Ignore previous instructions (016)
Welcome to Artificial Insights: your weekly review of how to better collaborate with AI.
Double Issue. Every week I usually feature five links to articles and videos that have sparked my interest around artificial intelligence and our slow, certain transition toward AGI. This week features twice the links in order to 1) reduce my publication backlog and 2) keep you all extra-busy reading and learning about important aspects about our developing relationship to autonomous systems.
Judging from Substack stats, most of you don’t click through the links – which tells me my commentary and insight is valued more than the linked content itself. With that in mind, I have tried describing the material and value of each post with a bit more care this week. Hope this makes sense to you and helps brighten your week a little bit.
Matt Clancy and Tamay Besiroglu dive deep into AI's potential future in a lively debate. With discussions ranging from AI taking on human tasks to heralding a new era of explosive economic growth, they juxtapose caution with optimism. As AI inches closer to human-like capabilities, will we witness an unparalleled growth surge in this century? Dive in to explore these riveting perspectives.
Linus Lee takes us beyond the usual chat paradigms in his deep-dive. He stresses the need to push past current chat limitations, introducing innovative "point and select" interfaces and emphasizing the blend of context and interactivity. With five compelling strategies, Lee promises dialogue interfaces that are both intuitive and packed with the magic of language models.
Andrew Ng busts myths around AI’s perceived dangers. Through intriguing insights, like a study on the game model Othello, Ng underscores AI’s potential to tackle pressing global challenges, from pandemics to climate change. While acknowledging risks like deep fakes, Ng stresses the merits of an open-source approach.
Professor Michael Levin dives deep into the mesmerizing world of developmental bioelectrics. With examples from flatworms to axolotls, he showcases the intelligence of cells and their bioelectric decision-making. His research paves the way for potential breakthroughs in birth defects, regenerative medicine, and AI.
In this inspiring video, Aza Raskin, from Earth Species Project, takes us on a journey to understand animal talk through the lens of AI. By bridging human and animal worlds, he paints a hopeful future where we decode animal languages, enriching conservation research and deepening our cosmic connection. With potential as profound as Hubble's revelations, the endeavor also prompts us to tread responsibly, ensuring we respect, not disrupt, animal cultures.
Cognitive scientist Joscha Bach embarks on a captivating exploration of consciousness and AI. From the profound "longest possible game" to the potential of a "biological internet," Bach touches on the interplay of beliefs, self-realization, and telepathy. He cuts through the AI buzz, presenting a unique vision of AGI living harmoniously with life, enriching complexity and awareness.
Satya Nadella and Sam Altman, leading voices from Microsoft and OpenAI, delve deep into AI's vast universe. From its intimate role in our lives to the looming concerns of job loss and ethical dilemmas, they navigate the terrain, emphasizing the need for education, careful governance, and global regulations.
Ben Goertzel, a trailblazer in AGI, sheds light on the monumental shift from basic AI to the vast landscape of AGI. With candor, he addresses societal challenges like job losses and nation disparities. Yet, he paints a hopeful picture of AGI's potential to transform societies. Dive into the nuances of recursive self-improvement, grasp the intricacies of intelligence, and understand the essence of embodied communication and the heart of social intelligence.
Long interview where Anthropic CEO, Dario Amodei, unpacks the layers of AI from model scaling to the surprises within emergent abilities. With a clear stress on empirical safety checks and real-world AI impacts, Amodei voices concerns on potential pitfalls and emphasizes talent density's role. The video serves as a deep-dive into the fast-paced AI universe, offering a blend of caution, experience, and optimism.
This insightful video dives into neural networks, portraying them as master sculptors of complex functions. From the basics of training to using Fourier features for detailed data capture, it demonstrates their artistry, even in approximating the intricate Mandelbrot set. While highlighting both prowess and limitations, the takeaway is clear: one size doesn't fit all in function approximation.
The guidance or input provided to a system, typically with the intention to confuse, mislead, or exploit it in some way. While the term can be applied generally across various systems, it's frequently discussed in the context of machine learning models and AI systems. Adversarial instructions might involve phrasing questions or inputs in deceptive ways, trying to trick the model into giving incorrect or unintended responses. (AKA Jailbreaking)
Long Read from Substack
If Artificial Insights makes sense to you, please help us out by:
Subscribing to the weekly newsletter on Substack.
Following the weekly newsletter on LinkedIn.
Forwarding this issue to colleagues and friends wanting to learn about AGI.
Sharing the newsletter on your socials.
Commenting with your favorite talks and thinkers.
Artificial Insights is written by Michell Zappa, CEO and founder of Envisioning, a technology research institute.
You are receiving this newsletter because you signed up on envisioning.io.