI regulation generally,” says Amba Kak, co–executive director of the AI Now Institute. After years of relative inaction, politicians are getting concerned about the risks of unregulated artificial intelligence. 

Granted, there’s an argument to be made that the moratorium’s defeat was highly contingent. Blackburn appears to have been motivated almost entirely by concerns about children’s online safety and the rights of country musicians to control their own likenesses; state lawmakers, meanwhile, were affronted by the federal government’s attempt to defang legislation that they had already passed.

And even though powerful technology firms such as Andreessen Horowitz and OpenAI reportedly lobbied in favor of the moratorium, continuing to push for it might not have been worth it to the Trump administration and its allies—at least not at the expense of tax breaks and entitlement cuts. Baobao Zhang, an associate professor of political science at Syracuse University, says that the administration may have been willing to give up on the moratorium in order to push through the rest of the bill by its self-imposed Independence Day deadline.

Andreessen Horowitz did not respond to a request for comment. OpenAI noted that the company was opposed to a state-by-state approach to AI regulation but did not respond to specific questions regarding the moratorium’s defeat. 

It’s almost certainly the case that the moratorium’s breadth, as well as its decade-long duration, helped opponents marshall a diverse coalition to their side. But that breadth isn’t incidental—it’s related to the very nature of AI. Blackburn, who represents country musicians in Nashville, and Wiener, who represents software developers in San Francisco, have a shared interest in AI regulation precisely because such a powerful and general-purpose tool has the potential to affect so many people’s well-being and livelihood. “There are real anxieties that are touching people of all classes,” Kak says. “It’s creating solidarities that maybe didn’t exist before.”

Faced with outspoken advocates, concerned constituents, and the constant buzz of AI discourse, politicians from both sides of the aisle are starting to argue for taking AI extremely seriously. One of the most prominent anti-moratorium voices was Marjorie Taylor Greene, who voted for the version of the bill containing the moratorium before admitting that she hadn’t read it thoroughly and committing to opposing the moratorium moving forward. “We have no idea what AI will be capable of in the next 10 years,” she posted last month.

And two weeks ago, Pete Buttigieg, President Biden’s transportation secretary, published a Substack post entitled “We Are Still Underreacting on AI.” “The terms of what it is like to be a human are about to change in ways that rival the transformations of the Enlightenment or the Industrial Revolution, only much more quickly,” he wrote.

Wiener has noticed a shift among his peers. “More and more policymakers understand that we can’t just ignore this,” he says. But awareness is several steps short of effective legislation, and regulation opponents aren’t giving up the fight. The Trump administration is reportedly working on a slate of executive actions aimed at making more energy available for AI training and deployment, and Cruz says he is planning to introduce his own anti-regulation bill.

Meanwhile, proponents of regulation will need to figure out how to channel the broad opposition to the moratorium into support for specific policies. It won’t be a simple task. “It’s easy for all of us to agree on what we don’t want,” Kak says. “The harder question is: What is it that we do want?”

Ecommerce MGMT 0 Comments
Jul 10 2025
Inside OpenAI’s empire: A conversation with Karen Hao

In a wide-ranging Roundtables conversation for MIT Technology Review subscribers, AI journalist and author Karen Hao spoke about her new book, Empire of AI: Dreams and Nightmares in Sam Altman’s OpenAI. She talked with executive editor Niall Firth about how she first covered the company in 2020 while on staff at MIT Technology Review, and they discussed how the AI industry now functions like an empire and what ethically-made AI looks like. 

Read the transcript of the conversation, which has been lightly edited and condensed, below. Subscribers can watch the on-demand recording of the event here. 

Niall Firth: Hello, everyone, and welcome to this special edition of Roundtables. These are our subscriber-only events where you get to listen in to conversations between editors and reporters. Now, I’m delighted to say we’ve got an absolute cracker of an event today. I’m very happy to have our prodigal daughter, Karen Hao, a fabulous AI journalist, here with us to talk about her new book. Hello, Karen, how are you doing?

Karen Hao: Good. Thank you so much for having me back, Niall. 

Niall Firth: Lovely to have you. So I’m sure you all know Karen and that’s why you’re here. But to give you a quick, quick synopsis, Karen has a degree in mechanical engineering from MIT. She was MIT Technology Review’s senior editor for AI and has won countless awards, been cited in Congress, written for the Wall Street Journal and The Atlantic, and set up a series at the Pulitzer Center to teach journalists how to cover AI. 

But most important of all, she’s here to discuss her new book, which I’ve got a copy of here, Empire of AI. The UK version is subtitled “Inside the reckless race for total domination,” and the US one, I believe, is “Dreams and nightmares in Sam Altman’s OpenAI.”

It’s been an absolute sensation, a New York Times chart topper. An incredible feat of reporting—like 300 interviews, including 90 with people inside OpenAI. And it’s a brilliant look at not just OpenAI’s rise, and the character of Sam Altman, which is very interesting in its own right, but also a really astute look at what kind of AI we’re building and who holds the keys. 

Karen, the core of the book, the rise and rise of OpenAI, was one of your first big features at MIT Technology Review. It’s a brilliant story that lifted the lid for the first time on what was going on at OpenAI … and they really hated it, right?

Karen Hao: Yes, and first of all, thank you to everyone for being here. It’s always great to be home. I do still consider MIT Tech Review to be my journalistic home, and that story was—I only did it because Niall assigned it after I said, “Hey, it seems like OpenAI is kind of an interesting thing,” and he was like, you should profile them. And I had never written a profile about a company before, and I didn’t think that I would have it in me, and Niall believed that I would be able to do it. So it really didn’t happen other than because of you.

I went into the piece with an open mind about—let me understand what OpenAI is. Let me take what they say at face value. They were founded as a nonprofit. They have this mission to ensure artificial general intelligence benefits all of humanity. What do they mean by that? How are they trying to achieve that ultimately? How are they striking this balance between mission-driven AI development and the need to raise money and capital? 

And through the course of embedding within the company for three days, and then interviewing dozens of people outside the company or around the company … I came to realize that there was a fundamental disconnect between what they were publicly espousing and accumulating a lot of goodwill from and how they were operating. And that is what I ended up focusing my profile on, and that is why they were not very pleased.

Niall Firth: And how have you seen OpenAI change even since you did the profile? That sort of misalignment feels like it’s got messier and more confusing in the years since.

Karen Hao: Absolutely. I mean, it’s kind of remarkable that OpenAI, you could argue that they are now one of the most capitalistic corporations in Silicon Valley. They just raised $40 billion, in the largest-ever private fundraising round in tech industry history. They’re valued at $300 billion. And yet they still say that they are first and foremost a nonprofit. 

I think this really gets to the heart of how much OpenAI has tried to position and reposition itself throughout its decade-long history, to ultimately play into the narratives that they think are going to do best with the public and with policymakers, in spite of what they might actually be doing in terms of developing their technologies and commercializing them.

Niall Firth: You cite Sam Altman saying, you know, the race for AGI is what motivated a lot of this, and I’ll come back to that a bit before the end. But he talks about it as like the Manhattan Project for AI. You cite him quoting Oppenheimer (of course, you know, there’s no self-aggrandizing there): “Technology happens because it’s possible,” he says in the book. 

And it feels to me like this is one of the themes of the book: the idea that technology doesn’t just happen because it comes along. It comes because of choices that people make. It’s not an inevitability that things are the way they are and that people are who they are. What they think is important—that influences the direction of travel. So what does this mean, in practice, if that’s the case?

Karen Hao: With OpenAI in particular, they made a very key decision early on in their history that led to all of the AI technologies that we see dominating the marketplace and dominating headlines today. And that was a decision to try and advance AI progress through scaling the existing techniques that were available to them. At the time when OpenAI started, at the end of 2015, and then, when they made that decision, in roughly around 2017, this was a very unpopular perspective within the broader AI research field. 

There were kind of two competing ideas about how to advance AI progress, or rather a spectrum of ideas, bookended by two extremes. One extreme being, we have all the techniques we need, and we should just aggressively scale. And the other one being that we don’t actually have the techniques we need. We need to continue innovating and doing fundamental AI research to get more breakthroughs. And largely the field assumed that this side of the spectrum [focusing on fundamental AI research] was the most likely approach for getting advancements, but OpenAI was anomalously committed to the other extreme—this idea that we can just take neural networks and pump ever more data, and train on ever larger supercomputers, larger than have ever been built in history.

The reason why they made that decision was because they were competing against Google, which had a dominant monopoly on AI talent. And OpenAI knew that they didn’t necessarily have the ability to beat Google simply by trying to get research breakthroughs. That’s a very hard path. When you’re doing fundamental research, you never really know when the breakthrough might appear. It’s not a very linear line of progress, but scaling is sort of linear. As long as you just pump more data and more compute, you can get gains. And so they thought, we can just do this faster than anyone else. And that’s the way that we’re going to leap ahead of Google. And it particularly aligned with Sam Altman’s skillset, as well, because he is a once-in-a-generation fundraising talent, and when you’re going for scale to advance AI models, the primary bottleneck is capital.

And so it was kind of a great fit for what he had to offer, which is, he knows how to accumulate capital, and he knows how to accumulate it very quickly. So that is ultimately how you can see that technology is a product of human choices and human perspectives. And they’re the specific skills and strengths that that team had at the time for how they wanted to move forward.

Niall Firth: And to be fair, I mean, it works, right? It was amazing, fabulous. You know the breakthroughs that happened, GPT-2 to GPT-3, just from scale and data and compute, kind of were mind-blowing really, as we look back on it now.

Karen Hao: Yeah, it is remarkable how much it did work, because there was a lot of skepticism about the idea that scale could lead to the kind of technical progress that we’ve seen. But one of my biggest critiques of this particular approach is that there’s also an extraordinary amount of costs that come with this particular pathway to getting more advancements. And there are many different pathways to advancing AI, so we could have actually gotten all of these benefits, and moving forward, we could continue to get more benefits from AI, without actually engaging in a hugely consumptive, hugely costly approach to its development.

Niall Firth: Yeah, so in terms of consumptive, that’s something we’ve touched on here quite recently at MIT Technology Review, like the energy costs of AI. The data center costs are absolutely extraordinary, right? Like the data behind it is incredible. And it’s only gonna get worse in the next few years if we continue down this path, right? 

Karen Hao: Yeah … so first of all, everyone should read the series that Tech Review put out, if you haven’t already, on the energy question, because it really does break down everything from what is the energy consumption of the smallest unit of interacting with these models, all the way up until the highest level. 

The number that I have seen a lot, and that I’ve been repeating, is there was a McKinsey report that was looking at if we continue to just look at the pace at which data centers and supercomputers are being built and scaled, in the next five years, we would have to add two to six times the amount of energy consumed by California onto the grid. And most of that will have to be serviced by fossil fuels, because these data centers and supercomputers have to run 24/7, so we cannot rely solely on renewable energy. We do not have enough nuclear power capacity to power these colossal pieces of infrastructure. And so we’re already accelerating the climate crisis. 

And we’re also accelerating a public-health crisis, the pumping of thousands of tons of air pollutants into the air from coal plants that are having their lives extended and methane gas turbines that are being built in service of powering these data centers. And in addition to that, there’s also an acceleration of the freshwater crisis, because these pieces of infrastructure have to be cooled with freshwater resources. It has to be fresh water, because if it’s any other type of water, it corrodes the equipment, it leads to bacterial growth.

And Bloomberg recently had a story that showed that two-thirds of these data centers are actually going into water-scarce areas, into places where the communities already do not have enough fresh water at their disposal. So that is one dimension of many that I refer to when I say, the extraordinary costs of this particular pathway for AI development.

Niall Firth: So in terms of costs and the extractive process of making AI, I wanted to give you the chance to talk about the other theme of the book, apart from just OpenAI’s explosion. It’s the colonial way of looking at the way AI is made: the empire. I’m saying this obviously because we’re here, but this is an idea that came out of reporting you started at MIT Technology Review and then continued into the book. Tell us about how this framing helps us understand how AI is made now.

Karen Hao: Yeah, so this was a framing that I started thinking a lot about when I was working on the AI Colonialism series for Tech Review. It was a series of stories that looked at the way that, pre-ChatGPT, the commercialization of AI and its deployment into the world was already leading to entrenchment of historical inequities into the present day.

And one example was a story that was about how facial recognition companies were swarming into South Africa to try and harvest more data from South Africa during a time when they were getting criticized for the fact that their technologies did not accurately recognize black faces. And the deployment of those facial recognition technologies into South Africa, into the streets of Johannesburg, was leading to what South African scholars were calling a recreation of a digital apartheid—the controlling of black bodies, movement of black people.

And this idea really haunted me for a really long time. Through my reporting in that series, there were so many examples that I kept hitting upon of this thesis, that the AI industry was perpetuating. It felt like it was becoming this neocolonial force. And then, when ChatGPT came out, it became clear that this was just accelerating. 

When you accelerate the scale of these technologies, and you start training them on the entirety of the Internet, and you start using these supercomputers that are the size of dozens—if not hundreds—of football fields. Then you really start talking about an extraordinary global level of extraction and exploitation that is happening to produce these technologies. And then the historical power imbalances become even more obvious. 

And so there are four parallels that I draw in my book between what I have now termed empires of AI versus empires of old. The first one is that empires lay claim to resources that are not their own. So these companies are scraping all this data that is not their own, taking all the intellectual property that is not their own.

The second is that empires exploit a lot of labor. So we see them moving to countries in the Global South or other economically vulnerable communities to contract workers to do some of the worst work in the development pipeline for producing these technologies—and also producing technologies that then inherently are labor-automating and engage in labor exploitation in and of themselves. 

And the third feature is that the empires monopolize knowledge production. So, in the last 10 years, we’ve seen the AI industry monopolize more and more of the AI researchers in the world. So AI researchers are no longer contributing to open science, working in universities or independent institutions, and the effect on the research is what you would imagine would happen if most of the climate scientists in the world were being bankrolled by oil and gas companies. You would not be getting a clear picture, and we are not getting a clear picture, of the limitations of these technologies, or if there are better ways to develop these technologies.

And the fourth and final feature is that empires always engage in this aggressive race rhetoric, where there are good empires and evil empires. And they, the good empire, have to be strong enough to beat back the evil empire, and that is why they should have unfettered license to consume all of these resources and exploit all of this labor. And if the evil empire gets the technology first, humanity goes to hell. But if the good empire gets the technology first, they’ll civilize the world, and humanity gets to go to heaven. So on many different levels, like the empire theme, I felt like it was the most comprehensive way to name exactly how these companies operate, and exactly what their impacts are on the world.

Niall Firth: Yeah, brilliant. I mean, you talk about the evil empire. What happens if the evil empire gets it first? And what I mentioned at the top is AGI. For me, it’s almost like the extra character in the book all the way through. It’s sort of looming over everything, like the ghost at the feast, sort of saying like, this is the thing that motivates everything at OpenAI. This is the thing we’ve got to get to before anyone else gets to it. 

There’s a bit in the book about how they’re talking internally at OpenAI, like, we’ve got to make sure that AGI is in US hands where it’s safe versus like anywhere else. And some of the international staff are openly like—that’s kind of a weird way to frame it, isn’t it? Why is the US version of AGI better than others? 

So tell us a bit about how it drives what they do. And AGI isn’t an inevitable fact that’s just happening anyway, is it? It’s not even a thing yet.

Karen Hao: There’s not even consensus around whether or not it’s even possible or what it even is. There was recently a New York Times story by Cade Metz that was citing a survey of long-standing AI researchers in the field, and 75% of them still think that we don’t have the techniques yet for reaching AGI, whatever that means. And the most classic definition or understanding of what AGI is, is being able to fully recreate human intelligence in software. But the problem is, we also don’t have scientific consensus around what human intelligence is. And so one of the aspects that I talk about a lot in the book is that, when there is a vacuum of shared meaning around this term, and what it would look like, when would we have arrived at it? What capabilities should we be evaluating these systems on to determine that we’ve gotten there? It can basically just be whatever OpenAI wants. 

So it’s kind of just this ever-present goalpost that keeps shifting, depending on where the company wants to go. You know, they have a full range, a variety of different definitions that they’ve used throughout the years. In fact, they even have a joke internally: If you ask 13 OpenAI researchers what AGI is, you’ll get 15 definitions. So they are kind of self-aware that this is not really a real term and it doesn’t really have that much meaning. 

But it does serve this purpose of creating a kind of quasi-religious fervor around what they’re doing, where people think that they have to keep driving towards this horizon, and that one day when they get there, it’s going to have a civilizationally transformative impact. And therefore, what else should you be working on in your life, but this? And who else should be working on it, but you? 

And so it is their justification not just for continuing to push and scale and consume all these resources—because none of that consumption, none of that harm matters anymore if you end up hitting this destination. But they also use it as a way to develop their technologies in a very deeply anti-democratic way, where they say, we are the only people that have the expertise, that have the right to carefully control the development of this technology and usher it into the world. And we cannot let anyone else participate because it’s just too powerful of a technology.

Niall Firth: You talk about the factions, particularly the religious framing. AGI has been around as a concept for a while—it was very niche, very kind of nerdy fun, really, to talk about—to suddenly become extremely mainstream. And they have the boomers versus doomers dichotomy. Where are you on that spectrum?

Karen Hao: So the boomers are people who think that AGI is going to bring us to utopia, and the doomers think AGI is going to devastate all of humanity. And to me these are actually two sides of the same coin. They both believe that AGI is possible, and it’s imminent, and it’s going to change everything. 

And I am not on this spectrum. I’m in a third space, which is the AI accountability space, which is rooted in the observation that these companies have accumulated an extraordinary amount of power, both economic and political power, to go back to the empire analogy. 

Ultimately, the thing that we need to do in order to not return to an age of empire and erode a lot of democratic norms is to hold these companies accountable with all the tools at our disposal, and to recognize all the harms that they are already perpetuating through a misguided approach to AI development.

Niall Firth: I’ve got a couple of questions from readers. I’m gonna try to pull them together a little bit because Abbas asks, what would post-imperial AI look like? And there was a question from Liam basically along the same lines. How do you make a more ethical version of AI that is not within this framework? 

Karen Hao: We sort of already touched a little bit upon this idea. But there are so many different ways to develop AI. There are myriads of techniques throughout the history of AI development, which is decades long. There have been various shifts in the winds of which techniques ultimately rise and fall. And it isn’t based solely on the scientific or technical merit of any particular technique. Oftentimes certain techniques become more popular because of business reasons or because of the funder’s ideologies. And that’s sort of what we’re seeing today with the complete indexing of AI development on large-scale AI model development.

And ultimately, these large-scale models … We talked about how it’s a remarkable technical leap, but in terms of social progress or economic progress, the benefits of these models have been kind of middling. And the way that I see us shifting to AI models that are going to be A) more beneficial and B) not so imperial is to refocus on task-specific AI systems that are tackling well-scoped challenges that inherently lend themselves to the strengths of AI systems that are inherently computational optimization problems. 

So I’m talking about things like using AI to integrate more renewable energy into the grid. This is something that we definitely need. We need to more quickly accelerate our electrification of the grid, and one of the challenges of using more renewable energy is the unpredictability of it. And this is a key strength of AI technologies, being able to have predictive capabilities and optimization capabilities where you can match the energy generation of different renewables with the energy demands of different people that are drawing from the grid.

Niall Firth: Quite a few people have been asking, in the chat, different versions of the same question. If you were an early-career AI scientist, or if you were involved in AI, what can you do yourself to bring about a more ethical version of AI? Do you have any power left, or is it too late? 

Karen Hao: No, I don’t think it’s too late at all. I mean, as I’ve been talking with a lot of people just in the lay public, one of the biggest challenges that they have is they don’t have any alternatives for AI. They want the benefits of AI, but they also do not want to participate in a supply chain that is really harmful. And so the first question is, always, is there an alternative? Which tools do I shift to? And unfortunately, there just aren’t that many alternatives right now. 

And so the first thing that I would say to early-career AI researchers and entrepreneurs is to build those alternatives, because there are plenty of people that are actually really excited about the possibility of switching to more ethical alternatives. And one of the analogies I often use is that we kind of need to do with the AI industry what happened with the fashion industry. There was also a lot of environmental exploitation, labor exploitation in the fashion industry, and there was enough consumer demand that it created new markets for ethical and sustainably sourced fashion. And so we kind of need to see just more options occupying that space.

Niall Firth: Do you feel optimistic about the future? Or where do you sit? You know, things aren’t great as you spell them out now. Where’s the hope for us?

Karen Hao: I am. I’m super optimistic. Part of the reason why I’m optimistic is because you know, a few years ago, when I started writing about AI at Tech Review, I remember people would say, wow, that’s a really niche beat. Do you have enough to write about? 

And now, I mean, everyone is talking about AI, and I think that’s the first step to actually getting to a better place with AI development. The amount of public awareness and attention and scrutiny that is now going into how we develop these technologies, how we use these technologies, is really, really important. Like, we need to be having this public debate and that in and of itself is a significant step change from what we had before. 

But the next step, and part of the reason why I wrote this book, is we need to convert the awareness into action, and people should take an active role. Every single person should feel that they have an active role in shaping the future of AI development, if you think about all of the different ways that you interface with the AI development supply chain and deployment supply chain—like you give your data or withhold your data.

There are probably data centers that are being built around you right now. If you’re a parent, there’s some kind of AI policy being crafted at [your kid’s] school. There’s some kind of AI policy being crafted at your workplace. These are all what I consider sites of democratic contestation, where you can use those opportunities to assert your voice about how you want AI to be developed and deployed. If you do not want these companies to use certain kinds of data, push back when they just take the data. 

I closed all of my personal social media accounts because I just did not like the fact that they were scraping my personal photos to train their generative AI models. I’ve seen parents and students and teachers start forming committees within schools to talk about what their AI policy should be and to draft it collectively as a community. Same with businesses. They’re doing the same thing. If we all kind of step up to play that active role, I am super optimistic that we’ll get to a better place.

Niall Firth: Mark, in the chat, mentions the Māori story from New Zealand towards the end of your book, and that’s an example of sort of community-led AI in action, isn’t it?

Karen Hao: Yeah. There was a community in New Zealand that really wanted to help revitalize the Māori language by building a speech recognition tool that could recognize Māori, and therefore be able to transcribe a rich repository of archival audio of their ancestors speaking Māori. And the first thing that they did when engaging in that project was they asked the community, do you want this AI tool? 

Niall Firth: Imagine that.

Karen Hao: I know! It’s such a radical concept, this idea of consent at every stage. But they first asked that; the community wholeheartedly said yes. They then engaged in a public education campaign to explain to people, okay, what does it take to develop an AI tool? Well, we are going to need data. We’re going to need audio transcription pairs to train this AI model. So then they ran a public contest in which they were able to get dozens, if not hundreds, of people in their community to donate data to this project. And then they made sure that when they developed the model, they actively explained to the community at every step how their data was being used, how it would be stored, how it would continue to be protected. And any other project that would use the data has to get permission and consent from the community first. 

And so it was a completely democratic process, for whether they wanted the tool, how to develop the tool, and how the tool should continue to be used, and how their data should continue to be used over time.

Niall Firth: Great. I know we’ve gone a bit over time. I’ve got two more things I’m going to ask you, basically putting together lots of questions people have asked in the chat about your view on what role regulations should play. What are your thoughts on that?

Karen Hao: Yeah, I mean, in an ideal world where we actually had a functioning government, regulation should absolutely play a huge role. And it shouldn’t just be thinking about once an AI model is built, how to regulate that. But still thinking about the full supply chain of AI development, regulating the data and what’s allowed to be trained in these models, regulating the land use. And what pieces of land are allowed to build data centers? How much energy and water are the data centers allowed to consume? And also regulating the transparency. We don’t know what data is in these training data sets, and we don’t know the environmental costs of training these models. We don’t know how much water these data centers consume and that is all information that these companies actively withhold to prevent democratic processes from happening. So if there were one major intervention that regulators could have, it should be to dramatically increase the amount of transparency along the supply chain.

Niall Firth: Okay, great. So just to bring it back around to OpenAI and Sam Altman to finish with. He famously sent an email around, didn’t he? After your original Tech Review story, saying this is not great. We don’t like this. And he didn’t want to speak to you for your book, either, did he?

Karen Hao: No, he did not.

Niall Firth: No. But imagine Sam Altman is in the chat here. He’s subscribed to Technology Review and is watching this Roundtables because he wants to know what you’re saying about him. If you could talk to him directly, what would you like to ask him? 

Karen Hao: What degree of harm do you need to see in order to realize that you should take a different path? 

Niall Firth: Nice, blunt, to the point. All right, Karen, thank you so much for your time. 

Karen Hao: Thank you so much, everyone.

MIT Technology Review Roundtables is a subscriber-only online event series where experts discuss the latest developments and what’s next in emerging technologies. Sign up to get notified about upcoming sessions.

Ecommerce MGMT 0 Comments