Machine learning, a branch of artificial intelligence, utilizes artificial neural networks inspired by the structure and function of the human brain. In these networks, individual nodes represent neurons, and connections between nodes mimic synapses. Hopfield and Hinton have both made significant contributions to advancing this technology since the 1980s.

John Hopfield, a professor at Princeton University, created an associative memory within an artificial neural network that can store and reconstruct images and patterns in data. His network, known as the Hopfield network, uses principles from physics to describe the energy of the system. By finding values for the connections between nodes, the network is trained to minimize the system’s energy, resulting in the recovery of the stored images from distorted or incomplete input.

Geoffrey Hinton, a professor at the University of Toronto, built upon Hopfield’s work and developed the Boltzmann machine. Using tools from statistical physics, the Boltzmann machine can recognize characteristic elements in a given type of data. Hinton trained this machine by providing it with examples that are likely to occur in real-world scenarios. The Boltzmann machine can then be utilized to classify images or generate new examples of the trained pattern type.

The impact of Hopfield and Hinton’s work extends far beyond the field of physics. These artificial neural networks have found applications in a wide range of areas, including the development of new materials with specific properties. The laureates’ contributions have been of the utmost benefit and have driven the explosive growth of machine learning in recent years.

The Nobel Prize in Physics comes with a prize amount of 11 million Swedish kronor, which will be equally shared between Hopfield and Hinton. Their work exemplifies the ability of physics and mathematics to shape and advance technology, benefiting society on a global scale.

Related: Ambros & Ruvkun’s MicroRNA Breakthrough Wins 2024 Nobel Prize in Physiology & Medicine

About Laureates:

John J. Hopfield:
Born 1933 in Chicago, IL, USA
PhD 1958 from Cornell University, Ithaca, NY, USA
Professor at Princeton University, NJ, USA

Geoffrey E. Hinton:
Born 1947 in London, UK
PhD 1978 from The University of Edinburgh, UK
Professor at the University of Toronto, Canada

Related: Geoffrey Hinton’s Red Flags: Uncovering the Risks in AI’s Future

About the Nobel Prize:

The Nobel Prize is annually awarded by The Royal Swedish Academy of Sciences. The Academy, founded in 1739, is an independent organization dedicated to promoting the sciences and strengthening their influence in society. While it focuses on the natural sciences and mathematics, the Academy actively encourages interdisciplinary collaboration and the exchange of ideas between different fields. For further information, visit www.kva.se and www.nobelprize.org

Here’s an official press release: The Nobel Prize in Physics 2024

Mohsin Rasheed

Co-Founder & Chief Editor of Everyman Science. I view science not just as a collection of facts, but as the ultimate guide for human survival. From medical breakthroughs to the logistics of space exploration, I am dedicated to documenting how scientific reasoning uplifts the human spirit and provides the blueprints to save our planet. I believe that by unleashing the power of nature through disciplined inquiry, we can secure a sustainable future for humanity.

The post Hopfield and Hinton Win 2024 Nobel Prize for Breakthroughs in Machine Learning appeared first on Everyman Science.

The post Hopfield and Hinton Win 2024 Nobel Prize for Breakthroughs in Machine Learning appeared first on Everyman Science.

The stock prices of other US-listed companies, such as semiconductor manufacturer Advanced Micro Devices and AI firm C3.ai, also increased as investors seek to capitalize on the booming AI sector. However, Matt Wacher, the chief investment officer at Morningstar Asia Pacific, cautions that while there can be significant rewards for investing in tech stocks, there are also substantial risks involved.

Nvidia co-founder and CEO Jensen Huang. Image: Bloomberg

Global AI Market to Hit $207B by 2022

A new report by Grand View Research estimates that the global AI market is expected to reach an impressive $US136.55 billion ($207 billion) in 2022, and is projected to grow at an average compound annual growth rate of 37.3 per cent from 2023 to 2030. This remarkable growth is attributed to the continuous research and innovation efforts of tech giants, which are driving the adoption of advanced technologies across various industry verticals such as automotive, healthcare, retail, finance and manufacturing. AI technology has the potential to revolutionize these industries by providing more efficient and cost-effective solutions for businesses. As such, it is likely that AI will continue to be a major driver of economic growth in the coming years.

Alphabet’s Chatbot Failure

Though the potential of Artificial Intelligence (AI) may seem like a surefire investment for some, it is important to remember that tech has seen some major downturns in the past. The most notable example of this was during the “tech wreck” of early 2000, when many companies saw their share prices plummet drastically. This was demonstrated again last year when Alphabet [owner of Google] saw $US100 billion wiped from its value after its chatbot failed to meet expectations during its first demonstration. This serves as a reminder that even the most promising technologies can experience significant setbacks, and investors should be aware of the risks associated with investing in AI.

Invest Beyond AI

For those looking to invest in Artificial Intelligence (AI), Cameron Gleeson, senior investment strategist at exchange-traded funds provider Betashares, recommends diversifying your portfolio by investing in a well-constructed basket of stocks. This will help reduce the risk of concentrated bets on specific AI names. There are several Exchange Traded Funds (ETFs) listed on the Australian Securities Exchange that track an index of technology stocks, which can provide investors with a way to spread their investment risk. Morningstar’s Wacher also suggests that investors should only allocate a small portion of their portfolio to higher-risk sectors such as AI. By doing so, investors can benefit from the potential upside of investing in AI while mitigating the risks associated with it.

admin

www.everymansci.com

The post Nvidia’s Soaring Stock Price Triggers Warning to Investors Eyeing the AI Boom appeared first on Everyman Science.

The post Nvidia’s Soaring Stock Price Triggers Warning to Investors Eyeing the AI Boom appeared first on Everyman Science.

EU’s AI Act

The EU’s AI Act is set to become the first law on AI by a major regulator anywhere, according to its website. It focuses on regulating AI and protecting Europeans from certain AI risks, which are divided into three categories. The European parliament voted in favor of adopting the act by a large majority and June 14 has been set as the tentative date for its adoption.

OpenAI’s Systems and the AI Act

Altman is reportedly concerned that OpenAI’s systems such as ChatGPT and GPT-4 could be designated as “high risk” under the regulation, according to Time. This would mean that the company would have to meet certain requirements over safety and transparency, such as disclosing that its content was AI-generated. The highest risk category of the AI Act would be banned, while systems in the second risk category would be subject to specific legal requirements and those in the third category would be largely left unregulated. Additionally, companies will have to design their AI models so they don’t generate illegal content and publish summaries of copyrighted data used for training.

OpenAI’s GPT-4 Model

When OpenAI released GPT-4 in March, some in the AI community were disappointed that OpenAI did not disclose information on what data was used to train the model, how much it cost, and how it was created. Ilya Sutskever, OpenAI’s cofounder and chief scientist, previously told The Verge that this was due to competition and safety concerns. He said that it took almost all of OpenAI working together for a very long time to produce GPT-4 and there are many companies who want to do the same thing. He also noted that while competition is top-of-mind now, safety will become more important in the future.

Also Read: Chat GPT-4 vs. Its Predecessor: What You Need to Know

Altman’s Suggestion for Government Oversight

Altman has expressed his concern about how the AI Act will affect OpenAI’s presence in Europe but he recently told US Senate members that there should be a government agency overseeing AI projects with “above a certain scale of capabilities.” He suggested granting licenses to AI companies and taking them away if they overstep safety rules.

It remains unclear how exactly these proposed regulations will affect OpenAI’s operations in Europe but one thing is certain: The details really do matter when it comes to regulating artificial intelligence. Companies like OpenAI must ensure they comply with all safety requirements or face potentially serious consequences if they don’t.

admin

www.everymansci.com

The post OpenAI’s Sam Altman Warns of Ceasing Operations in Europe appeared first on Everyman Science.

The post OpenAI’s Sam Altman Warns of Ceasing Operations in Europe appeared first on Everyman Science.

Scholar warns against placing trust in unreliable and easily manipulated detectors.

Study Findings

A new study from Stanford scholars has revealed that AI detectors, designed to detect content written by AI, are biased against non-native English writers. The research found that the detectors were near-perfect in evaluating essays written by U.S.-born eighth-graders, but classified more than half of TOEFL essays (61.22%) written by non-native English students as AI-generated. In addition, all seven AI detectors unanimously identified 18 of the 91 TOEFL student essays (19%) as AI-generated and a remarkable 89 of the 91 TOEFL essays (97%) were flagged by at least one of the detectors.

Image: Bias in GPT detectors against non-native English writing samples.

Issue with AI Detectors

The issue lies in how these detectors detect AI; they typically score based on a metric known as ‘perplexity’, which correlates with the sophistication of the writing – something in which non-native speakers are naturally going to trail their U.S.-born counterparts. As such, this raises serious questions about the objectivity of Artificial Intelligence detectors and the potential for foreign-born students and workers to be unfairly accused or penalized for cheating.

Potential Subversion of AI Detectors

The findings of the study have indicated that these detectors can be easily circumvented by the use of prompt engineering, which involves the application of generative AI to rephrase essays in order to incorporate more elaborate vocabulary. This renders them even less dependable and raises further moral quandaries.

Recommendations

Professor James Zou, a senior author of the study and a Professor of Biomedical Data Science at Stanford University, has recommended that we abstain from relying on these detectors in educational settings with a high concentration of non-native English speakers until they have been thoroughly assessed and improved. Additionally, he has proposed that developers move beyond using perplexity as their primary metric and instead implement watermarks in which generative AI embeds subtle indicators of its identity into the content it generates.

Access the complete study by clicking here.

admin

www.everymansci.com

The post Stanford Scholars find AI Detectors Unfairly Penalize Non-Native English Speakers appeared first on Everyman Science.

The post Stanford Scholars find AI Detectors Unfairly Penalize Non-Native English Speakers appeared first on Everyman Science.

Unlike conventional machine learning systems that require massive amounts of computation to fine-tune an AI system, this electrical circuit teaches itself. It offers a potential solution to a significant roadblock in the development of increasingly complex AI programs. The research was presented by physicist Samuel Dillavou at the annual meeting of the American Physical Society held in Chicago. This marks a significant step forward in our understanding of how machines learn.

It’s independent – doesn’t rely on computers

Dillavou emphasizes that the study is a “proof of principle,” highlighting the potential for further research in this field. The findings demonstrates capabilities of electrical circuits previously unknown. They suggest that there may be alternative methods to teach AI systems that do not rely on computers.

The implications of this research are significant for the future of AI. As AI programs grow more complex, the amount of computation required to fine-tune them grows exponentially. However, the ability of this simple electrical circuit to learn independently could pave the way for more efficient and sustainable approaches to developing advanced AI systems.

A Self-Learning Physical System

Artificial neural networks have revolutionized the field of machine learning. But, they come with a significant challenge. They exist only in the digital world, and their training requires extensive computation. This limitation has driven physicists to explore AI electrical circuit that can tune themselves without external computation. This leads to a groundbreaking discovery by Samuel Dillavou and his team that could transform how we teach machines.

Their system uses simple electrical components to tune itself without external help from a computer. The researchers created a network using 16 adjustable resistors, and the system learned to produce the desired output, teaching itself how to perform a given task. What’s remarkable about this system is that it mimics how a living brain learns. It outperformed traditional machine learning systems in its ability to teach itself.

Artificial neural networks rely on connections between nodes, and the optimization problem becomes daunting with the size of the network. Dillavou’s self-taught system could be a solution to this challenge. This model uses physical systems that can efficiently tune themselves, minimizing the amount of computing and memory required to train a such a system. This breakthrough could lead to more complex and efficient AI systems.

Bringing AI Closer to the Future

The potential applications of this discovery are vast. Machines that learn like humans, with the ability to adapt and improve their performance over time, could transform multitude of industries.

“If it’s made out of electrical components, then you should be able to scale it down to a microchip. I think that’s where they’re going with this.”

Jason Rocks, a physicist at Boston University

This breakthrough could pave the way for the creation of more efficient and intelligent machines, bringing us one step closer to the future of AI.

Is this AI electrical circuit different from Traditional Machines?

Unlike traditional machine learning systems, this system mimics how a living brain learns, teaching itself how to perform a given task without requiring vast amounts of computation. Using a network of 16 adjustable resistors, each acting as an edge in the network, the team was able to create a system that was capable of producing desired output by adjusting the resistors.

“It’s a proof of principle. We are learning something about learning.”

According to Samuel Dillavou, the lead researcher on the project.

The system was tested on a variety of simple AI tasks, including the ability to distinguish between three species of iris. Results showed 95% accuracy based on physical measurements of the flower. This is a canonical AI test that traditionally requires a standard catalog of 150 sets of measurements; with 30 used to train the network. It shows this approach may be highly effective.

What’s Next?

While it’s unlikely that the AI electrical circuit network will replace standard neural networks, which can make fine distinctions with great accuracy, the discovery has enormous potential for developing more efficient and intelligent machines. As physicist Jason Rocks notes, “If it’s made out of electrical components, then you should be able to scale it down to a microchip. I think that’s where they’re going with this.”

The ability to create physical systems that can teach themselves has the potential to transform whole industries. This is paving the way for the development of machines that learn as humans do. A significant step forward in the development of more complex and efficient AI systems – with their endless usecases.

admin

www.everymansci.com

The post Self-Learning electrical circuit: A groundbreaking AI discovery appeared first on Everyman Science.

The post Self-Learning electrical circuit: A groundbreaking AI discovery appeared first on Everyman Science.