Connect with us

A new tool lets artists add invisible changes to the pixels in their art before they upload it online so that if it’s scraped into an AI training set, it can cause the resulting model to break in chaotic and unpredictable ways. 

The tool, called Nightshade, is intended as a way to fight back against AI companies that use artists’ work to train their models without the creator’s permission. Using it to “poison” this training data could damage future iterations of image-generating AI models, such as DALL-E, Midjourney, and Stable Diffusion, by rendering some of their outputs useless—dogs become cats, cars become cows, and so forth. MIT Technology Review got an exclusive preview of the research, which has been submitted for peer review at computer security conference Usenix.

AI companies such as OpenAI, Meta, Google, and Stability AI are facing a slew of lawsuits from artists who claim that their copyrighted material and personal information was scraped without consent or compensation. Ben Zhao, a professor at the University of Chicago, who led the team that created Nightshade, says the hope is that it will help tip the power balance back from AI companies towards artists, by creating a powerful deterrent against disrespecting artists’ copyright and intellectual property. Meta, Google, Stability AI, and OpenAI did not respond to MIT Technology Review’s request for comment on how they might respond. 

Zhao’s team also developed Glaze, a tool that allows artists to “mask” their own personal style to prevent it from being scraped by AI companies. It works in a similar way to Nightshade: by changing the pixels of images in subtle ways that are invisible to the human eye but manipulate machine-learning models to interpret the image as something different from what it actually shows.

The team intends to integrate Nightshade into Glaze, and artists can choose whether they want to use the data-poisoning tool or not. The team is also making Nightshade open source, which would allow others to tinker with it and make their own versions. The more people use it and make their own versions of it, the more powerful the tool becomes, Zhao says. The data sets for large AI models can consist of billions of images, so the more poisoned images can be scraped into the model, the more damage the technique will cause.

targeted attack

Nightshade exploits a security vulnerability in generative AI models, one arising from the fact that they are trained on vast amounts of data—in this case, images that have been hoovered from the internet. Nightshade messes with those images.

Artists who want to upload their work online but don’t want their images to be scraped by AI companies can upload them to Glaze and choose to mask it with an art style different from theirs. They can then also opt to use Nightshade. Once AI developers scrape the internet to get more data to tweak an existing AI model or build a new one, these poisoned samples make their way into the model’s data set and cause it to malfunction.

Poisoned data samples can manipulate models into learning, for example, that images of hats are cakes, and images of handbags are toasters. The poisoned data is very difficult to remove, as it requires tech companies to painstakingly find and delete each corrupted sample.

The researchers tested the attack on Stable Diffusion’s latest models and on an AI model they trained themselves from scratch. When they fed Stable Diffusion just 50 poisoned images of dogs and then prompted it to create images of dogs itself, the output started looking weird—creatures with too many limbs and cartoonish faces. With 300 poisoned samples, an attacker can manipulate Stable Diffusion to generate images of dogs to look like cats.

A table showing a grid of thumbnails of generated images of Hemlock attack-poisoned concepts from SD-XL models contrasted with images from the clean SD-XL model in increments of 50, 100, and 300 poisoned samples.
COURTESY OF THE RESEARCHERS

Generative AI models are excellent at making connections between words, which helps the poison spread. Nightshade infects not only the word “dog” but all similar concepts, such as “puppy,” “husky,” and “wolf.” The poison attack also works on tangentially related images.

Read More

————

By: Melissa Heikkilä
Title: This new data poisoning tool lets artists fight back against generative AI
Sourced From: www.technologyreview.com/2023/10/23/1082189/data-poisoning-artists-fight-generative-ai/
Published Date: Mon, 23 Oct 2023 17:26:40 +0000

Total
0
Shares
Share 0
Tweet 0
Pin it 0
Share 0
Related Topics:

Tech

Climate tech is back—and this time, it can’t afford to fail

Lost in a stupor of déjà vu, I rang the intercom buzzer a second time. I had the odd sensation of being unstuck in time. The headquarters of this solar startup looked strangely similar to its previous offices, which I had visited more than a decade before. The name of the company had changed from 1366 Technologies to CubicPV, and it had moved about a mile away. But the rest felt familiar, right down to what I had come to talk about: a climate-tech boom.

A surge in cleantech investments, which had begun in 2006 with the high-profile entry of some of Silicon Valley’s leading venture capitalists, was still going strong during my first visit, in 2010—or at least it seemed to be. But a year later, it had begun to collapse. The rise of fracking was making natural gas cheap and abundant. US government funding for clean-energy research and deployment was falling. Meanwhile, China had begun to dominate solar and battery manufacturing. By the end of 2011, almost all the renewable-energy startups in the US were dead or struggling to survive.

The list of eventual casualties included headline grabbers like the solar-cell maker Solyndra and the high-flying battery company A123, as well as numerous less well-known startups in areas like advanced biofuels, innovative battery tech, and solar power. How, I was wondering, had CubicPV survived when nearly all its peers had failed?

Ushering me into the conference room (was that the same photo of a solar panel hanging on the wall that I had seen a decade before?), Frank van Mierlo, who is still the CEO, seemed almost giddy. And why not? After more than 10 years in photovoltaic limbo, with few opportunities to scale up its process for making the silicon wafers used in solar cells, the venture-backed company had suddenly seen its fortunes turn around.

The excitement around cleantech investments and manufacturing is back, and the money is flowing again. The 2022 US Inflation Reduction Act, which provides strong incentives for US domestic solar manufacturing, changed everything, says van Mierlo. As of this summer, some 44 new US plants had been planned, providing CubicPV with a huge potential demand for its silicon wafers.

Call it cleantech 2.0. In recent years, there has been a huge increase in public and private spending, both in the US and elsewhere, on technologies and infrastructure to address climate change. A recent analysis estimates that total green investments reached $213 billion in the US during the 12 months beginning July, 2022. Most of that spending is allocated to building sources of renewable energy, such as wind or solar, as well as to supporting battery and EV manufacturing and creating green hydrogen infrastructure. And the enormous amount of money is creating potential opportunities for the next generation of technologies to feed the expanding markets.

For startups like CubicPV, this means that after years of little market demand, the appetite for its products is suddenly almost insatiable. The company is designing a billion-dollar plant to make the silicon wafers needed to feed the rapid expansion in US solar production. What’s more, a bigger solar manufacturing base could eventually provide the startup with a lucrative future market for its next innovation: a new type of solar panel that is far more efficient at capturing sunlight than conventional silicon ones.

Silicon Valley and venture capitalists everywhere have fallen in love with the virtues and the promise of new catalysts and electrodes. Innovations in solar cells no longer seem like a lost cause. Startups are boasting radical new technologies for energy storage and carbon-free processes for making chemicals, steel, and cement. Investors are risking billions on scaling up nascent technologies such as geothermal power, fusion reactors, and ways to capture carbon dioxide directly from the air.

These innovations in what is being called “deep” or “hard” tech—products and processes based on science and engineering advances—could be critical in addressing climate change. While the past few years have seen remarkable progress in deploying relatively mature renewables such as solar and wind power, as well as strong growth in electric-vehicle sales, large gaps in the cleantech portfolio remain. In its most recent report this fall, the International Energy Agency estimates that around 35% of the emissions cuts needed to meet 2050 climate goals will have to come from technologies not yet available.

Key industrial sectors of the economy, in particular, have largely been untouched. Nearly a third of carbon emissions come from industrial processes used to make steel, cement, chemicals, and other commodities; concrete alone accounts for more than 7% of global emissions, while steel production is responsible for another 7% to 9%. Cleaning up these industries will take an almost unlimited supply of cheap, steady, and

Read More

————

By: David Rotman
Title: Climate tech is back—and this time, it can’t afford to fail
Sourced From: www.technologyreview.com/2023/12/02/1084059/climate-tech-startups-are-back-and-this-time-they-might-survive/
Published Date: Sat, 02 Dec 2023 12:00:00 +0000

Continue Reading

Tech

The Download: generative AI’s carbon footprint, and a CRISPR patent battle

This is today’s edition of The Download our weekday newsletter that provides a daily dose of what’s going on in the world of technology.

Making an image with generative AI uses as much energy as charging your phone

The news: Generating a single image using a powerful AI model takes as much energy as fully charging your smartphone, according to a new study. This is the first time researchers have calculated the carbon emissions caused by using an AI model for different tasks.

The significance: These emissions will add up quickly. The generative-AI boom has led big tech companies to integrate powerful AI models into many different products, from email to word processing. They are now used millions, if not billions, of times every single day.

The bigger picture: The study shows that while training massive AI models is incredibly energy intensive, it’s only one part of the puzzle. Most of their carbon footprint comes from their actual use. Read the full story.

—Melissa Heikkilä

The first CRISPR cure might kickstart the next big patent battle

By the middle of December, Vertex Pharmaceuticals is expected to receive FDA approval to sell a revolutionary new treatment for sickle-cell disease that’s the first in the US to use CRISPR to alter the DNA inside human cells. (Vertex has already received regulatory approval in the UK.)

But there’s a problem. The US patent on editing human cells with CRISPR isn’t owned by Vertex—it is owned by the Broad Institute of MIT and Harvard, probably America’s largest gene research center, and exclusively licensed to a Vertex competitor, Editas Medicine, which has its own sickle-cell treatment in testing.

That means Editas will want Vertex to pay. And if it doesn’t, Editas and Broad could go to the courts to claim patent infringement, demand royalties and damages, or even potentially try to stop the treatment from being sold. Odds are we’re about to see a blockbuster lawsuit. Read the full story.

—Antonio Regalado

This story is from The Checkup, our weekly newsletter giving you the inside track on all things health and biotech. Sign up to receive it in your inbox every Thursday.

A high school’s deepfake porn scandal is pushing US lawmakers into action

On October 20, Francesca Mani was called to the counselor’s office at her New Jersey high school. A 14-year-old sophomore and a competitive fencer, Francesca wasn’t one for getting in trouble. But it turned out that over the summer, boys in the school had used artificial intelligence to create sexually explicit pictures of some of their classmates. The school administration told Francesca that she was one of more than 30 girls who had been victimized.

Francesca didn’t see the photo of herself that day. And she still doesn’t intend to. Instead, she’s put all her energy into ensuring that no one else is targeted this way.

And, in the past few weeks, her advocacy has already fueled new legislative momentum to regulate nonconsensual deepfake pornography in the US. Read the full story.

—Tate Ryan-Mosley 

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 This is why we’re all sick right now
We’re contending with a lot more illnesses than we did in the pre-covid world. (The Atlantic $)
And covid hasn’t gone away either. (MIT Technology Review)

2 Climate disinformation is a big obstacle to action
And much of it is generated by influential nations, including China and Russia. (NYT $)
The US government has stopped warning social networks about foreign disinformation campaigns. (WP $)

3 Is the Turing Test dead?
It was arguably never that reliable a measure of intelligence to begin with. (IEEE Spectrum)
Mustafa Suleyman: My new Turing test would see if AI can make $1 million. (MIT Technology Review)
Hiring is still hot for prompt engineers, a year since ChatGPT launched. (Bloomberg $)

4 The long-delayed Tesla Cybertruck is finally on sale
And the price tag starts at $60,990. (The Guardian)
It has its detractors. But it has plenty of fans, too. (The Atlantic $)

5 College students are subject to alarming levels of surveillance
Which is adding to their stress levels at an already stressful time in their lives. (The Markup)
Computer scientists at Carnegie Mellon University can’t agree on what privacy means. (MIT Technology Review)

6 How Huawei stunned the US with a new Chinese-made chip
Getting around sanctions will have been difficult, and very expensive. (FT $)
Huawei’s 5G chip breakthrough needs a reality check. (MIT Technology Review)

7 Anduril has launched a wild new jet-powered AI drone
The company says it could be used in Ukraine to intercept Russian drones. (Wired $)

8 Startups have had a bad

Read More

————

By: Charlotte Jee
Title: The Download: generative AI’s carbon footprint, and a CRISPR patent battle
Sourced From: www.technologyreview.com/2023/12/01/1084204/the-download-ai-carbon-footprint-crispr-battle/
Published Date: Fri, 01 Dec 2023 13:10:00 +0000

Continue Reading

Tech

The Download: abandoning carbon offsets, and creating new materials

1f321

This is today’s edition of The Download our weekday newsletter that provides a daily dose of what’s going on in the world of technology.

The University of California has all but dropped carbon offsets—and thinks you should, too

In the fall of 2018, the University of California tasked a team of researchers with identifying projects from which it could confidently purchase carbon offsets that would reliably cancel out greenhouse gas emissions across its campuses. They found next to nothing.

The findings helped prompt the entire university system to radically rethink its sustainability plans. Now the researchers are sharing the lessons they learned over the course of the project, in the hopes of helping other universities and organizations consider what role, if any, offsets should play in sustainability strategies, MIT Technology Review can report.

The project’s leaders have three main takeaways for what others should do. Read our story to find out what they are.

—James Temple

Google DeepMind’s new AI tool helped create more than 700 new materials

The news: Google DeepMind has created a tool that uses deep learning to dramatically speed up the process of discovering new materials. The technology, which is called graphical networks for material exploration (GNoME), has already been used to predict structures for 2.2 million new materials, of which more than 700 have gone on to be created and tested in the lab.

Why it matters: From EV batteries to solar cells to microchips, new materials can supercharge technological breakthroughs. But discovering them usually takes months or even years of trial-and-error research. Thanks to GNoME, the number of known stable materials has grown almost tenfold, to 421,000. Read the full story.

—June Kim

The X Prize is taking aim at aging with a new $101 million award

Money can’t buy happiness, but X Prize founder Peter Diamandis hopes it might be able to buy better health. The X Prize Foundation, which funds global competitions to spark development of breakthrough technologies, has announced a new $101 million prize—the largest yet—to address the mental and physical decline that comes with aging.

The winners will have to prove by 2030 that their intervention can turn back the clock in older adults by at least a decade in three key areas: cognition, immunity, and muscle function. Its organizers are hoping the large prize will convince hundreds or even thousands of teams to compete. Read the full story.

—Cassandra Willyard

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 Ilya Sutskever is leaving the OpenAI board  
But the chief scientist is staying at the firm—for now. (Bloomberg $)
Microsoft has been added as a non voting member of the board. (NYT $)
Sam Altman says he initially felt furious after being asked to return to the company. (The Verge)
Inside the mind of OpenAI’s chief scientist. (MIT Technology Review)

2 Elon Musk isn’t playing nice with X’s worried advertisers
In fact, he went on a foul-mouthed rant railing against them. (CNBC)
Disney boss Bob Iger was a target of Musk’s ire. (WP $)
Musk just can’t help himself. (Slate $)

3 Next year is going to be even hotter
🌡
Thanks, in part, to the El Niño weather phenomenon. (FT $)
Methane is due to be a hot topic at COP28. (Economist $)+ Climate action is gaining momentum. So are the disasters. (MIT Technology Review)

4 Google has agreed to pay Canadian news outlets $100 million a year
It’s a rare win for publishers in their fight to get Big Tech to pay for content. (Motherboard)
Supporters say it’s the first step towards creating a sustainable news ecosystem. (BBC)

5 India is determined to clean up the Ganges river
The sacred waterway is incredibly polluted. Cleaning it up is both a holy and a scientific mission. (Wired $)
El Paso was “drought-proof.” Climate change is pushing its limits. (MIT Technology Review)

6 US border control is planning to trial Palmer Luckey’s AI surveillance towers
The autonomous towers track objects even in the coldest conditions. (404 Media)

7 Inside one man’s mission to track America’s gun violence
No one federal agency charts it, so Dan Kois has stepped up to fill the void. (Bloomberg $)

8 Please don’t follow TikTok’s dating advice
It’s bleak at best, outrageously sexist at worst. (Vox)
Here’s how the net’s newest matchmakers help you find love. (MIT Technology Review)

9 Cutting virtual grass is deeply satisfying
Just ask the video games fans transfixed by maintaining their lawns. (The Guardian)

10 What Spotify

Read More

————

By: Rhiannon Williams
Title: The Download: abandoning carbon offsets, and creating new materials
Sourced From: www.technologyreview.com/2023/11/30/1084123/the-download-abandoning-carbon-offsets-and-creating-new-materials/
Published Date: Thu, 30 Nov 2023 13:10:00 +0000

Did you miss our previous article…
https://mansbrand.com/sustainability-starts-with-the-data-center/

Continue Reading

Trending