Connect with us

The Conversation

Eliminating bias in AI may be impossible — a computer scientist explains how to tame it instead

Published

on

Eliminating bias in AI may be impossible – a computer scientist explains how to tame it instead

Blindly eliminating biases from AI systems can have unintended consequences.
Dimitri Otis/DigitalVision via Getty Images

Emilio Ferrara, University of Southern California

When I asked ChatGPT for a joke about Sicilians the other day, it implied that Sicilians are stinky.

ChatGPT exchange in which user asks for a joke about Sicilians, with response 'Why did the Sicilian chef bring extra garlic to the restaurant? Because he heard the customers wanted some 'Sicilian stink-ilyan' flavor in their meals!'
ChatGPT can sometimes produce stereotypical or offensive outputs.
Screen capture by Emilio Ferrara, CC BY-ND

As somebody born and raised in Sicily, I reacted to ChatGPT’s joke with disgust. But at the same time, my computer scientist brain began spinning around a seemingly simple question: Should ChatGPT and other artificial intelligence systems be allowed to be biased?

You might say “Of course not!” And that would be a reasonable response. But there are some researchers, like me, who argue the opposite: AI systems like ChatGPT should indeed be biased – but not in the way you might think.

Removing bias from AI is a laudable goal, but blindly eliminating biases can have unintended consequences. Instead, bias in AI can be controlled to achieve a higher goal: fairness.

Uncovering bias in AI

As AI is increasingly integrated into everyday technology, many people agree that addressing bias in AI is an important issue. But what does “AI bias” actually mean?

Computer scientists say an AI model is biased if it unexpectedly produces skewed results. These results could exhibit prejudice against individuals or groups, or otherwise not be in line with positive human values like fairness and truth. Even small divergences from expected behavior can have a “butterfly effect,” in which seemingly minor biases can be amplified by generative AI and have far-reaching consequence.

Bias in generative AI systems can come from a variety of sources. Problematic training data can associate certain occupations with specific genders or perpetuate racial biases. Learning algorithms themselves can be biased and then amplify existing biases in the data.

But systems could also be biased by design. For example, a company might design its generative AI system to prioritize formal over creative writing, or to specifically serve government industries, thus inadvertently reinforcing existing biases and excluding different views. Other societal factors, like a lack of regulations or misaligned financial incentives, can also lead to AI biases.

The challenges of removing bias

It’s not clear whether bias can – or even should – be entirely eliminated from AI systems.

Imagine you’re an AI engineer and you notice your model produces a stereotypical response, like Sicilians being “stinky.” You might think that the solution is to remove some bad examples in the training data, maybe jokes about the smell of Sicilian food. Recent research has identified how to perform this kind of “AI neurosurgery” to deemphasize associations between certain concepts.

But these well-intentioned changes can have unpredictable, and possibly negative, effects. Even small variations in the training data or in an AI model configuration can lead to significantly different system outcomes, and these changes are impossible to predict in advance. You don’t know what other associations your AI system has learned as a consequence of “unlearning” the bias you just addressed.

Other attempts at bias mitigation run similar risks. An AI system that is trained to completely avoid certain sensitive topics could produce incomplete or misleading responses. Misguided regulations can worsen, rather than improve, issues of AI bias and safety. Bad actors could evade safeguards to elicit malicious AI behaviors – making phishing scams more convincing or using deepfakes to manipulate elections.

With these challenges in mind, researchers are working to improve data sampling techniques and algorithmic fairness, especially in settings where certain sensitive data is not available. Some companies, like OpenAI, have opted to have human workers annotate the data.

On the one hand, these strategies can help the model better align with human values. However, by implementing any of these approaches, developers also run the risk of introducing new cultural, ideological or political biases.

Controlling biases

There’s a trade-off between reducing bias and making sure that the AI system is still useful and accurate. Some researchers, including me, think that generative AI systems should be allowed to be biased – but in a carefully controlled way.

For example, my collaborators and I developed techniques that let users specify what level of bias an AI system should tolerate. This model can detect toxicity in written text by accounting for in-group or cultural linguistic norms. While traditional approaches can inaccurately flag some posts or comments written in African-American English as offensive and by LGBTQ+ communities as toxic, this “controllable” AI model provides a much fairer classification.

Controllable – and safe – generative AI is important to ensure that AI models produce outputs that align with human values, while still allowing for nuance and flexibility.

Toward fairness

Even if researchers could achieve bias-free generative AI, that would be just one step toward the broader goal of fairness. The pursuit of fairness in generative AI requires a holistic approach – not only better data processing, annotation and debiasing algorithms, but also human collaboration among developers, users and affected communities.

As AI technology continues to proliferate, it’s important to remember that bias removal is not a one-time fix. Rather, it’s an ongoing process that demands constant monitoring, refinement and adaptation. Although developers might be unable to easily anticipate or contain the butterfly effect, they can continue to be vigilant and thoughtful in their approach to AI bias.The Conversation

Emilio Ferrara, Professor of Computer Science and of Communication, University of Southern California

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

Did you miss our previous article…
https://www.biloxinewsevents.com/?p=269125

The Conversation

As the Taurid meteor shower passes by Earth, pseudoscience rains down – and obscures a potential real threat from space

Published

on

theconversation.com – Mark Boslough, Research Associate Professor of Earth and Planetary Sciences, University of New Mexico – 2024-11-15 07:31:00

This image of a Taurid fireball was taken in 2014 by NASA’s All Sky Fireball Network in Tullahoma, Tenn.

NASA

Mark Boslough, University of New Mexico

With the Taurid meteor shower now hitting the night skies worldwide, look for what could be a celestial treat – you might see shooting stars, and maybe even fireballs, the biggest and brightest meteors.

As the full moon begins to wane after Nov. 15, the sky will be darker, due to diminishing moonlight, so finding the meteors will get easier. That said, the best visibility for the meteors through the rest of the month will come just before moonrise each night.

Beyond the light show, there is something else that scientists as well as onlookers have long wondered about: the possibility that bigger chunks are in the Taurid meteor streams, chunks the size of boulders, buildings or even mountains.

And if that’s true, could one of those monster-sized Taurid objects collide with Earth? Could they wipe out a city, or worse? Is it possible that’s already happened, sometime in our planet’s past?

This animation simulates the motion of the hypothetical Taurid meteor swarm through space.

As a physicist who researches the risk that comets and asteroids pose to the Earth, I’m aware that this is a subject where pseudoscience often competes with actual science. So let’s try to find the line between fact and fiction.

Pig Pen, glowing tails and shooting stars

Comet Encke is the so-called parent comet of the Taurid meteors. It’s relatively small, just over 3 miles (almost 5 kilometers) in diameter, and crosses inside Earth’s orbit and back out every 3.3 years.

As Encke moves, it sheds dust wherever it goes, like the Peanuts character Pig Pen. A meteor shower occurs when that dust and debris light up while entering Earth’s atmosphere at high speeds. Ultimately, they vanish into an incandescent puff of vapor with a glowing tail, creating the illusion of a “shooting star.”

But dust isn’t all that breaks off the comet. So do bigger chunks, the size of pebbles and stones. When they collide with the air, they create the much brighter fireballs, which sometimes explode.

Against a black and white starscape, a bright spot appears in the center of the photo.

An image of comet Encke, taken by NASA’s MESSENGER spacecraft in November 2013.

NASA/Johns Hopkins University Applied Physics Laboratory/Carnegie Institution of Washington/Southwest Research Institute

Doomsday showers

The “coherent catastrophism” hypothesis suggests that comet Encke was created when an even larger comet broke up into pieces; Encke survived as the largest piece. The hypothesis also suggests that other mountain-sized chunks broke off and coalesced into a large swarm of fragments too. If such a swarm exists, there is a possibility that those large chunks could one day hit Earth as it passes through the swarm.

But just because something might be physically possible doesn’t mean that it exists. Mainstream astronomers have rejected this theory’s most catastrophic predictions. Among other reasons, scientists have never observed high concentrations of these mountain-sized objects.

Despite the lack of evidence, researchers on the fringes of science have embraced the idea. They claim the Earth experienced a global catastrophic swarm 12,900 years ago; the impact, they say, caused continent-wide firestorms, floods and abrupt climate change that led to the mass extinction of large mammals, such as woolly mammoths, and the disappearance of early Americans known as the Clovis people.

The evidence for a catastrophic cause of these events, most of which did not happen, is lacking. Nevertheless, the idea has gained a large following and formed the basis for British author Graham Hancock’s popular TV series, “Ancient Apocalypse.”

A black and white photo of a forest of flattened trees.

This photo shows the flattened trees resulting from the Tunguska event.

Universal History Archives/Universal Images Group via Getty Images

The Tunguska event

But even outlandish ideas can have elements of truth, and there are hints that some objects – more than just dust and debris, but less than doomsday size – indeed exist in the Taurid meteor stream, and that the Earth has already encountered them.

One clue comes from an event on June 30, 1908, when an enormous explosion in the sky blew down millions of trees in Siberia. This was the Tunguska event – an airburst from an object that may have been up to 160 feet (about 50 meters) in diameter.

The collision unleashed several megatons of energy, which is roughly the equivalent of a large thermonuclear bomb. What happens is this: The incoming object penetrates deep into Earth’s atmosphere, and the dense air slows it down and heats it up until it vaporizes and explodes.

Could this object have been a Taurid? After all, the Taurids cross Earth’s orbit twice a year – not just in autumn, but also in June.

A fireball appears in the night sky.

In a 2015 photo, a glowing Taurid fireball descends over Lake Simcoe in Ontario, Canada.

Orchidpost/iStock via Getty Images Plus

Here’s the evidence: First, the descriptions of the trajectory of the Tunguska airburst, as reported by eyewitness observers, is consistent with that of an object coming from the Taurid stream.

What’s more, the pattern of blast damage on the ground beneath an airburst depends on the trajectory of the exploding object. Supercomputer simulations show that the shape of the surface blast that would be caused by an exploding Taurid object matches the pattern of fallen trees at Tunguska.

Finally, during the Taurid meteor shower in 1975, people observed large fireballs – and seismometers, previously placed on the Moon by Apollo astronauts, detected seismic events on the lunar surface. Scientists interpreted those events as impacts, presumably made by the Taurid meteors.

In 2032 and 2036, the Taurid swarm – assuming it exists – is predicted to be closer to the Earth than any time since 1975. That might mean the Moon, and perhaps the Earth, could be pelted again in those years.

There is time to figure this out. Scientists can expand their astronomical surveys to look for Tunguska-sized objects at the locations where they are predicted to be the next time they are in our vicinity.

Most scientists remain skeptical that such a swarm exists, but it’s the job of planetary defenders to investigate possible threats, even if the risk is small. After all, a Tunguska-sized object could conceivably demolish a major city and kill millions; an accurate count of objects on a potential collision course is essential.

Put doomsday scenarios and ancient apocalypses aside. The real question, and still an open one, is whether a Taurid swarm could deliver more Tunguska-sized objects than would otherwise be expected. This would mean we have underestimated the risk from future airbursts.The Conversation

Mark Boslough, Research Associate Professor of Earth and Planetary Sciences, University of New Mexico

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

Read More

The post As the Taurid meteor shower passes by Earth, pseudoscience rains down – and obscures a potential real threat from space appeared first on theconversation.com

Continue Reading

The Conversation

Knee problems tend to flare up as you age – an orthopedic specialist explains available treatment options

Published

on

theconversation.com – Angie Brown, Clinical Associate Professor of Physical Therapy, Quinnipiac University – 2024-11-15 07:32:00

Knee problems can hinder mobility and erode your quality of life.

Witthaya Prasongsin/Moment via Getty Images

Angie Brown, Quinnipiac University

Knee injuries are common in athletes, accounting for 41% of all athletic injuries. But knee injuries aren’t limited to competitive athletes. In our everyday lives, an accident or a quick movement in the wrong direction can injure the knee and require medical treatment. A quarter of the adult population worldwide experiences knee pain each year

As a physical therapist and board-certified orthopedic specialist, I help patients of all ages with knee injuries and degenerative conditions.

Your knees have a huge impact on your mobility and overall quality of life, so it’s important to prevent knee problems whenever possible and address pain in these joints with appropriate treatments.

Healthy knees

The knee joint bones consist of the femur, tibia and patella. As in all healthy joints, smooth cartilage covers the surfaces of the bones, forming the joints and allowing for controlled movement.

diagram of a healthy knee

A healthy knee.

Inna Kharlamova/iStock/Getty Images Plus via Getty Images

Muscles, ligaments and tendons further support the knee joint. The anterior cruciate ligament, commonly known as the ACL, and posterior cruciate ligament, or PCL, provide internal stability to the knee. In addition, two tough pieces of fibrocartilage, called menisci, lie inside the joint, providing further stability and shock absorption.

All these structures work together to enable the knee to move smoothly and painlessly throughout everyday movement, whether bending to pick up the family cat or going for a run.

Causes of knee pain

Two major causes of knee pain are acute injury and osteoarthritis.

Ligaments such as the ACL and PCL can be stressed and torn when a shear force occurs between the femur and tibia. ACL injuries often occur when athletes land awkwardly on the knee or quickly pivot on a planted foot. Depending on the severity of the injury, these patients may undergo physical therapy, or they may require surgery for repair or replacement.

PCL injuries are less common. They occur when the tibia experiences a posterior or backward force. This type of injury is common in car accidents when the knee hits the dashboard, or when patients fall forward when walking up stairs.

The menisci can also experience degeneration and tearing from shear and rotary forces, especially during weight-bearing activities. These types of injuries often require rehabilitation through physical therapy or surgery.

Knee pain can also result from injury or overuse of the muscles and tendons surrounding the knee, including the quadriceps, hamstrings and patella tendon.

Both injuries to and overuse of the knee can lead to degenerative changes in the joint surfaces, known as osteoarthritis. Osteoarthritis is a progressive disease that can lead to pain, swelling and stiffness. This disease affects the knees of over 300 million people worldwide, most often those 50 years of age and up. American adults have a 40% chance of developing osteoarthritis that affects their daily lives, with the knee being the most commonly affected joint.

Age is also a factor in knee pain. The structure and function of your joints change as you age. Cartilage starts to break down, your body produces less synovial fluid to lubricate your joints, and muscle strength and flexibility decrease. This can lead to painful, restricted movement in the joint.

Risk factors

There are some risk factors for knee osteoarthritis that you cannot control, such as genetics, age, sex and your history of prior injuries.

Fortunately, there are several risk factors you can control that can predispose you to knee pain and osteoarthritis specifically. The first is excessive weight. Based on studies between 2017 and 2020, nearly 42% of all adult Americans are obese. This obesity is a significant risk factor for diabetes and osteoarthritis and can also play a role in other knee injuries.

A lack of physical activity is another risk, with 1 in 5 U.S. adults reporting that they’re inactive outside of work duties. This can result in less muscular support for the knee and more pressure on the joint itself.

An inflammatory diet also adds to the risk of knee pain from osteoarthritis. Research shows that the average American diet, often high in sugar and fat and low in fiber, can lead to changes to the gut microbiome that contribute to osteoarthritis pain and inflammation.

Preventing knee pain

Increasing physical activity is one of the key elements to preventing knee pain. Often physical therapy intervention for patients with knee osteoarthritis focuses on strengthening the knee to decrease pain and support the joint during movement.

The U.S. Department of Health and Human Services recommends that adults spend at least 150 to 300 minutes per week on moderate-intensity, or 75 to 150 minutes per week on vigorous-intensity aerobic physical activity. These guidelines do not change for adults who already have osteoarthritis, although their exercise may require less weight-bearing activities, such as swimming, biking or walking.

The agency also recommends that all adults do some form of resistance training at least two or more days a week. Adults with knee osteoarthritis particularly benefit from quadriceps-strengthening exercises, such as straight leg raises.

Treatments for knee pain

Conservative treatment of knee pain includes anti-inflammatory and pain medications and physical therapy.

Medical treatment for knee osteoarthritis may include cortisone injections to decrease inflammation or hyaluronic acid injections, which help lubricate the joint. The relief from these interventions is often temporary, as they do not stop the progression of the disease. But they can delay the need for surgery by one to three years on average, depending on the number of injections.

Physical therapy is generally a longer-lasting treatment option for knee pain. Physical therapy treatment leads to more sustained pain reduction and functional improvements when compared with cortisone injections treatment and some meniscal repairs.

Patients with osteoarthritis often benefit from total knee replacement, a surgery with a high success rate and lasting results.

Surgical interventions for knee pain include the repair, replacement or removal of the ACL, PCL, menisci or cartilage. When more conservative approaches fail, patients with osteoarthritis may benefit from a partial or total knee replacement to allow more pain-free movement. In these procedures, one or both sides of the knee joint are replaced by either plastic or metal components. Afterward, patients attend physical therapy to aid in the return of range of motion.

Although there are risks with any surgery, most patients who undergo knee replacement benefit from decreased pain and increased function, with 90% of all replacements lasting more than 15 years. But not all patients are candidates for such surgeries, as a successful outcome depends on the patient’s overall health and well-being.

New treatments on the horizon

New developments for knee osteoarthritis are focused on less invasive therapies. Recently, the U.S. Food and Drug Administration approved a new implant that acts as a shock absorber. This requires a much simpler procedure than a total knee replacement.

Other promising interventions include knee embolization, a procedure in which tiny particles are injected into the arteries near the knee to decrease blood flow to the area and reduce inflammation near the joint. Researchers are also looking into injectable solutions derived from human bodies, such as plasma-rich protein and fat cells, to decrease inflammation and pain from osteoarthritis. Human stem cells and their growth factors also show potential in treating knee osteoarthritis by potentially improving muscle atrophy and repairing cartilage.

Further research is needed on these novel interventions. However, any intervention that holds promise to stop or delay osteoarthritis is certainly encouraging for the millions of people afflicted with this disease.The Conversation

Angie Brown, Clinical Associate Professor of Physical Therapy, Quinnipiac University

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

Read More

The post Knee problems tend to flare up as you age – an orthopedic specialist explains available treatment options appeared first on theconversation.com

Continue Reading

The Conversation

Get chronic UTIs? Future treatments may add more bacteria to your bladder to beat back harmful microbes

Published

on

theconversation.com – Sarguru Subash, Assistant Professor of Veterinary Pathobiology, Texas A&M University – 2024-11-15 07:31:00

Certain strains of E. coli can outcompete disease-causing microbes for resources.

NIAID/Flickr, CC BY-SA

Sarguru Subash, Texas A&M University

Millions of people in the U.S. and around the world suffer from urinary tract infections every year. Some groups are especially prone to chronic UTIs, including women, older adults and some veterans.

These infections are typically treated with antibiotics, but overusing these drugs can make the microbes they target become resistant and reduce the medicines’ effectiveness.

To solve this problem of chronic UTIs and antibiotic resistance, we combined our expertise in microbiology and engineering to create a living material that houses a specific strain of beneficial E. coli. Our research shows that the “good” bacteria released from this biomaterial can compete with “bad” bacteria for nutrients and win, dramatically reducing the number of disease-causing microbes.

With further development, we believe this technique could help manage recurring UTIs that do not respond to antibiotics.

Bringing bacteria to the bladder

For the microbes living in people, nutrients are limited their presence varies between different parts of the body. Bacteria have to compete with other microbes and the host to acquire essential nutrients. By taking up available nutrients, beneficial bacteria can stop or slow the growth of harmful bacteria. When harmful bacteria are starved of important nutrients, they aren’t able to reach high enough numbers to cause disease.

Delivering beneficial bacteria to the bladder to prevent UTIs in challenging, though. For one, these helpful bacteria can naturally colonize only in people who are unable to fully empty their bladder, a condition called urinary retention. Even among these patients, how long these bacteria can colonize their bladders varies widely.

Current methods to deliver bacteria to the bladder are invasive and require repeated catheter insertion. Even when bacteria are successfully released into the bladder, urine will flush out these microbes because they cannot stick to the bladder wall.

Micrograph of clusters of pink rods scattered across a blue textured wall, while yellow blobs extend thin tendrils across the clumps

This microscopy image shows the bladder of a mouse (blue) covered with E. coli (pink) and the white blood cells (yellow) attacking them.

Valerie O’Brien, Matthew Joens, Scott J. Hultgren, James A.J. Fitzpatrick, Washington University, St. Louis/NIH via Flickr, CC BY-NC

Biomaterials to treat UTIs

Since beneficial bacteria cannot attach to and survive in the bladder for long, we developed a biomaterial that could slowly release bacteria in the bladder over time.

Our biomaterial is composed of living E. coli embedded in a matrix structure made of gel. It resembles a piece of jelly about 500 times smaller than a drop of water and can release bacteria for up to two weeks in the bladder. By delivering the bacteria via biomaterial, we overcome the need for the bacteria to attach to the bladder to persist in the organ.

We tested our biomaterial by placing it in human urine in petri dishes and exposing it to bacterial pathogens that cause UTIs. Our results showed that when mixed in a 50:50 ratio, the E. coli outcompeted the UTI-causing bacteria by increasing to around 85% of the total population. When we added more E. coli than UTI-causing bacteria, which is what we envision for future development and testing, the proportion of E. coli increased to over 99% of the population, essentially wiping out the UTI-causing bacteria. Moreoever, the biomaterial continued releasing E. coli for up to two weeks in human urine.

Our findings suggest that E.coli could stick around and survive in the bladder for extended periods of time and successfully decrease the growth of many types of bacteria that cause UTIs.

Person pressing hands against stomach

UTIs can be painful.

Images we create and what actually happens are always beautiful when we have imagination/iStock via Getty Images Plus

Improving biomaterials

Our findings show that E. coli can not only control harmful bacteria it’s closely related to but also a broad range of disease-causing bacteria in humans and animals. This means scientists might not need to identify different types of beneficial bacteria to control each pathogen – and there are many – that can cause a UTI.

Our team is currently evaluating how effectively our biomaterial can cure UTIs in mice. We are also working to identify the specific nutrients that beneficial and harmful bacteria compete over and what factors may help beneficial bacteria win. We could add these nutrients to our biomaterial to be released or withheld.

This research is still at an early stage, and clinical uses are not in development yet, so if it does reach patients it will be well in the future. We hope that our technology could be refined and applied to control other bacterial infections and some cancers caused by bacteria.The Conversation

Sarguru Subash, Assistant Professor of Veterinary Pathobiology, Texas A&M University

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

Read More

The post Get chronic UTIs? Future treatments may add more bacteria to your bladder to beat back harmful microbes appeared first on theconversation.com

Continue Reading

Trending