Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog – Page 3

Large Language Models (LLMs) both threaten the uniqueness of human social intelligence and promise opportunities to better understand it. In this talk, I evaluate the extent to which distributional information learned by LLMs allows them to approximate human behavior on tasks that appear to require social intelligence. In the first half, I will compare human and LLM responses in experiments designed to measure theory of mind—the ability to represent and reason about the mental states of other agents. Second, I present the results of an evaluation of LLMs using the Turing test, which measures a machine’s ability to imitate humans in a multi-turn social interaction.

Cameron Jones recently graduated with a PhD in Cognitive Science from the Language and Cognition Lab at UC San Diego. His work focuses on comparing humans and Large Language Models (LLMs) to learn more about how each of those systems works. He is interested in the extent to which LLMs can explain human behavior that appears to rely on world knowledge, reasoning, and social intelligence. In particular, he is interested in whether LLMs can approximate human social behavior, for instance in the Turing test, or by persuading or deceiving human interlocutors.

https://camrobjones.com/

https://scholar.google.com/citations?… / camrobjones.

/ camrobjones.

Read more | >

Convergent engagement of neural and computational sciences and technologies are reciprocally enabling rapid developments in current and near-future military and intelligence operations. In this podcast, Prof. James Giordano of Georgetown University will provide an overview of how these scientific and technological fields can be — and are being — leveraged for non-kinetic and kinetic what has become known as cognitive warfare; and will describe key issues in this rapidly evolving operational domain.

James Giordano PhD, is the Pellegrino Center Professor in the Departments of Neurology and Biochemistry; Chief of the Neuroethics Studies Program; Co-director of the Project in Brain Sciences and Global Health Law and Policy; and Chair of the Subprogram in Military Medical Ethics at Georgetown University Medical Center, Washington DC. Professor Giordano is Senior Bioethicist of the Defense Medical Ethics Center, and Adjunct Professor of Psychiatry at the Uniformed Services University of Health Sciences; Distinguished Stockdale Fellow in Science, Technology, and Ethics at the United States Naval Academy; Senior Science Advisory Fellow of the SMA Branch, Joint Staff, Pentagon; Non-resident Fellow of the Simon Center for the Military Ethic at the US Military Academy, West Point; Distinguished Visiting Professor of Biomedical Sciences, Health Promotions, and Ethics at the Coburg University of Applied Sciences, Coburg, GER; Chair Emeritus of the Neuroethics Project of the IEEE Brain Initiative; and serves as Director of the Institute for Biodefense Research, a federally funded Washington DC think tank dedicated to addressing emerging issues at the intersection of science, technology and national defense. He previously served as Donovan Group Senior Fellow, US Special Operations Command; member of the Neuroethics, Legal, and Social Issues Advisory Panel of the Defense Advanced Research Projects Agency (DARPA); and Task Leader of the Working Group on Dual-Use of the EU-Human Brain Project. Prof. Giordano is the author of over 350 peer-reviewed publications, 9 books and 50governmental reports on science, technology, and biosecurity, and is an elected member of the European Academy of Science and Arts, a Fellow of the Royal Society of Medicine (UK), and a Fulbright Professorial Fellow. A former US Naval officer, he was winged as an aerospace physiologist, and served with the US Navy and Marine Corps.

Read more | >

Originally released December 2023._ In today’s episode, host Luisa Rodriguez speaks to Nita Farahany — professor of law and philosophy at Duke Law School — about applications of cutting-edge neurotechnology.

They cover:
• How close we are to actual mind reading.
• How hacking neural interfaces could cure depression.
• How companies might use neural data in the workplace — like tracking how productive you are, or using your emotional states against you in negotiations.
• How close we are to being able to unlock our phones by singing a song in our heads.
• How neurodata has been used for interrogations, and even criminal prosecutions.
• The possibility of linking brains to the point where you could experience exactly the same thing as another person.
• Military applications of this tech, including the possibility of one soldier controlling swarms of drones with their mind.
• And plenty more.

In this episode:
• Luisa’s intro [00:00:00]
• Applications of new neurotechnology and security and surveillance [00:04:25]
• Controlling swarms of drones [00:12:34]
• Brain-to-brain communication [00:20:18]
• Identifying targets subconsciously [00:33:08]
• Neuroweapons [00:37:11]
• Neurodata and mental privacy [00:44:53]
• Neurodata in criminal cases [00:58:30]
• Effects in the workplace [01:05:45]
• Rapid advances [01:18:03]
• Regulation and cognitive rights [01:24:04]
• Brain-computer interfaces and cognitive enhancement [01:26:24]
• The risks of getting really deep into someone’s brain [01:41:52]
• Best-case and worst-case scenarios [01:49:00]
• Current work in this space [01:51:03]
• Watching kids grow up [01:57:03]

The 80,000 Hours Podcast features unusually in-depth conversations about the world’s most pressing problems and what you can do to solve them.

Learn more, read the summary and find the full transcript on the 80,000 Hours website:

Nita Farahany on the neurotechnology already being used to convict criminals and manipulate workers

Read more | >

Check out my quantum mechanics course on Brilliant! First 30 days are free and 20% off the annual premium subscription when you use our link ➜ https://brilliant.org/sabine.

In 2020, a group of MIT researchers detected an anomaly in the nuclei of ytterbium atoms. They said that the nuclei’s strange behavior might be indicative of a “dark force” caused by a currently-undiscovered mystery particle that might make up dark matter. In 2020, the anomaly only had a significance of 3 sigma. But now, another group has confirmed it at a whopping 23 sigma! What does that mean for physics? Let’s find out.

Paper: https://journals.aps.org/prl/abstract… Check out my new quiz app ➜ http://quizwithit.com/ 💌 Support me on Donorbox ➜ https://donorbox.org/swtg 📝 Transcripts and written news on Substack ➜ https://sciencewtg.substack.com/ 👉 Transcript with links to references on Patreon ➜ / sabine 📩 Free weekly science newsletter ➜ https://sabinehossenfelder.com/newsle… 👂 Audio only podcast ➜ https://open.spotify.com/show/0MkNfXl… 🔗 Join this channel to get access to perks ➜ / @sabinehossenfelder 🖼️ On instagram ➜ / sciencewtg #science #sciencenews #physics #darkmatter.

🤓 Check out my new quiz app ➜ http://quizwithit.com/
💌 Support me on Donorbox ➜ https://donorbox.org/swtg.
📝 Transcripts and written news on Substack ➜ https://sciencewtg.substack.com/
👉 Transcript with links to references on Patreon ➜ / sabine.
📩 Free weekly science newsletter ➜ https://sabinehossenfelder.com/newsle…
👂 Audio only podcast ➜ https://open.spotify.com/show/0MkNfXl…
🔗 Join this channel to get access to perks ➜
/ @sabinehossenfelder.
🖼️ On instagram ➜ / sciencewtg.

#science #sciencenews #physics #darkmatter

Read more | >

By taking two flakes of special materials that are just one atom thick and twisting them at high angles, researchers at the University of Rochester have unlocked unique optical properties that could be used in quantum computers and other quantum technologies.

Read more | >

As the world increasingly prioritizes sustainable energy solutions, solar power stands out as a leading candidate for clean energy generation. However, traditional solar cells have encountered several challenges, particularly regarding efficiency and stability. But what if there was a better alternative? Imagine a solar cell that is affordable, more stable and highly efficient. Does it sound like science fiction? Not anymore. Meet SrZrSe3 chalcogenide perovskite, a rising star in the world of photovoltaics.

Our research team at the Autonomous University of Querétaro in Mexico has recently unveiled a solar cell crafted from a unique material called SrZrSe3. This novel approach is turning heads in the pursuit of affordable and efficient solar energy.

For the first time, we have successfully integrated advanced inorganic metal sulfide layers, known as hole transport layers (HTLs), with SrZrSe3 using SCAPS-1D simulations. Our work, published in Energy Technology, has significantly raised the (PCE) to an impressive rate of more than 27%, marking an advancement in solar technology.

Read more | >

As the global demand for sustainable energy solutions continues to grow, Lithuanian researchers have taken a step forward by developing a technology that not only transforms waste into valuable hydrogen but also eliminates a long-standing issue in gasification—the presence of tar. This new method offers an efficient and eco-friendly way to produce high-purity hydrogen from various waste materials, representing a significant advancement in clean energy production.

Hydrogen is a key element in the transition to cleaner energy. However, conventional gasification methods are often unable to ensure its high purity—synthesis gases contain very low concentrations of hydrogen.

This inefficiency limits the industrial application of hydrogen as a clean gas fuel, highlighting the need for more advanced production methods.

Read more | >

Reinforcement learning (RL) has become central to advancing Large Language Models (LLMs), empowering them with improved reasoning capabilities necessary for complex tasks. However, the research community faces considerable challenges in reproducing state-of-the-art RL techniques due to incomplete disclosure of key training details by major industry players. This opacity has limited the progress of broader scientific efforts and collaborative research.

Researchers from ByteDance, Tsinghua University, and the University of Hong Kong recently introduced DAPO (Dynamic Sampling Policy Optimization), an open-source large-scale reinforcement learning system designed for enhancing the reasoning abilities of Large Language Models. The DAPO system seeks to bridge the gap in reproducibility by openly sharing all algorithmic details, training procedures, and datasets. Built upon the verl framework, DAPO includes training codes and a thoroughly prepared dataset called DAPO-Math-17K, specifically designed for mathematical reasoning tasks.

DAPO’s technical foundation includes four core innovations aimed at resolving key challenges in reinforcement learning. The first, “Clip-Higher,” addresses the issue of entropy collapse, a situation where models prematurely settle into limited exploration patterns. By carefully managing the clipping ratio in policy updates, this technique encourages greater diversity in model outputs. “Dynamic Sampling” counters inefficiencies in training by dynamically filtering samples based on their usefulness, thus ensuring a more consistent gradient signal. The “Token-level Policy Gradient Loss” offers a refined loss calculation method, emphasizing token-level rather than sample-level adjustments to better accommodate varying lengths of reasoning sequences. Lastly, “Overlong Reward Shaping” introduces a controlled penalty for excessively long responses, gently guiding models toward concise and efficient reasoning.

Read more | >

Researchers at the University of Liverpool and the University of Southampton have used computational design methods to develop non-metal organic porous framework materials, with potential applications in areas such as catalysis, water capture or hydrogen storage.

In a study published in the journal Nature, the research team used inexpensive and abundant non-metallic elements, such as , to design non-metal organic porous frameworks (N-MOFs).

The new materials offer an alternative to (MOFs), a class of porous, crystalline materials made up of metals connected by organic linker compounds.

Read more | >

Hydrogen is often seen as the fuel of the future on account of its zero-emission and high gravimetric energy density, meaning it stores more energy per unit of mass compared to gasoline. Its low volumetric density, however, means it takes up a large amount of space, posing challenges for efficient storage and transport.

In order to address these deficiencies, hydrogen must be compressed in tanks to 700-bar pressure, which is extremely high. This situation not only incurs but also raises safety concerns.

For hydrogen-powered fuel-cell vehicles (FCVs) to become widespread, the US Department of Energy (DOE) has set specific targets for : 6.5% of the storage material’s weight should be hydrogen (gravimetric storage capacity of 6.5 wt%), and one liter of storage material should hold 50 grams of hydrogen (a volumetric storage capacity of 50 g L‒1). These targets ensure that vehicles can travel reasonable distances without excessive fuel.

Read more | >