Hello SundAI - our world through the lense of AI

Latest Episodes:

Have you heard from Cyc? The first Human-Like AI Through Knowledge Dec 01, 2024

Show notes

Today we dive into the fascinating world of Cyc, an ambitious AI project initiated in 1984 by Douglas Lenat aimed at creating a massive knowledge base to enable human-like reasoning.

Lenat posited that achieving human-like intelligence in a machine would require several million rules, leading to the development of a knowledge database containing entries ranging from common sense to specialized expertise.

Cyc's knowledge base is built around "frames," conceptual units with slots for properties and entries for values, all organized in a global ontology and connected through a constraint language that allows for the expression of logical concepts such as quantification and disjunction. The system also employs "microtheories" to reconcile seemingly contradictory facts from different domains.

Originally funded as part of the Microelectronics and Computer Technology Corporation (MCC) consortium to counter a Japanese government computer initiative, Cyc eventually spun off into Cycorp after MCC dissolved. Cycorp has since utilized Cyc in various applications, from assisting researchers at the Cleveland Clinic to supporting US intelligence agencies in building a knowledge base on terrorism.

The podcast is based on Fisher, I. (2024, 17. April). Cyc: History’s forgotten AI project. Outsider. It can be found here.

Disclaimer: This podcast is generated by Roger Basler de Roca (contact) by the use of AI. The voices are artificially generated and the discussion is based on public research data. I do not claim any ownership of the presented material as it is for education purpose only.

There is just a small "AI Class" - Insights from the AI Proficiency Report Nov 27, 2024

Show notes

In this episode, we delve into the "AI Proficiency Report" from Section, an online business training company, which offers a compelling analysis of AI use and understanding in the workplace. Drawing on a survey of over 1,000 knowledge workers in the USA, Canada, and the UK, the report evaluates their skills based on their ability to create simple prompts for large language models (LLMs).

The findings reveal the emergence of an "AI class," consisting of about 7% of surveyed workers who use AI daily, particularly the paid versions of LLMs, integrating it effectively into their workflows and saving up to 12 hours per week. In contrast, the majority, around 57%, are "AI novices" who have only experimented with tools like ChatGPT occasionally and have not learned to use them effectively.

Key drivers of AI proficiency include employer approval of AI use, training support, and access to LLMs provided by companies or teams. Additionally, the report underscores the advantages of using paid AI tools, which correlate with higher levels of competence compared to users of free versions.

The report can be found here.

Rethinking AI Intelligence: Beyond The Turing Test Nov 24, 2024

Show notes

In this episode, we delve into David Eagleman's thought-provoking article on the measurement of intelligence in AI systems.

Eagleman critiques traditional intelligence tests like the Turing Test, introduced in 1950, which judges a machine's intelligence based on its indistinguishability from humans in conversation. He also discusses the Lovelace Test from 2003, focusing on an AI's ability to create original works. Despite their historical significance, Eagleman argues these tests fall short as they do not require creative thought processes or true originality.

Eagleman proposes a new benchmark: the Scientific Discovery Test. This test assesses AI on its ability to make scientific discoveries, divided into two levels. Level-1 discoveries involve synthesizing scattered facts from scientific literature—a task well-suited to large language models (LLMs) due to their capacity for extensive memory. However, Eagleman points out that this doesn't necessarily denote intelligence, as it largely leverages their ability to recall vast amounts of data.

More crucially, Level-2 discoveries require conceptualizing and re-conceptualizing ideas to form new world models, akin to groundbreaking theories like Einstein’s theory of relativity or Darwin’s theory of evolution through natural selection. Eagleman posits that if AI can achieve Level-2 science, it would truly match or even surpass human intelligence.

Eagleman's insights provide a fascinating glimpse into the future possibilities of AI, emphasizing the need for a more nuanced approach to measuring AI intelligence that goes beyond mere data recollection to genuine conceptual innovation.

The paper can be found here.

Disclaimer: This podcast is generated by Roger Basler de Roca (contact) by the use of AI. The voices are artificially generated and the discussion is based on public research data. I do not claim any ownership of the presented material as it is for education purpose only.

Why do large language models not understand words and characters? Nov 20, 2024

Show notes

In this episode, we tackle an intriguing aspect of artificial intelligence: the challenges large language models (LLMs) face in understanding character composition. Despite their remarkable capabilities in handling complex tasks at the token level, LLMs struggle with tasks that require a deep understanding of how words are composed from characters.

The findings reveal a significant performance gap in these character-focused tasks compared to token-level tasks. LLMs particularly struggle with understanding the position of characters within words, especially when positions are numerically specified.

This limitation is suspected to stem from the training approach of LLMs, which typically treats words as indivisible units (tokens) without considering the underlying character composition.

The episode also delves into potential solutions proposed by experts, including embedding character-level information into word embeddings and employing techniques from visual recognition to simulate human character perception.

Join us as we discuss these innovative approaches to enhancing the understanding of character composition in LLMs and their implications for the development of more nuanced and capable AI systems.

This podcast is based on Shin, A. and Kaneko, K. (2024) Large language models lack understanding of character composition of words, arXiv.org. Available at: https://arxiv.org/abs/2405.11357

Do you trust AI more than your coworker? Nov 17, 2024

Show notes

Today we explore the intricate relationship between trust in humans and trust in artificial intelligence (AI), drawing from the insightful study "On trust in humans and trust in artificial intelligence: A study with samples from Singapore and Germany extending recent research" by Montag et al. (2024). The authors delve into how trust is a crucial prerequisite for the acceptance and usage of AI technologies and how understanding this relationship can enhance AI's integration into society.

The study examines large samples from Singapore and Germany, where participants were asked about their trust in humans and AI, their personality traits using the Big Five model, and their general attitudes towards AI. Findings reveal a positive, yet varying correlation between trust in humans and AI across the two countries. In Singapore, the correlation was moderate, whereas in Germany, it was weak. The authors attribute these differences to cultural factors and suggest that trust may be interpreted differently across cultures.

This episode discusses why, despite some linkage, trust in humans and AI should largely be considered separate constructs. It also highlights the significant role cultural differences play in shaping trust in AI. By integrating these insights, the authors urge educational institutions, policymakers, and educators to consider these nuances when promoting AI technologies.

The podcast is based on Montag, C., Becker, B. and Li, B. J. (2024). From trust in humans to trust in artificial intelligence: a study of samples from Singapore and Germany that extends recent research. *Computers in Human Behavior: Artificial Humans, 2*, 100070. It can be found here.

One year later: ChatGPT in the Classroom is it revolutionizing education or stifling Learning? Nov 13, 2024

Show notes

ChatGPT offers significant advantages by enabling personalized learning experiences. It can tailor instructions to individual needs, provide round-the-clock support, and facilitate interactive learning sessions. Furthermore, it can reduce the pressure on learners by creating a safer environment for asking questions and making mistakes.

However, the authors caution against the risks of becoming overly dependent on ChatGPT. Excessive reliance may lead to diminished critical thinking, superficial engagement with learning materials, and reduced human interaction. The constant availability of answers from AI could also deter students from developing essential critical thinking and problem-solving skills necessary for academic and professional success.

In this insightful episode, we delve into the impact of ChatGPT, an advanced AI language model, on education and learning. Drawing from the study by Bai, Liu, and Su titled "ChatGPT: Cognitive Impacts on Learning and Memory," we explore both the potential benefits and the challenges of integrating ChatGPT into educational environments.

Bai L, Liu X, Su J. ChatGPT: Die kognitiven Auswirkungen auf Lernen und Gedächtnis. Brain-X. 2023;1:e30. https://doi.org/10.1002/brx2.30 it can be found here.

Disclaimer: This podcast is generated by Roger Basler de Roca (contact) by the use of AI. The voices are artificially generated and the discussion is based on public research data. I do not claim any ownership of the presented material as it is for education purpose only.

Generative AI in Academia: A Double-Edged Sword? Nov 10, 2024

Show notes

In this episode toda, we dive into the intriguing findings from the article "Is It Harmful or Helpful? Investigating the Causes and Consequences of Generative AI Use Among University Students" by Abbas, Jam, and Khan. The study focuses on why students turn to generative AI like ChatGPT for academic purposes and the implications of this usage.

The research comprises two distinct studies. The first developed a questionnaire to gauge how frequently students use ChatGPT for their studies. The second study used this questionnaire to explore how factors like workload, time pressure, reward sensitivity, and quality sensitivity influence the use of ChatGPT and its effects on students' propensity to procrastinate, experience memory loss, and perform poorly academically.

Results indicate that students are more likely to use ChatGPT under high workload and time pressure. Interestingly, those highly sensitive to rewards used it less frequently, possibly fearing poor grades if caught. Surprisingly, quality consciousness did not significantly affect ChatGPT usage.

The study also revealed that using ChatGPT likely leads to procrastination and memory loss, ultimately impairing academic performance. The authors suggest that while students may use ChatGPT to cope with high demands, this could backfire, resulting in procrastination, memory issues, and deteriorating grades.

The implications of these findings are significant for universities, policymakers, educators, and students. The authors recommend that institutions support students in efficiently managing their time and workload and encourage students to use ChatGPT as a learning supplement rather than a substitute for their own thinking. They also advise educators to develop new assessment criteria that motivate students to apply their creative abilities and critical thinking.

This podcast is based on the research of Abbas, M., Jam, F. A., & Khan, T. I. (2024). Is it harmful or helpful? Investigating the causes and consequences of generative AI use among university students. International Journal of Educational Technology in Higher Education, 21(10). It can be found here.

What is data poisoning in AI? Nov 06, 2024

Show notes

Today we delve into the hidden dangers lurking within artificial intelligence, as discussed in the paper titled "Turning Generative Models Degenerate: The Power of Data Poisoning Attacks." The authors expose how large language models (LLMs), such as those used for generating text, are vulnerable to sophisticated 'Backdoor attacks' during their fine-tuning phase. Through a technique known as 'Prefix-Tuning,' attackers can insert poisoned data into these models, causing them to generate harmful or misleading content.

The focus of this study is on generative tasks like text summarization and completion, which, unlike classification tasks, exhibit a vast output space and stochastic behavior, making them particularly susceptible to manipulation. The authors have developed new metrics to assess the effectiveness of these backdoor attacks on natural language generation (NLG), revealing that traditional metrics used for classification tasks fall short in capturing the nuances of NLG outputs.

Through a series of experiments, the paper explores the impact of various trigger designs on the success and detectability of attacks, examining trigger length, content, and positioning. Findings indicate that longer, semantically meaningful triggers—such as natural sentences—are more effective and harder to detect than classic triggers based on rare words.

Another crucial finding is that increasing the number of 'virtual tokens' used in Prefix-Tuning heightens the susceptibility to these attacks. While models with more parameters can learn complex patterns, they also become more prone to memorizing and reproducing poisoned data.

This podcast is based on the research from Jiang, S., Kadhe, S. R., Zhou, Y., Ahmed, F., Cai, L., & Baracaldo, N. (2023). Turning Generative Models Degenerate: The Power of Data Poisoning Attacks. It can be found here.

Navigating the AI Revolution: The Good, the Bad, and the Scary Nov 03, 2024

Show notes

In this thought-provoking episode, we delve into the paper "Navigating the AI Revolution: The Good, the Bad, and the Scary" which explores the multifaceted impact of artificial intelligence (AI) on our world.

AI is identified as a key driver of the Fourth Industrial Revolution, poised to revolutionize numerous facets of life.

We explore the positive and negative impacts of AI, highlighting breakthroughs such as DeepMind's AlphaFold in medicine, AI's precision in India's Chandrayaan lunar mission, and its role in combating climate change through data processing innovations.

This podcast is based on Krishna, V.V. (2024). AI and contemporary challenges: The good, bad and the scary. Journal of Open Innovation: Technology, Market, and Complexity, 10(1), 100178. https://doi.org/10.1016/j.joitmc.2023.100178

The role of AI in education, according to the World Economic Forum (WEF) Oct 30, 2024

Show notes

In this thought-provoking episode, we dive into the 2024 report by the World Economic Forum on the potential of artificial intelligence (AI) to address some of the most pressing challenges faced by educational systems globally. Titled "Shaping the Future of Learning: The Role of AI in Education 4.0," the report illustrates how AI, when effectively managed, could revolutionize the educational landscape.

We begin by examining the three major challenges currently plaguing education: a global shortage of teachers, inefficient administrative and assessment processes, and a significant digital skills gap. The report presents AI as a powerful tool capable of reshaping how education is conceptualized and delivered, offering solutions such as automating administrative tasks to free up teachers for more personalized student interaction and enhancing socio-emotional skills development.

Furthermore, AI's role in improving assessment and decision-making processes is highlighted, providing educators with timely feedback and data-driven insights to optimize teaching and learning experiences. The integration of AI in classrooms also presents a unique opportunity to educate students about AI concepts, their societal impacts, and the ethical considerations of AI development.

Personalized learning experiences are another significant advantage, with AI acting much like human tutors to tailor content and provide real-time feedback to meet individual learners' needs.

The episode also explores nine case studies from the report, showcasing the broad range of AI applications in education and demonstrating how AI is currently being used to enhance access to education and optimize learning outcomes. Highlights include initiatives like "Letrus" in Brazil, which significantly improves literacy outcomes through early intervention, and UNICEF's "Accessible Digital Textbooks," which enhance educational opportunities for children with disabilities.

This episode is based on The World Economic Forum. (2024). Shaping the Future of Learning: The Role of AI in Education 4.0. World Economic Forum. It can be found here.

Our TOPPODCAST Picks

Follow Us

Stay Connected

Hello SundAI – our world through the lense of AI

Related Podcasts

Links

Stay Connected