Copyright: ft.com – “Does Investment Research Make Sense in the Age of AI?”

Large language models can outperform analysts in some areas but human intuition is invaluable.

The late Byron Wien, a prominent markets strategist of the 1990s, defined the best research as a non-consensus recommendation that turned out to be right. Could AI pass Wien’s test of worthwhile research and make the analyst job redundant? Or at the very least increase the probability of a recommendation to be right more than 50 per cent of the time? Well, it is important to understand that most analyst reports are devoted to the interpretation of financial statements and news. This is about facilitating the job of investors. Here, modern large language models simplify or displace this analyst function.

Next, a good amount of effort is spent predicting earnings. Given that most of the time profits tend to follow a pattern, as good years follow good years and vice versa, it is logical that a rules-based engine would work. And because the models do not need to “be heard” by standing out from the crowd with outlandish projections, their lower bias and noise can outperform most analysts’ estimates in periods where there is limited uncertainty. Academics wrote about this decades ago, but the practice did not take off in mainstream research. To scale, it required a good dose of statistics or building a neural network. Rarely in the skillset of an analyst.

Change is under way. Academics from University of Chicago trained large language models to estimate variance of earnings. These outperformed median estimates when compared with those of analysts. The results are fascinating because LLMs generate insights by understanding the narrative of the earnings release, as they do not have what we may call numerical reasoning — the edge of a narrowly trained algorithm.[…]

Read more: www.ft.com

Der Beitrag Does Investment Research Make Sense in the Age of AI? erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

SwissCognitive Guest Blogger: Utpal Chakraborty, Chief Digital Officer, Allied Digital Services Ltd., AI & Quantum Scientist – “AI Pioneers Claim Nobel Prizes: Transforming the Future of Science”

The year 2024 will be remembered for generations, marking a historic milestone as artificial intelligence researchers make unprecedented strides in multiple Nobel Prize categories. For the first time, AI pioneers were recognized not solely for advancements in AI itself but for groundbreaking contributions to physics and chemistry. This achievement highlights how the lines between traditional sciences and computer science (specifically AI) are blurring in ways that would have seemed unimaginable just a few decades ago.

The announcement sent ripples through the scientific community when Geoffrey Hinton and John Hopfield shared the Nobel Prize in Physics, while Demis Hassabis along with two other scientists claimed the Chemistry prize. Three brilliant minds known primarily for their AI work, now recognized for transforming our understanding of the physical world.

Geoffrey Hinton and John Hopfield received the Physics Nobel for their work on understanding phase transitions in complex systems through the lens of Neural Computation. Their groundbreaking discovery showed how the mathematics of phase transitions in materials shares fundamental principles with how Neural Networks learn and process information.

Hopfield’s contribution stemmed from his revolutionary 1982 paper (Neural networks and physical systems with emergent collective computational abilities) introducing the Hopfield network, a mathematical model that showed how collections of simple units could exhibit complex behavior similar to phase transitions in physics. The model demonstrated how memory could emerge from the collective behavior of simple components, much like how magnetic properties emerge in materials.

Hinton’s work complemented this by revealing how the principles of statistical mechanics, traditionally used to understand particle behavior in physics, could explain deep learning’s success. His breakthrough came from showing that the way neural networks optimize their weights (Backpropagation) follows the same mathematical principles that govern how physical systems find their lowest energy states.

Of course, many of us know these scientists primarily for their AI contributions:

– Hopfield’s neural networks revolutionized our understanding of associative memory and laid the groundwork for modern deep learning.

– Hinton’s work on backpropagation and deep belief networks essentially created the deep learning revolution we’re experiencing today.

But it’s their ability to bridge these seemingly disparate fields that makes their Physics Nobel Prize so significant. As Hinton once said at a conference, “The brain is a physical system. Why shouldn’t its principles help us understand other physical systems?”

On the other hand, Demis Hassabis’s Chemistry Nobel came for something equally remarkable – using AI principles to solve one of chemistry’s grand challenges, protein folding. His work at DeepMind led to AlphaFold2, but the Nobel recognized his deeper insights into how the principles of reinforcement learning could reveal fundamental rules governing molecular interactions.

The prize specifically acknowledged his team’s discovery of new chemical principles through AI analysis, principles that classical scientists had missed. By training AI systems to understand molecular behavior, they uncovered previously unknown patterns in how proteins fold and interact, revolutionizing our understanding of chemical processes at the molecular level.

Most know Hassabis as the founder of DeepMind and the mind behind AlphaGo, but his journey from AI to chemistry illustrates a broader trend in science. His background in neuroscience and computer games gave him a unique perspective on how complex systems organize themselves, whether they are neural networks, game strategies, or molecular structures.

What makes these Nobel Prizes so fascinating is how they highlight the convergence of different scientific disciplines.

The work of Hinton, Hopfield, and Hassabis shows us that these aren’t separate fields anymore, they are different lenses for viewing the same reality. Their discoveries reveal a deeper unity in science that we are only beginning to appreciate.

As I write this article, I can’t help but feel we are living through a new scientific revolution. The tools of AI aren’t just helping us do traditional science faster; they are fundamentally changing how we think about science itself.

Young researchers today don’t see themselves as just physicists, chemists, or computer scientists. They are explorers in a unified landscape where:

– Physical laws inform neural network design.

– Chemical principles inspire new computing architectures.

– AI algorithms reveal new patterns in nature.

What strikes me most about these Nobel laureates is their humanity. Despite working with machines and mathematical abstractions, they never lost sight of the human element in science.

As someone who has worked in these intersecting fields, I see these Nobel Prizes as more than just recognition of brilliant work. They are a signal that the future of science lies not in specialization, but in synthesis. The next generation of scientists won’t just cross boundaries – they’ll erase them.

These Nobel Prizes aren’t just awards; they are a glimpse of science’s future. A future where the boundaries between classical physics, quantum physics, chemistry, and computation disappear and where artificial intelligence helps us see the unity that was always there.

Der Beitrag AI Pioneers Claim Nobel Prizes: Transforming the Future of Science erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

Here is the latest AI news collection for you that we think you should read:

LG’s AI slashes battery cell design time by 93%
The urgent need for global AI literacy
AI revolutionizes operations management
Samsung introduces “Hybrid AI” to balance smartphone AI
New neural network model adapts and learns like the human brain
… and more!

Look into these insights and #ShareForSuccess!

Best regards,

The Team of SwissCognitive

Der Beitrag Urgent Need For Global AI Literacy erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

Copyright: htworld.co.uk – “‘Fitbit for the gut’: researchers create GPS-like smart pills with AI”

Developed by the Khan Lab, these capsules are tailored to identify gasses associated with gastritis and gastric cancers.

The research, to be published in Cell Reports Physical Science, shows how these smart pills have been accurately monitored through a newly designed wearable system.

This breakthrough represents a significant step forward in ingestible technology, which Yasser Khan, an Assistant Professor of Electrical and Computer Engineering at USC, believes could someday serve as a ‘Fitbit for the gut’ and for early disease detection.

While wearables with sensors hold a lot of promise to track body functions, the ability to track ingestible devices within the body has been limited.

However, with innovations in materials, the miniaturisation of electronics, as well as new protocols developed by Khan, researchers have demonstrated the ability to track the location of devices specifically in the GI tract.

Khan’s team with the USC Institute for Technology and Medical Systems Innovation (ITEMS) at the Michelson Center for Convergent Biosciences, placed a wearable coil that generates a magnetic field on a t-shirt.

This field coupled with a trained neural network, allows his team to locate the capsule within the body.

According to Ansa Abdigazy, lead author of the work and a PhD student in the Khan Lab, this has not been demonstrated with a wearable before.

The second innovation within this device is the newly created “sensing” material. Capsules are outfitted not just with electronics for tracking location but with “optical sensing membrane that is selective to gasses.

This membrane is comprised of materials whose electrons change their behavior within the presence of ammonia gas.[…]

Read more: www.htworld.co.uk

Der Beitrag ‘Fitbit For The Gut’: Researchers Create GPS-Like Smart Pills With AI erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

Copyright: spectrum.ieee.org – “Will Scaling Solve Robotics?”

Last year’s Conference on Robot Learning (CoRL) was the biggest CoRL yet, with over 900 attendees, 11 workshops, and almost 200 accepted papers. While there were a lot of cool new ideas (see this great set of notes for an overview of technical content), one particular debate seemed to be front and center: Is training a large neural network on a very large dataset a feasible way to solve robotics?

Of course, some version of this question has been on researchers’ minds for a few years now. However, in the aftermath of the unprecedented success of ChatGPT and other large-scale “foundation models” on tasks that were thought to be unsolvable just a few years ago, the question was especially topical at this year’s CoRL. Developing a general-purpose robot, one that can competently and robustly execute a wide variety of tasks of interest in any home or office environment that humans can, has been perhaps the holy grail of robotics since the inception of the field. And given the recent progress of foundation models, it seems possible that scaling existing network architectures by training them on very large datasets might actually be the key to that grail.

Given how timely and significant this debate seems to be, I thought it might be useful to write a post centered around it. My main goal here is to try to present the different sides of the argument as I heard them, without bias towards any side. Almost all the content is taken directly from talks I attended or conversations I had with fellow attendees. My hope is that this serves to deepen people’s understanding around the debate, and maybe even inspire future research ideas and directions.

I want to start by presenting the main arguments I heard in favor of scaling as a solution to robotics.

Why Scaling Might Work

It worked for Computer Vision (CV) and Natural Language Processing (NLP), so why not robotics? This was perhaps the most common argument I heard, and the one that seemed to excite most people given recent models like GPT4-V and SAM. The point here is that training a large model on an extremely large corpus of data has recently led to astounding progress on problems thought to be intractable just 3-4 years ago.[…]

Read more: www.spectrum.ieee.org

This post was originally published on the author’s personal blog.

Der Beitrag Will Scaling Solve Robotics? erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

Copyright: news.mit.edu – “New AI Model Could Streamline Operations In A Robotic Warehouse”

Hundreds of robots zip back and forth across the floor of a colossal robotic warehouse, grabbing items and delivering them to human workers for packing and shipping. Such warehouses are increasingly becoming part of the supply chain in many industries, from e-commerce to automotive production.

However, getting 800 robots to and from their destinations efficiently while keeping them from crashing into each other is no easy task. It is such a complex problem that even the best path-finding algorithms struggle to keep up with the breakneck pace of e-commerce or manufacturing.

In a sense, these robots are like cars trying to navigate a crowded city center. So, a group of MIT researchers who use AI to mitigate traffic congestion applied ideas from that domain to tackle this problem.

They built a deep-learning model that encodes important information about the warehouse, including the robots, planned paths, tasks, and obstacles, and uses it to predict the best areas of the warehouse to decongest to improve overall efficiency.

Their technique divides the warehouse robots into groups, so these smaller groups of robots can be decongested faster with traditional algorithms used to coordinate robots. In the end, their method decongests the robots nearly four times faster than a strong random search method.

In addition to streamlining warehouse operations, this deep learning approach could be used in other complex planning tasks, like computer chip design or pipe routing in large buildings.

“We devised a new neural network architecture that is actually suitable for real-time operations at the scale and complexity of these warehouses. It can encode hundreds of robots in terms of their trajectories, origins, destinations, and relationships with other robots, and it can do this in an efficient manner that reuses computation across groups of robots,” says Cathy Wu, the Gilbert W. Winslow Career Development Assistant Professor in Civil and Environmental Engineering (CEE), and a member of a member of the Laboratory for Information and Decision Systems (LIDS) and the Institute for Data, Systems, and Society (IDSS).[…]

Read more: www.news.mit.edu

Der Beitrag New AI Model Could Streamline Operations In A Robotic Warehouse erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

Copyright: medicalxpress.com – “Research Team Breaks Down Musical Instincts With AI”

A team of KAIST researchers led by Professor Hawoong Jung from the Department of Physics have used an artificial neural network model to identify the principle by which musical instincts emerge from the human brain without special learning.

The research, conducted by first author Dr. Gwangsu Kim of the KAIST Department of Physics (current affiliation: MIT Department of Brain and Cognitive Sciences) and Dr. Dong-Kyum Kim (current affiliation: IBS) is published in Nature Communications under the title “Spontaneous emergence of rudimentary music detectors in deep neural networks.”

Previously, researchers have attempted to identify the similarities and differences among the music that exists in various cultures, and have tried to understand the origin of the universality. A paper published in Science in 2019 revealed that music is produced in all ethnographically distinct cultures, and that similar forms of beats and tunes are used. Neuroscientists have also learned that a specific part of the human brain, the auditory cortex, is responsible for processing musical information.

Professor Jung’s team used an artificial neural network model to show that cognitive functions for music forms spontaneously as a result of processing auditory information received from nature, without being taught music. The research team utilized AudioSet, a large-scale collection of sound data provided by Google, and taught the artificial neural network to learn the various sounds.[…]

Read more: www.medicalxpress.com

Der Beitrag Research Team Breaks Down Musical Instincts With AI erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

SwissCognitive Guest Blogger:  Zachary Amos – “How AI Can Address Global Water Scarcity Issues”

AI has become a part of everyday life, helping people work, manage their fitness and analyze their finances. However, it can also aid humanity on a grander scale, such as addressing global water scarcity.

Water scarcity affects people worldwide. Unfortunately, climate change exasperates the issue as regions experience drier and hotter conditions. AI algorithms are helping conserve this precious resource and assist those who need it most.

Understanding Water Scarcity

The Earth is made mostly of water, but only a fraction is drinkable. About 3% is freshwater, but of that percentage, only 0.5% is fit for human consumption. This becomes problematic when considering the increasing worldwide population and continued urban development, as well as the effects of climate change.

Water shortages have disastrous effects on human health. Dehydration leads to constipation, cramping and abdominal pain. About one-tenth of the world’s population lacks clean water, putting them at risk of developing dysentery, cholera and typhoid fever. Water scarcity makes agriculture challenging as well, leading to poorer harvests and nutritional deficiencies.

AI to the Rescue

Artificial intelligence has been making waves in addressing the water crisis. Governments, nonprofits and entrepreneurs are embracing AI algorithms to identify areas where water scarcity is a problem, and take action by providing better sanitation and management.

Quality and Treatment

Many shortages stem from poor water quality and contamination. It can be challenging for water boards to combat bacteria, lead and other contaminants before they sicken a population. The events that occurred in Flint, Michigan, are a perfect example.

AI can help improve water quality by using a deep-learning neural network to find harmful particles. Officials can examine water at microscopic levels in real time and devices can help them detect contamination much quicker than humanly possible. Models and conclusions derived from machine learning have been applied to all aspects of water treatment and management systems, including filtration, disinfection and desalination. AI can also provide solutions for pollution control and watershed ecosystem security management.

Conservation and Efficiency

AI has proven helpful in optimizing water usage in various sectors, such as agriculture and urban environments. It is being used for precision irrigation, leak detection and strengthening water distribution systems.

These algorithms can help farmers monitor their crops’ health and identify issues related to irrigation needs, helping them determine the best action to take. They can recommend best practices by combining and analyzing data from environmental and plant health sensors, weather forecasts, and more.

Smart meters equipped with AI can provide real-time feedback to consumers about their water usage. This helps them identify where they can reduce their consumption and actively participate in conservation.

Monitoring and Prediction

Water scarcity can best be managed by monitoring resources through satellite imagery, AI-powered sensors and remote sensing. Taking action is vital since climate change significantly contributes to water scarcity and tracking helps predict potential issues. Although some say AI is part of the problem through its intensive energy use — data centers consume about 200 terawatt hours of power annually — it’s proving valuable.

AI algorithms can analyze large data sets, and offer accurate predictions on water availability and demand. Satellite imaging technology has proven especially beneficial by providing detailed images of water sources in various areas. Researchers can examine rivers, lakes and other waterways to determine their viability.

Tackling Water Scarcity Through AI

Water scarcity is an issue everyone must address as climate change progresses and populations grow. AI is doing its part to mitigate problems and ensure people have access to clean, usable water for years to come.

About the Author:

Zachary Amos is the Features Editor at ReHack, where he writes about artificial intelligence, cybersecurity and other technology-related topics.

Der Beitrag How AI Can Address Global Water Scarcity Issues erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

SwissCognitive Guest Blogger: Sam Bowman – “3 Ways in Which AI Web Design Will Impact User Experience and Conversion” Image by Pexels

In this field, there’s a strong interaction with digital technology, the designs are often data-influenced, and projects often benefit from tools that bolster the agility of a site. Yet, it’s also important to consider what role AI platforms can have in improving the way human users interact with online materials. What advantages does the technology offer businesses in getting potential consumers further down the sales funnel?

Let’s explore a few ways AI web design will impact user experience and conversion.

Design Automation

Professional web designers and developers will likely always be the driving force in effective website creation. However, it’s important to recognize that more AI web development tools are emerging that provide a certain amount of design automation. When experienced professionals collaborate with these platforms, they have the opportunity to create a more relevant and agile online customer experience that may translate to conversions.

Some of the most useful AI platforms are those that help developers create new pages, sites, and features faster. There are now tools — like the open-source Screenshot to Code — in which designers can provide mock-up images or screenshots to a neural network trained in web development. The software can then translate the images into code, which designers will only need to do minimal testing. This means that brands can more swiftly respond to trends or changes in preferences that impact user experience.

There are also AI-automated design tools that allow for greater personalization. Algorithms can pick up the nuances of consumer behavior as they navigate your website. The AI platforms then utilize this data alongside customer profile information to adjust visual components, landing pages, and product recommendations to best suit each user. This can help customers feel they’re gaining a more relevant and unique journey, which may feed into satisfaction and conversion likelihood.

Content Provision

Including content in web design is an important part of providing a user experience that results in conversions. These materials provide value to consumers in ways that boost their engagement with a brand. To have the most impact, it is vital to adopt methods of content creation quickly without sacrificing quality. This can involve planning ahead and utilizing templates. However, AI tools are also playing a role in ensuring web designers can incorporate content in a robust and efficient way.

Completely automated content generation is not yet accurate or nuanced enough to make for solid content. There are assistive tools that can make a difference, though. AI copy-creation platforms can provide designers with some basic content ideas and text based on topic prompts. Nevertheless, these should be subject to a certain amount of professional writing and editing, especially to ensure accuracy, credibility, and adherence to the brand voice. Some platforms — such as Ryter and Article Forge — also include features such as automatically generating royalty-free images and publishing them to the site.

It’s also important to remember that the accessibility of online content can also influence consumer engagement. There are AI-driven translation tools, such as Deepl, that web designers can use to generate text in a variety of languages to suit consumers’ needs. For video and image content on the site, automatic captioning and audio description tools can be vital in improving the user experience of people with vision impairments.

Customer Service

Customer service is deeply connected to user experience and conversions. Consumers are increasingly using websites not just to shop but also to gain advice, lodge complaints, and seek refunds from brands. Utilizing AI in web design can have a positive impact by making certain customers have easy and quick access to these components.

At the moment, AI chatbots are valuable tools to integrate into websites. Bots function as automated front-line support systems and can be coded to interact using a personality reflective of the brand voice. They can answer specific types of customer questions and provide frequently-sought advice in ways that free up already busy customer service teams.

This saves time for customers, which is key to a positive user experience. However, it can also play an important part in prompting users along the customer journey by funneling them through to sales teams based on their queries. This may result in more conversions. Not to mention that data gained from interactions can trigger marketing communications with relevant product offerings, discount codes, and content.

Conclusion

Moving forward, it is also worth considering the innovative potential for AI. Developments are already demonstrating that machine learning is capable of increasingly creative and practical tasks. Consider taking the time to explore thus-far untapped uses in customer experience. After all, it’s only through experimentation that breakthroughs occur. You may find your company at the cutting-edge of consumer interactions.

About the Author:

Sam Bowman is a published freelance writer from the West Coast who specializes in healthcare tech and artificial intelligence content. His experience in patient care directly translates into his work and his passion for industry technologies influences the content he creates. Sam has worked for years – directly in, and writing about – healthcare technology and the many benefits it offers to patients and doctors alike. He loves to watch as medical tech and business software grow and develop, ushering in a modern age of industry.

Der Beitrag 3 Ways in Which AI Web Design Will Impact User Experience and Conversion erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.

SwissCognitive Guest Blogger: Zachary Amos – “How Hackers Are Using AI to Subvert Facial Recognition Technology – Generative Adversarial Networks (GAN)”

Most people have used facial recognition technology before but may not understand its potential repercussions. It seems secure and can make life easier, but there are downsides to the convenience. Technology has progressed to the point where hackers can use AI in facial recognition to commit biometric identity theft.

What Is a Generative Adversarial Network?

A generative adversarial network (GAN) uses two competing neural network models to create fake data that appears real. They continuously train through unsupervised learning until they improve. The generator model learns to generate new data, while the discriminator model attempts to determine which data is real or fake.

The process utilizes artificial intelligence (AI) and machine learning so both models learn from each other. The generator aims to improve throughout the process until it creates something the discriminator classifies as real.

How Do GANs Connect to Facial Recognition?

Many people have used facial recognition technology before because most phones unlock with it. In fact, around 73% of people feel comfortable sharing their biometric data partly because they’ve grown comfortable using it in everyday situations. It scans the face using over 30,000 infrared dots invisible to the naked eye and stores it for future comparison. The technology references the face map whenever it needs to confirm identity.

It’s widely accepted and used worldwide, even though some disagree with it. For example, law enforcement agencies in London plan on adding facial recognition cameras throughout the city despite pushback from residents. They raised questions about the security of storage, citing privacy concerns.

People may be correct to assume their privacy is at risk. Many think their faces are entirely unique, much like their fingerprints. While that’s true to an extent, GANs can trick facial recognition technology into thinking someone else is you.

How Are Hackers Tricking Facial Recognition With GANs?

AI can search through a database of faces to learn how to generate a very realistic image. It slightly morphs the real people in the photos until it creates something new. The result isn’t instantly recognizable as fake to humans, so it’s no surprise technology often can’t tell the difference.

Since GANs pit machine learning models against each other, they eventually make something capable of tricking other machine learning models. For example, it would attempt to generate a picture of a face until it looks realistic. The first few attempts may look odd and have facial features in the wrong place, but it can learn from past data and eventually create pictures that look like actual humans.

It might seem like a distant issue, but there are real-world repercussions. For example, hackers might take advantage of airports because they use facial recognition technology to identify individuals on the no-fly list. It seems secure, but there might be better options.

Although the National Institute of Standards and Technology claims its biometric tool can scan a passenger’s face once and be 99.5% or more accurate, the result can look correct when it isn’t. For example, researchers at a cybersecurity firm successfully attempted to trick an airport’s facial recognition system using GANs.

They repeatedly trained the neural networks to create a single fake face from a combination of real faces. The resulting image looked like one person to the naked eye and someone else to the system. Even though the individual was on the no-fly list, it recognized them as an entirely different person on the passport database and allowed them to board.

Why Does It Matter AI Is Subverting Facial Recognition?

Many assume biometric technology is secure because it’s tied to their appearance, but GANs have shown that is not the case. Essentially, hackers can commit identity theft and trick systems into thinking they’re someone else. Not only is this a significant security issue, but it impacts the individuals who have their biometric data misused.

Although it’s largely challenging to accomplish because it takes time and computational resources, it’s not impossible for determined cybercriminals. Considering many people are growing comfortable with facial recognition technology, biometric identity theft may become an issue.

AI Is Subverting Facial Recognition Technology

Hackers can use GANs to pose as someone else by tricking facial recognition technology. While some have privacy concerns over the widespread use and storage of biometric data, others don’t mind because it makes their life easier. No technology is inherently bad, but combining AI and biometrics might prove an issue without increased security measures.

About the Author:

Zachary Amos is the Features Editor at ReHack, where he writes about artificial intelligence, cybersecurity and other technology-related topics.

Der Beitrag How Hackers Are Using AI to Subvert Facial Recognition Technology – Generative Adversarial Networks (GAN) erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.