Connect with us

Science

AI-powered text from this program could fool the government

Published

on

Elena Lacey | Getty Images

In October 2019, Idaho proposed changing its Medicaid program. The state needed approval from the federal government, which solicited public feedback via Medicaid.gov.

Roughly 1,000 comments arrived. But half came not from concerned citizens or even Internet trolls. They were generated by artificial intelligence. And a study found that people could not distinguish the real comments from the fake ones.

The project was the work of Max Weiss, a tech-savvy medical student at Harvard, but it received little attention at the time. Now, with AI language systems advancing rapidly, some say the government and Internet companies need to rethink how they solicit and screen feedback to guard against deepfake text manipulation and other AI-powered interference.

“The ease with which a bot can generate and submit relevant text that impersonates human speech on government websites is surprising and really important to know,” says Latanya Sweeney, a professor at Harvard’s Kennedy School who advised Weiss on how to run the experiment ethically.

Sweeney says the problems extend well beyond government services, but it is imperative that public agencies find a solution. “AI can drown speech from real humans,” she says. “Government websites have to change.”

The Centers for Medicare and Medicaid Services says it has added new safeguards to the public comment system in response to Weiss’ study, though it declines to discuss specifics. Weiss says he was contacted by the US General Services Administration, which is developing a new version of the federal government website for publishing regulations and comments, about ways to better protect it from fake comments.

Government systems have been the target of automated influence campaigns before. In 2017, researchers discovered that over a million comments submitted to the Federal Communications Commission regarding plans to roll back net neutrality rules had been auto-generated, with certain phrases copied and pasted into different messages.

“I was a bit shocked when I saw nothing more than a submit button standing in the way of your comment becoming a part of the public record.”

Weiss’ project highlights a more serious threat. There has been remarkable progress in applying AI to language over the past few years. When powerful machine-learning algorithms are fed huge amounts of training data—in the form of books and text scraped from the Web—they can produce programs capable of generating convincing text. Besides myriad useful applications, this raises the prospect that all sorts of Internet messages, comments, and posts could be faked easily and less detectably.

“As technology gets better,” Sweeney says, “human speech venues become subject to manipulation without human knowledge that it has happened.” Weiss was working at a health care consumer-advocacy organization in the summer of 2019 when he learned about the public feedback process required to make Medicaid changes. Knowing that these public comments had swayed previous efforts to change state Medicaid programs, Weiss looked for tools that could auto-generate comments.

“I was a bit shocked when I saw nothing more than a submit button standing in the way of your comment becoming a part of the public record,” he says.

Weiss discovered GPT-2, a program released earlier that year by OpenAI, an AI company in San Francisco, and realized he could generate fake comments to simulate a groundswell of public opinion. “I was also shocked at how easy it was to fine tune GPT-2 to actually spit out the comments,” Weiss says. “It’s relatively concerning on a number of fronts.”

Besides the comment-generating tool, Weiss built software for automatically submitting comments. He also conducted an experiment in which volunteers were asked to distinguish between the AI-generated comments and ones written by humans. The volunteers did no better than random guessing.

After submitting the comments, Weiss notified the Centers for Medicare and Medicaid Services. He had added a few characters to make it easy to identify each fake comment. Even so, he says, the AI feedback remained posted online for several months.

GPT-3

OpenAI released a more capable version of its text-generation program, called GPT-3, last June. So far, it has only been made available to a few AI researchers and companies, with some people building useful applications such as programs that generate email messages from bullet points. When GPT-3 was released, OpenAI said in a research paper that it had not seen signs of GPT-2 being used maliciously, even though it had been aware of Weiss’ research.

OpenAI and other researchers have released a few tools capable of identifying AI-generated text. These use similar AI algorithms to spot telltale signs in the text. It’s not clear if anyone is using these to protect online commenting platforms. Facebook declined to say if it is using such tools; Google and Twitter did not respond to requests for comment.

It also isn’t clear if sophisticated AI tools are yet being used to create fake content. In August, researchers at Google posted details of an experiment that used deepfake-text-detection tools to analyze over 500 million webpages. They found that the tools could identify pages hosting auto-generated text and spam. But it wasn’t clear if any of the content was made using an AI tool such as GPT-2.

Renée DiResta, research manager at the Stanford Internet Observatory, which tracks online abuse, says she expects more government sites to be targeted by fake text. “Any time you have new technology, it’s a tool in the hands of some and a weapon in the hands of others,” she says.

Politically driven misinformation has become a critical issue in American politics. Joan Donovan, research director of the Shorenstein Center on Media, Politics and Public Policy at the Harvard Kennedy School, warns that sophisticated AI may not be needed to erode people’s sense of what’s true. “People’s emotions are frayed, and that makes them very vulnerable to convenient explanations rather than difficult truths,” Donovan says.

This story originally appeared on wired.com.

Continue Reading

Science

Biden says US will halve carbon emissions by 2030

Published

on

Getty Images | Bloomberg

President Joe Biden announced Thursday that the US would cut carbon emissions by 50 to 52 percent by 2030 compared with 2005 levels. It’s an aggressive target that aims to put the country on a path to limit warming to below 1.5˚C.

The pledge comes as Biden hosts dozens of world leaders for a virtual climate summit today. It nearly doubles the targets set by former President Barack Obama when the Paris Agreement was initially signed.

“This is a moral imperative, an economic imperative. A moment of peril but also a moment of extraordinary possibilities,” Biden said.

The US reduced its carbon emissions to 13 percent below 2005 levels in 2019, the most recent “normal” year for energy use. While the number represents a notable decline given that the economy grew in that time, it’s still short of where the country should be if it is to achieve either Obama’s or Biden’s goals.

The US isn’t the only country that has come up short in terms of actual reductions. The world as a whole is far from meeting the goals set forth in the Paris Agreement. We’re currently on track for 3˚C of warming by 2100, an amount that would result in catastrophic consequences for people around the globe.

Emissions targets pledged as part of the Paris Agreement aren’t binding, but they do offer roadmaps for countries to decarbonize. The agreement also encourages countries to make their pledges more ambitious over time, with the hopes that emissions will decline more rapidly as technological advancements and diplomatic pressure make significant reductions more appealing.

Other leaders used the forum to announce new commitments as well. Prime Minister Yoshihide Suga of Japan said his country would aim for a reduction of 46 percent below 2013 levels, almost double its previous commitment. Canada’s Prime Minister, Justin Trudeau, increased his country’s pledge to 40 to 45 percent below 2005 levels by 2030, a bump above the previous 30 percent commitment. The European Union and the UK didn’t update their pledges, but both have already committed to major cuts over the next nine years, with the EU targeting 55 percent and the UK shooting for 68 percent.

China’s Xi Jinping also reiterated his country’s previous commitment to halt the growth of emissions by 2030 and reach net zero by 2060. China’s carbon pollution has grown precipitously over the last 20 years, to the point where the country now produces more than twice as much as the next largest polluter, the US. India, another country with swiftly rising emissions, did not put forth any targets, nor did it add to its previous commitments. Instead, Prime Minister Narendra Modi restated his country’s promise to install 450 GW of renewable energy capacity by 2030.

Globally, carbon emissions dropped 6.4 percent last year as the world hunkered down to weather the pandemic. The US contributed the most to that decline with a 13 percent drop, and most of that happened in the transportation sector, which produces the largest share of the nation’s carbon pollution. Globally, the decline in air travel allowed the aviation sector to halve its emissions.

The Biden administration again signaled that it’s considering a “border carbon adjustment”—basically a carbon tariff. Such a tariff would apply to imports from countries where carbon regulations aren’t as stringent.

For many countries, the energy transition has become less of a burden and more of an opportunity to leapfrog laggards by developing early leads in a host of promising and increasingly affordable clean technologies, including solar, wind, and batteries. Biden appears to agree with that idea and is hoping to position the US among countries taking the lead. “The countries that take decisive action now to create the industries of the future will be the ones who reap the economic benefits of the clean energy boom that’s coming,” he said.

Continue Reading

Science

Contractor that ruined 15M doses of J&J vaccine hiked price of another by 800%

Published

on

Enlarge / The Emergent BioSolutions plant, a manufacturing partner for Johnson & Johnson’s COVID-19 vaccine, in Baltimore, Maryland, on April 9, 2021.

Things are not looking good for Emergent BioSolutions, the contract manufacturer that ruined 15 million doses of Johnson & Johnson’s one-shot COVID-19 vaccine and millions more doses of AstraZeneca’s COVID-19 vaccine at its production facility in Baltimore.

The Food and Drug Administration on Wednesday released a searing inspection report of the facility, finding a slew of significant violations and failings.

Meanwhile, federal lawmakers have opened a multi-pronged investigation into whether Emergent used ties to the Trump administration to get billions of dollars in federal contracts despite a history of failing to complete contracts, inadequately training staff, persistent quality-control issues, and an “unjustified” 800% price increase for an anthrax vaccine.

In a letter sent to Emergent’s top executives Tuesday, Rep. Carolyn Maloney, chairwoman of the House Committee on Oversight and Reform, and Rep. James Clyburn, chairman of the Select Subcommittee on the Coronavirus Crisis, laid out the investigation, writing:

Emergent received $628 million in June 2020 to establish the primary US facility for manufacturing vaccines developed by Johnson & Johnson and AstraZeneca. Dr. Robert Kadlec, who served as Assistant Secretary for Preparedness and Response under President Trump and previously worked as a consultant for Emergent, appears to have pushed for this award despite indications that Emergent did not have the ability to reliably fulfill the contract.

800% drain

But the investigation stretches back much farther than the start of the pandemic—through years of questionable federal contracts.

In 1998, Emergent (then called BioPort) bought the license to an anthrax vaccine. The vaccine, BioThrax, was approved by the FDA in 1970 but remains the only FDA-approved vaccine for anthrax. As such, Emergent is the sole supplier of anthrax vaccine for the federal government’s Strategic National Stockpile (SNS). When Emergent bought the vaccine in 1998, a dose went for about $3.35. By 2010, the price was up to $28 a dose. Now, the price is over $30, and average wholesale prices are even higher, reaching $90 per dose, the lawmakers note.

“Emergent has raised the government purchasing price of the anthrax vaccine by 800% since acquiring the drug in 1998,” the lawmakers write in their letter. “As a result, through most of the last decade, nearly half of the SNS’s budget has been spent purchasing Emergent’s anthrax vaccine. These spiraling costs contributed to shortages of critical supplies, including ventilators, reusable respirator masks, and other personal protective equipment, which severely impacted the government’s ability to respond to the coronavirus crisis.”

This drain on the SNS budget was particularly apparent during the pandemic—which Robert Kadlec, the former Emergent consultant, himself acknowledged.

Kadlec was nominated in 2017 by President Trump to lead the Office of the Assistant Secretary for Preparedness and Response (ASPR). Following his confirmation, Emergent received millions of dollars in federal contracts from ASPR, including contracts for the SNS that were awarded without competitive bidding, the lawmakers note in their letter.

In 2020—just before the pandemic hit the US—Kadlec’s office awarded Emergent around $3 billion in long-term contracts for anthrax and other bioterrorism threats. According to the lawmakers, Kadlec later suggested this was a bad move, saying: “If I could spend less on anthrax replenishment, I could buy more N95s. I could buy more ventilators. I could buy more of other things that quite frankly I didn’t have the money to buy.”

Failures and an inspection

Aside from the skyrocketing prices, the lawmakers suggest Emergent didn’t even deserve the contracts in the first place. In 2012, the Department of Health and Human Services awarded Emergent a $163 million contract to renovate its (currently troubled) Baltimore manufacturing plant. The idea was for the plant to become a manufacturing hub for rapidly producing vaccine in the event of an infectious disease outbreak or bioterror attack. Part of the contract stipulated that Emergent would be required to do a test run, producing 50 million doses of a pandemic influenza vaccine in the span of four months, and obtain manufacturing approval from the FDA by June 2020.

Emergent failed to meet those requirements.

Reading the FDA’s inspection report of Emergent’s Baltimore facility, it’s clear why. During the nine-day inspection, which ended April 20, FDA inspectors logged a long list of problems at the facility.

First on the list is that Emergent failed to thoroughly investigate how the millions of Johnson & Johnson and AstraZeneca doses became contaminated. The agency concluded that, without a thorough review of what happened, it’s possible that other finished batches of vaccine may also be ruined. “There is no assurance that other batches have not been subject to cross contamination,” the inspectors wrote.

The FDA inspectors went on to note unsanitary conditions, paint peeling off of the walls and floors, residue on equipment, improperly trained staff, and numerous opportunities for vaccine products to be contaminated.

The potential for cross contamination—spread of viral ingredients back and forth between Johnson & Johnson’s vaccine and AstraZeneca’s vaccine—appeared rampant at the facility. Inspectors witnessed Emergent employees dragging unsealed, non-decontaminated bags of medical waste across different manufacturing areas. In some cases, employees tossed bags of medical waste, unsealed, into a service elevator. Emergent did not have proper written procedures for how to decontaminate waste, the inspection report notes. Security footage also caught employees moving from different areas of the facility without following proper procedures for donning and removing protective gowns.

At the request of the FDA, vaccine production at the Baltimore facility has been halted since April 16.

In a statement Wednesday, Emergent said that the FDA’s findings “provide direction on the necessary steps to improve operations.” The company went on:

The FDA’s feedback will help us continue to improve and strengthen the supply chain for Johnson & Johnson’s COVID-19 vaccine. While we are never satisfied to see shortcomings in our manufacturing facilities or process, they are correctable and we will take swift action to remedy them.

Continue Reading

Science

Trump EPA sidelined its own scientists when rewriting fuel economy rules

Published

on

Enlarge / Former EPA Administrator Scott Pruitt.

The Trump administration effectively muffled scientific staffers at the Environmental Protection Agency when it rewrote automobile pollution rules, the agency’s watchdog said.

When drafting fuel economy and greenhouse gas pollution rules for cars and light trucks, former EPA Administrator Scott Pruitt decided to cede various EPA duties to the National Highway Transportation Safety Administration in what is typically a collaborative process, the independent inspector general said in a report released yesterday. Though Pruitt signed the final report for the EPA, he allowed NHTSA staff to write a significant portion of the rules and to complete all modeling and analysis for both agencies.

The NHTSA’s modeling efforts did not use the EPA’s established tools that had been created to evaluate greenhouse gas emissions standards. Instead, the NHTSA hacked its own Corporate Average Fuel Economy models and sent EPA experts the results late in the process. “Technical personnel were unable to fully collaborate on rule development,” the report said.

The EPA and the NHTSA have collaborated on fuel economy and emissions standards since 2010 in an effort to streamline and harmonize regulations for automakers. In rules written during the Obama administration, the EPA and NHTSA called for 5 percent annual increases in fuel economy. After Trump was elected, though, automakers wrote to his transition team to request that the annual increases be frozen. The Trump administration obliged, reviewing the rule and issuing a new one calling for 1.5 percent increases in fuel economy. The reason for the reduction, the Trump administration said, was that it would help keep auto prices lower, allowing consumers to buy newer, safer vehicles. Not everyone was convinced by that logic.

That the final rule included such a rationale is not surprising in light of the inspector general’s report. When making fuel economy rules, the Energy Policy and Conservation Act directs the NHTSA to consider four different variables—“technological feasibility, economic practicability, the effect of other government standards on fuel economy, and the need of the nation to conserve energy,” the report points out. None of those includes pollution or its effects.

The EPA, on the other hand, is the only agency that the Clean Air Act allows to regulate greenhouse gas emissions from new vehicles.

Career EPA staff, who are not political appointees, reportedly felt shut out of the process, a drastic departure from the previous four times they worked with the NHTSA to jointly issue pollution and fuel economy rules.

The inspector general’s report also noted that the EPA skipped major milestones in its rule-making process. According to the report, the agency “did not document who decided to skip these milestones and why.”

New EPA Administrator Michael Regan said he would be reviewing the Trump-era fuel economy standards and will release a revised rule by the end of July. The new rules could effectively phase out fossil fuel vehicles.

Stifling science

This latest report isn’t the only one to find fault with the way the Trump administration went about writing or rewriting environmental regulations. EPA political appointees also ordered career staff to scale back efforts to monitor ethylene oxide pollution in the Chicago area. The flammable gas is used to sterilize medical equipment, but it’s also a proven carcinogen.

Chicago EPA staff, barred from an official effort to document ethylene oxide pollution, installed air pollution-monitoring equipment outside an agency warehouse that abutted a large sterilization plant owned by Sterigenics, a major emitter of ethylene oxide. The scientists discovered that pollution from the plant was at such high levels that it significantly increased the cancer risk in the surrounding neighborhood by orders of magnitude, according to a report by the Chicago Tribune. The company couldn’t get the plant’s pollution under control and ultimately closed the site after significant opposition from the community.

The EPA is also undertaking a broader review of decisions in which the agency’s scientific integrity may have been compromised. Regan asked employees to bring “items of concern” to scientific integrity officials or the inspector general, who would investigate accordingly.

Trump administration officials at the Centers for Disease Control and Prevention are also under fire for their meddling in scientific affairs. Earlier this month, documents released to the public showed political appointees pushing for changes in how the CDC framed reports about COVID-19 and the agency’s Morbidity and Mortality Weekly Reports. The alterations helped frame the CDC’s reporting on the pandemic in terms that hewed more closely to then-President Trump’s rosy outlook.

Continue Reading

Trending