fbpx
Connect with us

The Conversation

AI chatbots refuse to produce ‘controversial’ output − why that’s a free speech problem

Published

on

theconversation.com – Jordi Calvet-Bademunt, Research Fellow and Visiting Scholar of Political Science, Vanderbilt – 2024-04-18 07:23:58

AI chatbots restrict their output according to vague and broad policies.

taviox/iStock via Getty Images

Jordi Calvet-Bademunt, Vanderbilt University and Jacob Mchangama, Vanderbilt University

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.

Advertisement

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 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 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, 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 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.

Advertisement

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 to participate in women's 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.

Advertisement

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 . 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 search, word processors, email and other applications.

Advertisement

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 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 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.

Advertisement

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.The Conversation

Jordi Calvet-Bademunt, Research Fellow and Visiting Scholar of Political Science, Vanderbilt University and Jacob Mchangama, Research Professor of Political Science, Vanderbilt University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Advertisement

The Conversation

The perilous past and promising future of a toxic but nourishing crop

Published

on

theconversation.com – Stephen Wooding, Assistant Professor of Anthropology and Heritage Studies, University of California, Merced – 2024-05-01 07:36:48

A grower shows off his lush cassava garden.

Stephen Wooding, CC BY-ND

Stephen Wooding, University of California, Merced

The three staple crops dominating modern diets – corn, rice and wheat – are familiar to Americans. However, fourth place is held by a dark horse: cassava.

Advertisement

While nearly unknown in temperate climates, cassava is a key source of nutrition throughout the tropics. It was domesticated 10,000 years ago, on the southern margin of the Amazon basin in Brazil, and spread from there throughout the region. With a scraggly stem a few meters tall, a handful of slim branches and modest, hand-shaped leaves, it doesn't look like anything special. Cassava's humble appearance, however, belies an impressive combination of productivity, toughness and diversity.

Five people sit in background with several piles of peeled and unpeeled cassava tubers

People preparing to cassava, with some peeled tubers in the foreground.

Philippe Giraud/Corbis Historical via Getty Images

Over the course of millennia, Indigenous peoples bred it from a weedy wild plant into a crop that stores prodigious quantities of starch in potatolike tubers, thrives in Amazonia's poor soils and is nearly invulnerable to pests.

Cassava's many assets would seem to make it the ideal crop. But there's a problem: Cassava is highly poisonous.

Advertisement

How can cassava be so toxic, yet still dominate diets in Amazonia? It's all down to Indigenous ingenuity. For the past 10 years, my collaborator, César Peña, and I have been studying cassava gardens on the Amazon and its myriad tributaries in Peru. We have discovered scores of cassava varieties, growers using sophisticated breeding strategies to manage its toxicity, and elaborate methods for processing its dangerous yet nutritious products.

Long history of plant domestication

One of the most formidable challenges by early humans was getting enough to eat. Our ancient ancestors relied on hunting and gathering, catching prey on the and collecting edible plants at every . They were astonishingly good at it. So good that their populations soared, surging out of humanity's birthplace in Africa 60,000 years ago.

Even so, there was room for improvement. Searching the landscape for food burns calories, the very resource being sought. This paradox forced a trade-off for the hunter-gatherers: burn calories searching for food or conserve calories by staying home. The trade-off was nearly insurmountable, but humans found a way.

A little more than 10,000 years ago, they cleared the hurdle with one of the most transformative innovations in history: plant and animal domestication. People discovered that when plants and animals were tamed, they no longer needed to be chased down. And they could be selectively bred, producing larger fruits and seeds and bulkier muscles to eat.

Advertisement

Cassava was the champion domesticated plant in the neotropics. After its initial domestication, it diffused through the region, reaching sites as far north as Panama within a few thousand years. Growing cassava didn't completely eliminate people's need to search the forest for food, but it lightened the load, providing a plentiful, reliable food supply close to home.

, almost every rural family across the Amazon has a garden. Visit any household and you will find cassava roasting on the fire, being toasted into a chewy flatbread called casabe, fermenting into the beer called masato, and steaming in soups and stews. Before adopting cassava in these roles, though, people had to figure out how to deal with its toxicity.

Processing a poisonous plant

One of cassava's most important strengths, its pest resistance, is provided by a powerful defense system. The system relies on two chemicals produced by the plant, linamarin and linamarase.

These defensive chemicals are found inside cells throughout the cassava plant's leaves, stem and tubers, where they usually sit idle. However, when cassava's cells are damaged, by chewing or crushing, for instance, the linamarin and linamarase react, releasing a burst of noxious chemicals.

Advertisement

One of them is notorious: cyanide gas. The burst contains other nasty substances as well, including compounds called nitriles and cyanohydrins. Large doses of them are lethal, and chronic exposures permanently damage the nervous system. Together, these poisons deter herbivores so well that cassava is nearly impervious to pests.

Nobody knows how people first cracked the problem, but ancient Amazonians devised a complex, multistep process of detoxification that transforms cassava from inedible to delicious.

two women in hats peeling and shredding tubers

Women grind the cassava's starchy tubers into shreds.

Stephanie Maze/Corbis Historical via Getty Images

It begins with grinding cassava's starchy roots on shredding boards studded with fish teeth, chips of rock or, most often today, a rough sheet of tin. Shredding mimics the chewing of pests, causing the release of the root's cyanide and cyanohydrins. But they drift away into the , not into the lungs and stomach like when they are eaten.

Advertisement

Next, the shredded cassava is placed in rinsing baskets where it is rinsed, squeezed by hand and drained repeatedly. The action of the releases more cyanide, nitriles and cyanohydrins, and squeezing rinses them away.

Finally, the resulting pulp can be dried, which detoxifies it even further, or cooked, which finishes the process using heat. These steps are so effective that they are still used throughout the Amazon today, thousands of years since they were first devised.

man standing next to large vat with fire beneath, under thatch roof

People cooking cassava in the traditional way in the 1970s.

Education Images/Universal Images Group via Getty Images

A powerhouse crop poised to spread

Amazonians' traditional methods of grinding, rinsing and cooking are a sophisticated and effective means of converting a poisonous plant into a meal. Yet, the Amazonians pushed their efforts even further, taming it into a true domesticated crop. In addition to inventing new methods for processing cassava, they began keeping track and selectively growing varieties with desirable characteristics, gradually producing a constellation of types used for different purposes.

Advertisement

In our travels, we have found more than 70 distinct cassava varieties that are highly diverse, physically and nutritionally. They include types ranging in toxicity, some of which need laborious shredding and rinsing and others that can be cooked as is, though none can be eaten raw. There are also types with different tuber sizes, growth rates, starch production and drought tolerance.

Their diversity is prized, and they are often given fanciful names. Just as American supermarkets stock apples called Fuji, Golden Delicious and Granny Smith, Amazonian gardens stock cassavas called bufeo (dolphin), arpón (harpoon), motelo (tortoise) and countless others. This creative breeding cemented cassava's place in Amazonian cultures and diets, ensuring its manageability and usefulness, just as the domestication of corn, rice and wheat cemented their places in cultures elsewhere.

While cassava has been ensconced in South and Central America for millennia, its story is far from over. In the age of climate change and mounting efforts toward sustainability, cassava is emerging as a possible world crop. Its durability and resilience make it easy to grow in variable environments, even when soils are poor, and its natural pest resistance reduces the need to protect it with industrial pesticides. In addition, while traditional Amazonian methods for detoxifying cassava can be slow, they are easy to replicate and speed up with modern machinery.

two workers in white coats, hair caps and gloves show off white clumps they are bagging

Workers package frozen cassava in bags at a Florida food processing plant.

Juan Silva/The Image Bank via Getty Images

Advertisement

Furthermore, the preference of Amazonian growers to maintain diverse types of cassava makes the Amazon a natural repository for genetic diversity. In modern hands, they can be bred to produce new types, fitting purposes beyond those in Amazonia itself. These advantages spurred the first export of cassava beyond South America in the 1500s, and its range quickly spanned tropical Africa and Asia. Today, production in nations such as Nigeria and Thailand far outpaces production in South America's biggest producer, Brazil. These successes are raising optimism that cassava can become an eco-friendly source of nutrition for populations globally.

While cassava isn't a familiar name in the U.S. just yet, it's well on its way. It has long flown under the radar in the form of tapioca, a cassava starch used in pudding and boba tea. It's also the shelves in the snack aisle in the form of cassava chips and the baking aisle in naturally gluten- flour. Raw cassava is an emerging presence, too, showing up under the names “yuca” and “manioc” in stores catering to Latin American, African and Asian populations.

Track some down and give it a try. Supermarket cassava is perfectly safe, and recipes abound. Cassava fritters, cassava fries, cassava cakes … cassava's possibilities are nearly endless.


This article was co-authored by César Rubén Peña.The Conversation

Stephen Wooding, Assistant Professor of Anthropology and Heritage Studies, University of California, Merced

Advertisement

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Continue Reading

The Conversation

Sourdough under the microscope reveals microbes cultivated over generations

Published

on

theconversation.com – Daniel Veghte, Senior Research Associate Engineer, The Ohio – 2024-04-30 07:28:14

Microbes make a home among the starch grains of your sourdough starter.

Daniel Veghte, CC BY-SA

Daniel Veghte, The Ohio State University

Sourdough is the oldest kind of leavened bread in recorded history, and people have been eating it for thousands of years. The components of creating a sourdough starter are very simple – flour and . Mixing them produces a culture where yeast and bacteria ferment the sugars in flour, making byproducts that give sourdough its characteristic and smell. They are also what make it rise in the absence of other leavening agents.

Advertisement

My sourdough starter, affectionately deemed the “Fosters” starter, was passed down to me by my grandparents, who received it from my grandmother's college roommate. It has followed me throughout my academic career across the country, from undergrad in New Mexico to graduate school in Pennsylvania to postdoctoral work in Washington.

Currently, it resides in the Midwest, where I work at The Ohio State University as a senior research associate, collaborating with researchers to characterize samples in a wide variety of fields ranging from food science to material science.

As part of one of the microscopy courses I instruct at the university, I decided to take a closer look at the microbial community in my 's sourdough starter with the microscope I use in my day-to-day research.

Microscopy image of rod-shaped bacteria, elongated and spherical yeast, and globular starch grains

Each sourdough starter has a unique mix of microbes.

Daniel Veghte, CC BY-SA

Advertisement

Scanning electron microscopes

Scanning electron microscopy, or SEM, is a powerful tool that can image the surface of samples at the nanometer scale. For comparison, a human hair is between 10 to 150 micrometers, and SEM can observe features that are 10,000 times smaller.

Since SEM uses electrons instead of light for imaging, there are limitations to what can be imaged in the microscope. Samples must be electrically conductive and able to withstand the very low pressures in a vacuum. Low-pressure environments are generally unfavorable for microbes, since these conditions will cause the water in cells to evaporate, deforming their structure.

To prepare samples for SEM analysis, researchers use a method called critical point drying that carefully dries the sample to reduce unwanted artifacts and preserve fine details. The sample is then coated with a thin layer of iridium metal to make it conductive.

Round metal disk on a platform surrounded by a large cylindrical device

Scanning electron microscopes can image samples at the nanoscale level.

Daniel Veghte, CC BY-SA

Advertisement

Exploring a sourdough starter

Since sourdough starters are created from wild yeast and bacteria in the flour, it creates a favorable for many types of microbes to flourish. There can be more than 20 different species of yeast and 50 different species of bacteria in a sourdough starter. The most robust become the dominant species.

You can visually observe the microbial complexity of sourdough starter by imaging the different components that vary in size and morphology, yeast and bacteria. However, a full understanding of all the diversity present in the starter would require a complete gene sequencing.

The main component that gives the starter texture are starch grains from the flour. These grains, colored green in the image, are identifiable as relatively large globular structures approximately 8 micrometers in diameter.

Microscopy image of rod-shaped bacteria, elongated and spherical yeast, and globular starch grains

A false-colored scanning electron microscope image of a sourdough starter shows starch grains (green), yeast (red) and bacteria (blue).

Daniel Veghte, CC BY-SA

Advertisement

Giving rise to the starter is the yeast, colored red. As the yeast grows, it ferments sugars from the starch grains and releases carbon dioxide bubbles and alcohol as byproducts that make the dough rise. Yeast generally falls in the range of 2 to 10 micrometers in size and are round to elongated in shape. There are two distinct yeast types visible in this image, one that is nearly round, at the bottom left, and another that is elongated, at the top right.

Bacteria, colored blue, metabolize sugars and release byproducts such as lactic acid and acetic acid. These byproducts act as a preservative and are what give the starter its distinctive sour smell and taste. In this image, bacteria have pill-like shapes that are approximately 2 micrometers in size.

Now, the next time you eat sourdough bread or sourdough waffles – try them, they're delicious! – you can visualize the rich array of microorganisms that give each piece its distinctive flavor.The Conversation

Daniel Veghte, Senior Research Associate Engineer, The Ohio State University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Advertisement
Continue Reading

The Conversation

‘What is a fact?’ A humanities class prepares STEM students to be better scientists

Published

on

theconversation.com – Timothy Morton, Rita Shea Guffey Chair of English, Rice – 2024-04-30 07:29:12

A favorite class focuses on the tendency to see meaningful patterns where there aren't any, such as constellations of .

Yuga Kurita/Moment via Getty Images

Timothy Morton, Rice University

Text saying: Uncommon Courses, from The Conversation

Uncommon Courses is an occasional from The Conversation U.S. highlighting unconventional approaches to teaching.

Advertisement

Title of course:

What Is a Fact?

What prompted the idea for the course?

With all the conspiracy theories floating around in 2020 when hit, I wanted to help my learn to identify and deal with them. I was also concerned about political propaganda. And in my STEM-heavy school, I wanted to showcase what humanities scholars can do. So I created this class, which is distilled humanities for freshmen. Almost every student so far has been a science, technology, engineering and math major.

What does the course explore?

We start with a week called What Is Data? In Latin, “data” just means “things that are given.” Data can be in the form of measurements: “This bowlful of weighs x.” But data can also mean “it reminds me of my grandma.” How can you tell when something could be meaningful, or whether it's just nonsense?

A later class that students find especially interesting is on apophenia, the tendency to see patterns where there aren't any, like the man in the Moon, or constellations of stars.

Advertisement

chart illustrating dots of data, colored and connected in various ways as information, knowledge, insight, wisdom and conspiracy theory

Conspiracy theories connect a lot of dots, but that doesn't make them right.

Screenshot of a meme

Why is this course relevant now?

A fact is an interpretation of data. In physics class, you learn how to interpret physics data, find patterns, relate those patterns to other ones, and produce facts about them. If your argument hangs together logically, your interpretation can appear in the journal Nature Physics.

Humanities classes, however, prepare you to understand what facts are, period – whether they're based on biology or on the Bible, nutrition science or novels.

What's a critical lesson from the course?

One critical lesson is that many big conspiracy theories such as QAnon are about jumping to conclusions as quickly as possible. Being a good student and a good scholar means accepting that what you're examining might not be meaningful or might not indicate a pattern. What we're exploring here is how not to jump to conclusions. And this lesson applies as much to stuff in the real world as it does to lab work.

Advertisement

What materials does the course feature?

We watch YouTuber hbomberguy debunking global warming denialism. We read Kurt Gödel on how logical systems must always be flawed. We read poems and stories, introducing science majors to interpreting artistic data, a every bit as rigorous as interpreting scientific data.

What will the course prepare students to do?

Without the kinds of critical thinking this course teaches, scientists can be susceptible to propaganda and unable to share their ideas effectively, whether it's in the or to their colleagues, friends and .

Students learn to look at the world with fresh, skeptical eyes. They learn to identify illogical arguments and rhetorical strong-arm tactics. In the Middle Ages, humanities – grammar, logic, rhetoric – prepared you to do science. What Is a Fact? is like that, helping students see how collecting data and being skeptical don't stop once you've left the lab. A questioning, open-minded attitude is an essential skill.The Conversation

Timothy Morton, Rita Shea Guffey Chair of English, Rice University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Advertisement
Continue Reading

News from the South

Trending