cool work! this feels related to https://arxiv.org/abs/2304.11082 - what are your thoughts on the connection?
I find this paper mostly misleading. It assumes that the LLM is initially 99% certain to be friendly and 1% certain to me "malicious", and that "friendly" and "malicious" can be distinguished if you have a long enough prompt (more precisely, at no point have you gathered so much evidence for or against being malicious that you prob would not go up and down based on new information). Assuming those, it's pretty obvious that the LLM will say bad things if you have a long enough prompt.
The result is not very profound, and I like this paper mostly as a formali...
For what it's worth, I think most people I know expect most professed values to be violated most of the time, and so they think that libertarians advocating for this is perfectly ordinary; the surprising thing would be if professed libertarians weren't constantly showing up advocating for regulating things. Show don't tell in politics and ideology. That's not to say professing values is useless, just that there's not an inconsistency to be explained here, and if I link people in my circles this post, they'd respond with an eyeroll at the possibility that if only they were more libertarian they'd be honest - because the name is most associated with people using the name to lie.
it only works when you are able to reduce social anxiety by showing that they're welcome. someone who is cripplingly anxious typically wants to feel like they're safe, so showing them a clearer map to safety includes detecting the structure of their social anxiety first and getting in sync with it. then you can show them they're welcome in a way that makes them feel safer, not less. to do this requires gently querying their anxiety's agentic target and inviting the group to behave in ways that satisfy what their brain's overactivation wants.
I think the only content left would be the actual art. not the stuff that only deserves the name content.
Sam Altman's world tour has highlighted both the promise and risks of AI. While AI could solve major issues like climate change, super intelligence poses existential risks that require careful management. Current AI models may still provide malicious actors with expertise for causing mass harm. OpenAI aims to balance innovation with addressing risks, though some regulation of large models may be needed. Altman believes AI will be unstoppable and greatly improve lives, but economic dislocation from job loss will be significant and AI may profoundly change o...
the resolution criteria of a bet should not rely heavily on reasonableness of participants unless the bet is very small such that both parties can tolerate misresolution. the manifold folks can tell you all about how it goes when you get this wrong, there are many seemingly obvious questions that have been derailed by technicalities, and it was not the author's reasonableness most centrally at play. (edit: in fact, the author's reasonableness is why the author had to say "wait... uh... according to those criteria this pretty clearly went x way, which I didn't expect and so the resolution criteria were wrong")
Collective intelligence, which is groups of individuals acting together in intelligent ways, may be more intelligent than individuals. Tom Malone studies how to design groups for maximum effectiveness. To build a science of collective intelligence, they need to measure and develop theories of how it works. They can build on knowledge from many disciplines. To create a design space, they characterize different tasks and processes groups can use. Experiments can test how different processes work on tasks. Theories are modified based on results, guiding futur...
Gillian Hadfield discusses the importance of cooperative intelligence and normative systems for AI. She argues that humans have evolved the ability to create and enforce norms through third-party punishment, which allows for stable groups and cooperation. However, current AI approaches focus too much on individual optimization. Instead, AI systems should learn to participate in and maintain normative infrastructure, rather than simply mimic existing human behavior. Understanding the generative process behind human norms and the role of normative reasoning ...
Evolutionary game theory studies how strategies evolve and change over time, unlike classical game theory which focuses on static strategies. Natural selection, not rational choice, drives the evolution of strategies in biological systems. Initially, defectors outcompete cooperators but repeated interactions allow cooperation to evolve through strategies like tit-for-tat and generous tit-for-tat. Indirect reciprocity through reputation systems also enables cooperation in larger groups where people do not interact repeatedly. The evolution of cooperation th...
Survival of the fittest does not preclude altruism in nature. Simulations show that unconditionally sacrificing offspring for others does not work in the long run. For altruism to evolve, there must be some benefit to copies of the altruistic gene. Kin selection, where creatures help family members who likely share the same genes, can allow altruistic genes to spread through a population if the benefit of helping outweighs the cost. While the genes are selfish in seeking to replicate, the altruistic behavior of the creatures themselves is genuine.
The video argues that corporations have co-opted identity politics for their own interests, pretending to care about social justice issues while doing little to actually help marginalized groups. True progress requires solidarity and constructive politics that focus on positive outcomes for working people, uniting them against corporate elites. While identity and acknowledging differences are important, we must identify our shared adversaries and mutual interests to build solidarity and achieve true economic justice for all. Corporate capture of identity p...
The video provides advice and coping mechanisms for the author's past self to be more productive and happy. These include writing everything down, using a calendar for events, accepting one's autism diagnosis [lol @ how the ai generalized this], asking for clarification when communication is unclear, organizing notes with hyperlinks, and using timers and the Pomodoro Technique to structure work. An interesting point is that the author recommends storing one's "brain" in plain text files instead of proprietary apps to ensure longevity. Externalizing informa...
The video discusses the challenges of integrating knowledge across disciplines in research teams. Despite many attempts over the past 20 years, most interdisciplinary research remains multi-disciplinary with little true integration. The speaker argues that early interactions in interdisciplinary teams are crucial to develop a shared conceptualization and become a complex system capable of true knowledge integration. If teams allocate sufficient time in their first meetings to develop a co-created understanding of the research problem, it can help overcome ...
The video discusses 3 proven study techniques backed by scientific research: 1) Testing yourself early and often, even if you get answers wrong initially, to take advantage of the hypercorrection effect and better retain information. 2)Spacing out study sessions over time to improve long-term retention. 3) Interleaving different topics during study to develop broader strategies and more flexible knowledge. Interleaving, though more frustrating, leads to significantly better performance.The video is sponsored by the Bill and Melinda Gates Foundation, which ...
Michael Levin discusses how cells can organize themselves into complex structures through bioelectricity and collective intelligence. Cells have the ability to regenerate and adapt to perturbations through electrical networks that store and process information. His research aims to understand and manipulate this bioelectric software to control cell behavior and form complex structures. He demonstrates how altering the bioelectric patterns in flatworms can cause them to regenerate heads of different shapes and species, showing that cells can achieve differe...
This is the video transcript, not a summary.
0:01 Imagine two identical circles, and two balls - one in each - placed at almost, but not
0:07 exactly the same position.
0:11 Now, let them fall.
0:35 At first, they appear to be following the same path - but soon, their trajectories will diverge. N hits
1:04 This is a chaotic system.
1:06 And these tend to have rather beautiful patterns.
1:08 So let's try to visualize it!
1:12 To colour in a point, let a ball drop from it.
1:26 And read off the colour after the N-th hit. Time, revert
2:05 To put non-integer values of ...
Consciousness goes away when we sleep or are under anesthesia and comes back when we are awake. Consciousness is not the same as our senses or cognitive functions like memory and thinking, as shown by people who lack senses but are still conscious. Clive Wearing, who loses his memory every 30 seconds, shows that consciousness is not continuous, suggesting it is a momentary state. After removing all senses, functions and self, we are left with a bare nugget of being that is our raw consciousness, the closest we can get to understanding what consciousness tr...
AI art generators can produce novel and creative images by exploring the vast space of all possible images. While not at the same level as human artists, they can combine styles in new ways and make interesting mistakes that spark the imagination. They are trained on human creativity found in the data they learn from, imitating and reflecting human art. However, they lack human intent, expression and lived experience. When paired with a human, AI art can become a collaborative tool for exploration and expression of new kinds of art. The purpose of AI art i...
The video argues that movies have shifted from a modernist to a postmodernist and now to a metamodernist style. Modernist movies had straightforward stories and advocated for specific values. Postmodernist films questioned narrative itself and deconstructed traditional storytelling. Now, metamodern films incorporate elements of both modernism and postmodernism, oscillating between sincerity and deconstruction. They use meta elements and references not just to deconstruct but also to find meaning. This metamodern shift reflects broader cultural changes and ...
AI art has faced pushback for being built on stolen art without artists' consent. While AI can be used as a creative tool, many worry corporations will use it to cut costs by replacing human artists. There are concerns that media saturated with AI generations, driven by profit motives, could strangle human creativity. However, AI could also augment human creativity if used as a tool. The key issue is how AI is created and used, and people need to remain vigilant to ensure it is integrated ethically into society.
Daniel Dennett discusses the dangers of counterfeit people created by AI. While current AI may not be perfectly human-like, it is good enough to fool many people. This could undermine trust and communication on the internet. As AI improves, it will become harder to distinguish text generated by humans versus AI. Dennett argues that adopting an intentional stance and treating AI systems as agents can help us predict and understand them, though it also makes us vulnerable to being fooled. While Dennett acknowledges that agentiveness is a continuum, he still ...
The video discusses the concept of hierarchy and differentiates between hierarchy in general and hierarchical power structures. Hierarchy refers to any list of order or importance while hierarchical power structures concentrate power at the top and enforce it through domination. Hierarchical power structures are rare in nature and tend to exist in less intelligent species. In contrast, horizontal power structures distribute power through libertarianism and mutuality. While hierarchical power structures are universally bad, some hierarchies can be acceptabl...
The video discusses terms like "Himbo", "Soy Boy", and "Soft Boy" used to describe men who exhibit softer or more feminine traits. While some view these terms as insulting, others find Himbos pleasant and uncomplicated. However, embracing femininity can be risky for men in a patriarchal society where masculinity is tied to humanity and desirability. The video argues that men embracing a soft life can help escape narratives of Black male hypermasculinity and violence. Allowing men to express a wider range of emotions and traits could help them achieve self-...
- Many companies and platforms are becoming more restrictive and hostile towards developers, limiting what can be built on their sites. This reduces creativity and usefulness of the internet.
- Major platforms are deleting old content and inactive accounts in mass, resulting in a loss of internet history and institutional memory.
- Search engines are becoming less useful and filled with ads, clickbait content, and generic results that don't answer users' questions.
- Search engine optimization practices have homogenized the internet and sterilized content, focusi
... (read more)