You're reading Entrepreneur India, an international franchise of Entrepreneur Media.

A recently published study 'Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data' by DeepMind, a division of Google's parent company Alphabet revealed that the most common misuse of artificial intelligence is the creation of deepfakes of politicians and celebrities, rather than AI-assisted cyber attacks.

The findings are based on the analysis of media reports of GenAI misuse between January 2023 and March 2024.

Source: Report

The study notes that the most reported cases of GenAI misuse involve actors exploiting the capabilities of these systems, rather than launching direct attacks at the models themselves. Nine out of ten cases fall into this category, DeepMind shared.

Manipulation of human likeness is the most prevalent cluster of tactics. These may include Impersonation; Sockpuppeting: Appropriated Likeness: and NCII. Scaling & Amplification and Falsification are also prominent tactics, accounting for 13 per cent and 12 per cent of reported cases respectively.

When it comes to goals and strategies of misuse, Opinion Manipulation ranked first with 27 per cent of all reported cases, followed by Monetization & Profit at 21 per cent and Scam & Fraud at 18 per cent.

In case of Opinion Manipulation, a range of tactics are deployed to distort the public's perception of political realities. These include impersonating public figures, using synthetic digital personas to simulate grassroots support for or against a cause ('astroturfing'), and creating falsified media.

Deemed as the 'election year, 2024 will see at least 64 countries head to the polls. The influence of deepfakes has been felt in countries like the USA, Nigeria, and Bangladesh, particularly during elections. In India, voters saw deep fakes of politicians and celebrities such as Aamir Khan, Ranveer Singh, and KT Rama Rao make rounds during the election phases. Other cases include Vladimir Putin declaring martial law after Ukrainian forces entered Russian territory;

ALSO READ: Election Essentials: 4 Websites to Identify Deepfakes and Fake News During India's 2024 Elections

The study found defamation to be another central strategy for opinion manipulation. According to the study, data involved depicted electoral candidates spouting abuse towards protected groups, party staffers, or their own constituents; while in other cases, actors shared AI-generated images of politicians appearing visibly aged to make them look unfit for leadership, and showing them in intimate settings with other public figures.

A common factor was the lack of appropriate disclosure around the use of GenAI tools in the context of campaigning risks misleading users and causing harm through deception.

The second most common goal behind GenAI misuse was to monetize products and services. These tactics include content scaling, amplification, and falsification.

Content farming saw users producing low-quality AI-generated articles, books, and product ads for placement on websites such as Amazon and Etsy to cut costs and capitalize on advertising revenue. The creation of non-consensual intimate imagery (NCII) constituted a significant portion. This tactic saw the creation and selling of sexually explicit videos of celebrities who did not consent to the production of that content.

The Scam & Fraud misuse saw the leveraging of real identities to deceive victims. This included celebrity scam ads and phishing scams. These not only infringe upon the targeted individual or organization's rights and reputation but also inflict a financial and psychological cost on victims.

"Addressing these challenges will require not only technical advancements, but a multi-faceted approach to interventions, involving collaboration between policymakers, researchers, industry leaders, and civil society. We highlight these implications in our discussion," said Nahema Marchal, lead author of the study and researcher at Google DeepMind on X.

Visit link:
The Most Common Misuse of GenAI Is For Influencing Public Opinion: Google's DeepMind - Entrepreneur

As society embraces new technologies and scientific advancements, scientists must navigate the complex balance of exchanging ideas and preventing misinformation and polarization within public audiences.

Bryce Johnson, a computer sciences graduate student in the lab of Morgridge computation biologist Anthony Gitter, put this idea into practice through anopinion pieceabout the artificial intelligence software AlphaFold 3 in the digital magazine,Undark.

AlphaFold is an AI model developed by Google DeepMind, a private research subsidiary of Google. The most recent model,AlphaFold 3, was published inNature.

This model claims it can predict the structure of proteins and their interaction with DNA, RNA, and other types of biological molecules, says Johnson. The ability to know how these molecules interact with a protein can be really useful to revolutionize drug discovery.

This technology could be useful for other biotechnology applications beyond drug discovery, including solutions to mitigate the effects of climate change a topic that inspires Johnsons passion for science communication and policy.

Johnson appreciates Google DeepMinds decision to publish their model in a reputable scientific journal. But he takes issue with how theNatureeditors seemingly allowed the AlphaFold 3 model to be described without adhering to the journals usual open-source standards.

Their mission is to serve scientists in publishing academic material, so it didnt feel like they were staying true to their mission, he says. It felt like this was more about a promotion for a for-profit company.

Johnsons opinion piece grew out of an assignment for one of his classes with the UWMadison Department ofLife Sciences Communication(LSC). As the AlphaFold 3 model went public, he pitched his story to several publications that he hoped would be willing to work with PhD candidates.

Within a day, editors fromUndarkreplied and commissioned his piece, which Johnson said reaffirmed that his piece fit the criteria for news value timely, relevant with a clear conflict and argument.

He worked withUndarkeditors for about a month, as he was encouraged to take a step back and make sure he represented viewpoints from all stakeholders. This applied one of his major takeaways from his LSC class: Before you speak, listen.

Initially the piece stood as a demand that DeepMind release their code, Johnson explains. I had to address a lot of new and incoming information which changed the nature of my opinion piece from less of a demand to more of a brief of the situation itself with an assessment of my own.

Johnson is grateful for research mentor Gitter, as well as being in a working environment passionate about the public implications of science. The Morgridge Institute encourages all researchers to assess their relationships with science and society, with a commitment to programs that advancescience communicationandcommunity engagement.

This is a perfect example of the Morgridge mission of having scientists communicate to a broader audience about important issues, Gitter says.

Johnson hopes his experiences will one day lead to a career in science policy. Read his fullopinion piece at Undark.

Read the original post:
Grad student contemplates open-source conflicts with new AI software - Morgridge Institute for Research - Morgridge Institute for Research

Researchers from Google DeepMind and Harvard University have built a virtual rodent powered by artificial intelligence to better understand how the brain controls movement.

The virtual rat is powered by an artificial neural network that mimics the neural activity of its real-life counterpart, giving researchers a chance to compare the two.

While animals have exquisite control of their bodies, allowing them to perform a diverse range of bahaviors, how the brain implements such control remains unclear, researchers said.

To get a better understanding of how the brain works, researchers trained the virtual rodent to mimic the whole-body movements of freely moving rats in a physics simulator, where an artificial neural network actuated a biomechanically realistic model of the rat.

We then compared neural activity from the real rats brain to the activations of the virtual rodents artificial neural network when performing the same behaviors, lead author Diego Aldorando said in a thread of posts on X.

We found that the virtual rodents neural networks, which implement inverse dynamics models, were better predictors of neural activity than measurable features of movement, like the positions or velocities of the joints, or alternative control models, Aldorando said.

Researchers used deep reinforcement learning to train the virtual agent to imitate the behavior of freely moving rats, according to the paper published in Nature.

The results of the study demonstrated how physical simulation of biomechanically realistic virtual animals can help interpret the structure of neural activity across behavior and relate it to theoretical principles of motor control, the paper read.

According to Aldorando, their research approach can be applied in neuroscience and facilitate the study of aspects of neuromotor control that are hard to experimentally deduce. It could also be instrumental in modeling the neural control of increasingly complex animal behavior.

Rowan Cheung, founder of the Rundown AI newsletter, said the study could massively open up new research with testing on AI animals and expand robotics.

Book review: The Secret Life of Data is fascinating and disturbing

Honor launches a foldable phone challenging Samsung

Ascension hospitals breach caused by employee downloading malicious file

Russian Matryoshka strikes the media ahead of the Paris Olympics

Former NSA chief joins OpenAI's board of directors

Subscribe to our newsletter

Read the rest here:
Google DeepMind and Harvard build virtual rat with AI brain - CyberNews.com

Summary: Researchers created a virtual rat with an AI brain to study how real rats control movement. Using data from real rats, they trained the AI to mimic behaviors in a physics simulator.

The virtual rats neural activations closely matched those of real rats, offering new insights into brain function. This innovation could revolutionize neuroscience and improve robotic control systems.

Key Facts:

Source: Harvard

The agility with which humans and animals move is an evolutionary marvel thatno robot has yet been able to closely emulate.

To help probe the mystery of how brains control movement, Harvard neuroscientists have created a virtual rat with an artificial brain that can move around just like a real rodent.

Bence lveczky, professor in the Department of Organismic and Evolutionary Biology, led a group of researchers who collaborated with scientists at Googles DeepMind AI lab to build a biomechanically realistic digital model of a rat.

Using high-resolution data recorded from real rats, they trained an artificial neural network the virtual rats brain to control the virtual body in a physics simulator calledMuJoco, where gravity and other forces are present.

Publishing inNature,the researchers found that activations in the virtual control network accurately predicted neural activity measured from the brains of real rats producing the same behaviors, said lveczky, who is an expert at training (real) rats to learn complex behaviors in order to study their neural circuitry.

The feat represents a new approach to studying how the brain controls movement, lveczky said, by leveraging advances in deep reinforcement learning and AI, as well as 3D movement-tracking in freely behaving animals.

The collaboration was fantastic, lveczky said. DeepMind had developed a pipeline to train biomechanical agents to move around complex environments. We simply didnt have the resources to run simulations like those, to train these networks.

Working with the Harvard researchers was, likewise, a really exciting opportunity for us, said co-author and Google DeepMind Senior Director of Research Matthew Botvinick.

Weve learned a huge amount from the challenge of building embodied agents: AI systems that not only have to think intelligently, but also have to translate that thinking into physical action in a complex environment.

It seemed plausible that taking this same approach in a neuroscience context might be useful for providing insights in both behavior and brain function.

Graduate student Diego Aldarondo worked closely with DeepMind researchers to train the artificial neural network to implement what are called inverse dynamics models, which scientists believe our brains use to guide movement. When we reach for a cup of coffee, for example, our brain quickly calculates the trajectory our arm should follow and translates this into motor commands.

Similarly, based on data from actual rats, the network was fed a reference trajectory of the desired movement and learned to produce the forces to generate it. This allowed the virtual rat to imitate a diverse range of behaviors, even ones it hadnt been explicitly trained on.

These simulations may launch an untapped area of virtual neuroscience in which AI-simulated animals, trained to behave like real ones, provide convenient and fully transparent models for studying neural circuits, and even how such circuits are compromised in disease.

While lveczkys lab is interested in fundamental questions about how the brain works, the platform could be used, as one example, to engineer better robotic control systems.

A next step might be to give the virtual animal autonomy to solve tasks akin to those encountered by real rats.

From our experiments, we have a lot of ideas about how such tasks are solved, and how the learning algorithms that underlie the acquisition of skilled behaviors are implemented, lveczky continued.

We want to start using the virtual rats to test these ideas and help advance our understanding of how real brains generate complex behavior.

Author: Anne Manning Source: Harvard Contact: Anne Manning Harvard Image: The image is credited to Google DeepMind

Original Research: Closed access. A virtual rodent predicts the structure of neural activity across behaviors by Bence lveczky et al. Nature

Abstract

A virtual rodent predicts the structure of neural activity across behaviors

Animals have exquisite control of their bodies, allowing them to perform a diverse range of behaviors. How such control is implemented by the brain, however, remains unclear. Advancing our understanding requires models that can relate principles of control to the structure of neural activity in behaving animals.

To facilitate this, we built a virtual rodent, in which an artificial neural network actuates a biomechanically realistic model of the ratin a physics simulator.

We used deep reinforcement learningto train the virtual agent to imitate the behavior of freely-moving rats, thus allowing us to compare neural activity recorded in real rats to the network activity of a virtual rodent mimicking their behavior.

We found that neural activity in the sensorimotor striatum and motor cortex was better predicted by the virtual rodents network activity than by any features of the real rats movements, consistent with both regions implementing inverse dynamics.

Furthermore, the networks latent variability predicted the structure of neural variability across behaviors and afforded robustness in a way consistent with the minimal intervention principle of optimal feedback control.

These results demonstrate how physical simulation of biomechanically realistic virtual animals can help interpret the structure of neural activity across behavior and relate it to theoretical principles of motor control.

Read more from the original source:
Virtual Rat with AI Brain Mimics Real Rodent Movement - Neuroscience News

Current and former employees at OpenAI and Google DeepMind have signed an open letter warning of the risks of artificial intelligence (AI), highlighting insufficient whistleblower protections and the threat of retaliation.

The letter is signed confidentially by six current and former OpenAI employees; and publicly by five former OpenAI employees, one former DeepMind employee, and one current DeepMind employee.

The authors state they believe in the potential of AI technology, but believe it also poses numerous risks including the further entrenchment of existing inequalities, manipulation and misinformation, and the loss of control of autonomous AI systems leading to human extinction.

They argue that AI companies have a "financial incentive" to avoid effective AI oversight, and "weak obligation" to share capabilities, limitations, risks, and the adequacy of protective measures.

The outcome, according to the signatories, is that employees are among the few who can hold AI companies to account, but fear doing so due to fear of retaliation.

Broad confidentiality agreements block us from voicing our concerns, except to the very companies that may be failing to address these issues, the letter states, arguing that current protections for whistleblowers are insufficient as they are predicated on illegal activity, whereas much of the AI landscape remains unregulated.

Some of us reasonably fear various forms of retaliation, given the history of such cases across the industry, the current and former AI staffers write.

The workers ask advanced AI companies to commit to four principles they believe would mitigate retaliation against workers who raise concerns, criticism, limitations, and risks associated with AI technology.

The principles include commitments not to enforce any agreement that bans workers from disparagement or criticism and not to retaliate by hindering any vested economic benefit.

Companies are also asked to support a culture of open criticism, achieved in part by setting up an anonymous channel for current and former employees to raise concerns to the company's board, regulators, or an independent body.

The group also recommends companies do not retaliate against current or former workers who resort to publicly sharing concerns if their efforts to do so in other (private) channels have failed.

In response to the letter, an spokesperson told CNN OpenAI is proud of our track record providing the most capable and safest AI systems and believe in our scientific approach to addressing risk.

The spokesperson added OpenAI agrees with the need for rigorous debate and pointed out its anonymous integrity hotline and the recent announcement of its Safety and Security Committee.

However, one of the letters signatories, Daniel Ziegler, worked for OpenAI from 2018 and 2021, and questions the companys commitment to safety and transparency.

Read more from us

Its really hard to tell from the outside how seriously theyre taking their commitments for safety evaluations and figuring out societal harms, especially as there is such strong commercial pressures to move very quickly, he told CNN. Its really important to have the right culture and processes so that employees can speak out in targeted ways when they have concerns.

OpenAI came under fire earlier in May after Vox News reported the exit of two high-profile safety researchers whichrevealed clauses in non-disclosure and non-disparagement agreements (NDAs) that could have cost workers vested equity if they criticized the company.

CEO Sam Altman said he was genuinely embarrassed, but added the company had never clawed back equity from a current or former worker and would strike the policy from all paperwork of current and future staff.

See the original post:
OpenAI, Google DeepMind employees warn of a culture of retaliation in open letter - HR Grapevine