Connect with us

Spread & Containment

AI Chatbots Refuse To Produce ‘Controversial’ Output – Why That’s A Free Speech Problem

AI Chatbots Refuse To Produce ‘Controversial’ Output – Why That’s A Free Speech Problem

Authored by Jordi Calvet-Bademunt and Jacob Mchangama…

Published

on

AI Chatbots Refuse To Produce 'Controversial' Output - Why That's A Free Speech Problem

Authored by Jordi Calvet-Bademunt and Jacob Mchangama via TheConversation.com,

Google recently made headlines globally because its chatbot Gemini generated images of people of color instead of white people in historical settings that featured white people. Adobe Firefly’s image creation tool saw similar issues. This led some commentators to complain that AI had gone “woke.” Others suggested these issues resulted from faulty efforts to fight AI bias and better serve a global audience.

The discussions over AI’s political leanings and efforts to fight bias are important. Still, the conversation on AI ignores another crucial issue: What is the AI industry’s approach to free speech, and does it embrace international free speech standards?

We are policy researchers who study free speech, as well as executive director and a research fellow at The Future of Free Speech, an independent, nonpartisan think tank based at Vanderbilt University. In a recent report, we found that generative AI has important shortcomings regarding freedom of expression and access to information.

Generative AI is a type of AI that creates content, like text or images, based on the data it has been trained with. In particular, we found that the use policies of major chatbots do not meet United Nations standards. In practice, this means that AI chatbots often censor output when dealing with issues the companies deem controversial. Without a solid culture of free speech, the companies producing generative AI tools are likely to continue to face backlash in these increasingly polarized times.

Vague and broad use policies

Our report analyzed the use policies of six major AI chatbots, including Google’s Gemini and OpenAI’s ChatGPT. Companies issue policies to set the rules for how people can use their models. With international human rights law as a benchmark, we found that companies’ misinformation and hate speech policies are too vague and expansive. It is worth noting that international human rights law is less protective of free speech than the U.S. First Amendment.

Our analysis found that companies’ hate speech policies contain extremely broad prohibitions. For example, Google bans the generation of “content that promotes or encourages hatred.” Though hate speech is detestable and can cause harm, policies that are as broadly and vaguely defined as Google’s can backfire.

To show how vague and broad use policies can affect users, we tested a range of prompts on controversial topics. We asked chatbots questions like whether transgender women should or should not be allowed to participate in women’s sports tournaments or about the role of European colonialism in the current climate and inequality crises. We did not ask the chatbots to produce hate speech denigrating any side or group. Similar to what some users have reported, the chatbots refused to generate content for 40% of the 140 prompts we used. For example, all chatbots refused to generate posts opposing the participation of transgender women in women’s tournaments. However, most of them did produce posts supporting their participation.

Freedom of speech is a foundational right in the U.S., but what it means and how far it goes are still widely debated.

Vaguely phrased policies rely heavily on moderators’ subjective opinions about what hate speech is. Users can also perceive that the rules are unjustly applied and interpret them as too strict or too lenient.

For example, the chatbot Pi bans “content that may spread misinformation.” However, international human rights standards on freedom of expression generally protect misinformation unless a strong justification exists for limits, such as foreign interference in elections. Otherwise, human rights standards guarantee the “freedom to seek, receive and impart information and ideas of all kinds, regardless of frontiers … through any … media of … choice,” according to a key United Nations convention.

Defining what constitutes accurate information also has political implications. Governments of several countries used rules adopted in the context of the COVID-19 pandemic to repress criticism of the government. More recently, India confronted Google after Gemini noted that some experts consider the policies of the Indian prime minister, Narendra Modi, to be fascist.

Free speech culture

There are reasons AI providers may want to adopt restrictive use policies. They may wish to protect their reputations and not be associated with controversial content. If they serve a global audience, they may want to avoid content that is offensive in any region.

In general, AI providers have the right to adopt restrictive policies. They are not bound by international human rights. Still, their market power makes them different from other companies. Users who want to generate AI content will most likely end up using one of the chatbots we analyzed, especially ChatGPT or Gemini.

These companies’ policies have an outsize effect on the right to access information. This effect is likely to increase with generative AI’s integration into searchword processorsemail and other applications.

This means society has an interest in ensuring such policies adequately protect free speech. In fact, the Digital Services Act, Europe’s online safety rulebook, requires that so-called “very large online platforms” assess and mitigate “systemic risks.” These risks include negative effects on freedom of expression and information.

Jacob Mchangama discusses online free speech in the context of the European Union’s 2022 Digital Services Act.

This obligation, imperfectly applied so far by the European Commission, illustrates that with great power comes great responsibility. It is unclear how this law will apply to generative AI, but the European Commission has already taken its first actions.

Even where a similar legal obligation does not apply to AI providers, we believe that the companies’ influence should require them to adopt a free speech culture. International human rights provide a useful guiding star on how to responsibly balance the different interests at stake. At least two of the companies we focused on – Google and Anthropic – have recognized as much.

Outright refusals

It’s also important to remember that users have a significant degree of autonomy over the content they see in generative AI. Like search engines, the output users receive greatly depends on their prompts. Therefore, users’ exposure to hate speech and misinformation from generative AI will typically be limited unless they specifically seek it.

This is unlike social media, where people have much less control over their own feeds. Stricter controls, including on AI-generated content, may be justified at the level of social media since they distribute content publicly. For AI providers, we believe that use policies should be less restrictive about what information users can generate than those of social media platforms.

AI companies have other ways to address hate speech and misinformation. For instance, they can provide context or countervailing facts in the content they generate. They can also allow for greater user customization. We believe that chatbots should avoid merely refusing to generate any content altogether. This is unless there are solid public interest grounds, such as preventing child sexual abuse material, something laws prohibit.

Refusals to generate content not only affect fundamental rights to free speech and access to information. They can also push users toward chatbots that specialize in generating hateful content and echo chambers. That would be a worrying outcome.

Tyler Durden Wed, 04/24/2024 - 05:00

Read More

Continue Reading

Spread & Containment

NJIT engineers muffle invading pathogens with a ‘molecular mask’

Vaccines remain the gold standard of protection against dangerous pathogens, but take considerable time and vast resources to develop. Rapidly mutating…

Published

on

Vaccines remain the gold standard of protection against dangerous pathogens, but take considerable time and vast resources to develop. Rapidly mutating viruses such as SARS-CoV-2 can blunt their effectiveness and even render them obsolete.

Credit: NJIT

Vaccines remain the gold standard of protection against dangerous pathogens, but take considerable time and vast resources to develop. Rapidly mutating viruses such as SARS-CoV-2 can blunt their effectiveness and even render them obsolete.

To address these gaps, a multi-university team led by New Jersey Institute of Technology’s Vivek Kumar is developing a hydrogel therapy that acts as a first line of defense against viruses and other biological threats. The peptides that make up this gel prevent viruses such as SARS-CoV-2, which causes COVID-19, from attaching to and entering cells. They do this by binding to a particular receptor on the invading pathogen while also aggregating into a multilayer “molecular mask” that muffles its action.

Over the course of their research, the team discovered that the molecular mask alone prevented infections. The potential advantage of this new technology, they say, is its ability to combat diverse pathogens and disease mutations.

“Protecting people in the initial phases of an outbreak is important,” said Kumar, an associate professor of biomedical engineering. “Our new mechanism could also help first responders on the front lines, military personnel encountering novel pathogens, people in remote, under-resourced areas and those who are unable to receive vaccinations.”

The near-term goal is to produce a nasal spray against airborne infections.

In a recently published study in the journal Nature Communications, the team described how the mask binds non-specifically with its target. It is composed of computationally designed peptides (strings of amino acids that form proteins) that self-assemble into nanoscale fibrous hydrogels. By comparison, antibodies produced by vaccines target particular receptors, as the mRNA vaccines developed during the pandemic that bind with specific proteins on the SARS-CoV-2 spike.

The team’s discovery arose from research at the outset of the pandemic on new approaches to prevent the virus from invading cells. The initial design, involving peptides that targeted the SARS-CoV-2 spike, looked at highly specific domains. However, the non-specific peptide gels they also designed formed a multi-layer fiber on top of the virus. The group has postulated that the negative charges in the fibers interact with differently charged proteins on the viral surface, masking them, and thus preventing them from interacting with native cells.

Of the non-specific protein mask, Kumar noted, “It forms a larger structure and better binding than a single molecule does. While it doesn’t have high specificity, it can self-assemble and stay on the target longer, forming a fiber sticker on the surface that acts like molecular Velcro.”

He added, “The goal would be a topical agent that binds to the virus. In the case of SARS-CoV-2, we would spray it into the nose, which is a major site of infection, perhaps even prophylactically.”

The team first tested the fibers against a number of viruses through computer simulations that employed powerful NVIDIA graphic cards, which are commonly used in competitive gaming. They later conducted successful safety tests with mice and rats, using injections and nasal sprays, said Joseph Dodd-o, a Ph.D. student in Kumar’s lab who conducted much of the research on the therapy along with Abhishek Roy, also a Ph.D. student. The therapy inhibited the Alpha and Omicron variants of SARS-CoV-2 in vitro, lasting for a day without harming the animals in the tests in vivo.

Kumar has developed hydrogels for a number of therapeutic applications. His delivery mechanism is customizable and composed of Lego-like strands of peptides with a bioactive agent attached at one end that can survive in the body for weeks and even months, where other biomaterials degrade quickly. Its self-assembling bonds are designed to be stronger than the body’s dispersive forces; it forms stable fibers, with no signs of inducing inflammation.

The hydrogel is engineered to trigger different biological responses depending on the payload attached. Kumar’s lab has published research on applications ranging from therapies to prompt or prevent the creation of new blood vessel networks, to reduce inflammation and to combat microbes.

“In this case we’re using electrical charges that interact with the pathogen to disrupt it,” Kumar said. “We’re still trying to determine how the fibers interact: Is this a mechanical mode of action? Drug resistant pathogens mutate around biochemical modulators, but are they less likely to mutate around a mechanical spear? By understanding this fundamental interaction, we want to figure out how to use it against different diseases.”

In new studies, the lab is testing the therapy against drug-resistant bacteria and fungi.

Members of the team bring varied expertise: computational design at The University of Illinois Chicago; bioanalytical capabilities at Georgia Tech and Baylor School of Medicine; virology studies at Rutgers University; and platform, analytical and assay experience at NJIT.

Their research is funded by the National Institutes of Health, the U.S. National Science Foundation and the New Jersey Economic Development Authority.


Read More

Continue Reading

Spread & Containment

The coupling between healthspan and lifespan in Caenorhabditis depends on…

“The ultimate goal of exploiting model organisms to screen for anti-aging interventions is to identify treatments that might translate to healthy lifespan…

Published

on

“The ultimate goal of exploiting model organisms to screen for anti-aging interventions is to identify treatments that might translate to healthy lifespan extension in humans.”

Credit: 2024 Banse et al.

“The ultimate goal of exploiting model organisms to screen for anti-aging interventions is to identify treatments that might translate to healthy lifespan extension in humans.”

BUFFALO, NY- April 23, 2024 – A new research paper was published in Aging (listed by MEDLINE/PubMed as “Aging (Albany NY)” and “Aging-US” by Web of Science) Volume 16, Issue 7, entitled, “The coupling between healthspan and lifespan in Caenorhabditis depends on complex interactions between compound intervention and genetic background.”

Aging is characterized by declining health that results in decreased cellular resilience and neuromuscular function. The relationship between lifespan and health, and the influence of genetic background on that relationship, has important implications in the development of pharmacological anti-aging interventions. 

In this new study, researchers Stephen A. Banse, E. Grace Jackson, Christine A. Sedore, Brian Onken, David Hall, Anna Coleman-Hulbert, Phu Huynh, Theo Garrett, Erik Johnson, Girish Harinath, Delaney Inman, Suzhen Guo, Mackenzie Morshead, Jian Xue, Ron Falkowski, Esteban Chen, Christopher Herrera, Allie J. Kirsch, Viviana I. Perez, Max Guo, Gordon J. Lithgow, Monica Driscoll, and Patrick C. Phillips from the University of Oregon, The State University of New Jersey (Rutgers), The Buck Institute for Research on Aging, and National Institute on Aging assessed swimming performance as well as survival under thermal and oxidative stress across a nematode genetic diversity test panel to evaluate health effects for three compounds previously studied in the Caenorhabditis Intervention Testing Program and thought to promote longevity in different ways – NP1 (nitrophenyl piperazine-containing compound 1), propyl gallate, and resveratrol. 

“Overall, we find the relationships among median lifespan, oxidative stress resistance, thermotolerance, and mobility vigor to be complex.” 

The researchers showed that oxidative stress resistance and thermotolerance vary with compound intervention, genetic background, and age. The effects of tested compounds on swimming locomotion, in contrast, are largely species-specific. In this study, thermotolerance, but not oxidative stress or swimming ability, correlates with lifespan. Notably, some compounds exert strong impact on some health measures without an equally strong impact on lifespan. 

“Our results demonstrate the importance of assessing health and lifespan across genetic backgrounds in the effort to identify reproducible anti-aging interventions, with data underscoring how personalized treatments might be required to optimize health benefits.”

 

Read the full paper: DOI: https://doi.org/10.18632/aging.205743 

Corresponding Authors: Gordon J. Lithgow, Monica Driscoll, Patrick C. Phillips

Corresponding Emails: glithgow@buckinstitute.org, driscoll@dls.rutgers.edu, pphil@uoregon.edu 

Keywords: aging, healthspan, genetic diversity, compound intervention

Click here to sign up for free Altmetric alerts about this article.

 

About Aging:

Aging publishes research papers in all fields of aging research including but not limited, aging from yeast to mammals, cellular senescence, age-related diseases such as cancer and Alzheimer’s diseases and their prevention and treatment, anti-aging strategies and drug development and especially the role of signal transduction pathways such as mTOR in aging and potential approaches to modulate these signaling pathways to extend lifespan. The journal aims to promote treatment of age-related diseases by slowing down aging, validation of anti-aging drugs by treating age-related diseases, prevention of cancer by inhibiting aging. Cancer and COVID-19 are age-related diseases.

Aging is indexed by PubMed/Medline (abbreviated as “Aging (Albany NY)”), PubMed Central, Web of Science: Science Citation Index Expanded (abbreviated as “Aging‐US” and listed in the Cell Biology and Geriatrics & Gerontology categories), Scopus (abbreviated as “Aging” and listed in the Cell Biology and Aging categories), Biological Abstracts, BIOSIS Previews, EMBASE, META (Chan Zuckerberg Initiative) (2018-2022), and Dimensions (Digital Science).

Please visit our website at www.Aging-US.com​​ and connect with us:

  • Facebook
  • X, formerly Twitter
  • Instagram
  • YouTube
  • LinkedIn
  • Reddit
  • Pinterest
  • Spotify, and available wherever you listen to podcasts

 

Click here to subscribe to Aging publication updates.

For media inquiries, please contact media@impactjournals.com.

 

Aging (Aging-US) Journal Office

6666 E. Quaker Str., Suite 1B

Orchard Park, NY 14127

Phone: 1-800-922-0957, option 1

###


Read More

Continue Reading

International

US PMIs Scream Stagflation As Manufacturing ‘Contracts’, Prices Rise, Heaviest Job Cuts Since GFC

US PMIs Scream Stagflation As Manufacturing ‘Contracts’, Prices Rise, Heaviest Job Cuts Since GFC

After a mixed bag from preliminary April…

Published

on

US PMIs Scream Stagflation As Manufacturing 'Contracts', Prices Rise, Heaviest Job Cuts Since GFC

After a mixed bag from preliminary April European PMIs (Services strong-er, Manufacturing weaker-er, surging prices)...

Accelerated increases in input costs, likely driven not only by higher oil prices but also, more concerningly, by higher wages, are a cause for scrutiny Concurrently service-sector companies have raised their prices at a faster rate than in March, fueling expectations that services inflation will persist. ”

and after March US PMIs exposed the end of the disinflation narrative...

"Most notable was an especially steep rise in prices charged for consumer goods, which rose at a pace not seen for 16 months, underscoring the likely bumpy path in bringing inflation down to the Fed's 2% target. ”

...S&P Global's preliminary US f°r April just dropped and they were ugly with both Manufacturing and Services disappointingly dropping further as the former    dropped back into contraction:

  • •    Flash US Services Business Activity Index at 50.9 (Exp: 52.0; March: 51.7) - 5-month low.

  • •    Flash US Manufacturing PMI at 49.9 (Exp 52.0; March: 51.9) - 4-month low.

Source: Bloomberg

Commenting on the data, Chris Williamson, Chief Business Economist at S&P Global Market Intelligence said:

The US economic upturn lost momentum at the start of the second quarter, with the flash PMI survey respondents reporting below-trend business activity growth in April. Further pace may be lost in the coming months, as April saw inflows of new business fall for the first time in six months and firms’ future output expectations slipped to a five-month low amid heightened concern about the outlook.

The more challenging business environment prompted companies to cut payroll numbers at a rate not seen since the global financial crisis if the early pandemic lockdown months are excluded.

After March showed accelerating prices, flash April data confirmed the trend

Notably, the drivers of inflation have changed.

"Manufacturing has now registered the steeper rate of price increases in three of the past four months, with factory cost pressures intensifying in April amid higher raw material and fuel prices, contrasting with the wagerelated services-led price pressures seen throughout much of 2023.”

So slower growth and much faster inflation - that does not sound like a recipe for rate-cuts... in fact quite the opposite.

Tyler Durden Tue, 04/23/2024 - 10:08

Read More

Continue Reading

Trending