AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Around 600 million years ago, the first biotariants emerged with the ability to navigate and pursue food using a simple algorithm called taxes navigation. This algorithm allowed the organisms to turn away from dangers and move towards food by detecting smell gradients. This breakthrough introduced the concept of valence or reward, where certain stimuli were classified as good or bad, directly mapping to a movement repertoire. From an evolutionary perspective, the brains of these organisms needed to consolidate into a nucleus to make trade-offs and integrate different sensory inputs. This breakthrough marked the beginning of the exploration of the world based on valence and the development of the ability to navigate and find food.
Around 50 million years later, during the Cambrian period, fish and early vertebrates evolved with a more advanced capacity for learning through trial and error. This breakthrough allowed these organisms to learn and adapt behavior based on reinforcing actions that led to positive results. The key insight came from the concept of temporal difference learning, pioneered by Richard Sutton, which suggested that the reward signal for learning should be based on changes in predicted future reward rather than the actual reward itself. In vertebrate brains, the basal ganglia and cortex played a vital role in enabling reinforcement learning. The basal ganglia, with its dopamine and prediction error signals, was responsible for reinforcing behaviors based on expected future rewards. This breakthrough allowed for more complex learning and the recognition of patterns and 3D spatial mapping, enhancing the organism's ability to adapt and navigate its environment.
About 500 million years ago, the development of a structure called the hippocampus in vertebrates marked another breakthrough in brain evolution. The hippocampus enabled the simulation of mental maps and the ability to mentally represent and simulate future scenarios. This breakthrough allowed organisms to plan ahead, remember locations, and navigate through complex environments. It also laid the foundation for the emergence of episodic memory, incorporating the ability to form memories of specific events and experiences.
Around 350 million years ago, the emergence of the social brain in vertebrates brought about the ability to mentalize, or attribute mental states to oneself and others. This breakthrough enabled an understanding of intentions, desires, and beliefs, facilitating complex social interactions and cooperation. The development of structures like the prefrontal cortex and mirror neurons played a crucial role in mentalizing.
Finally, around 100,000 years ago, with the evolution of early Homo sapiens, the ability to produce complex vocalizations and language represented a significant breakthrough. Language formed the foundation for enhanced communication, cultural transmission, storytelling, and the expression of abstract concepts, further expanding the cognitive capabilities of humans.
The neocortex, which evolved in early mammals, enables the simulation of possible futures. It is the source of metacognition, allowing individuals to think about their own thinking. By generating models of the self and simulating different scenarios, mammals can plan and make decisions based on anticipated outcomes. This ability is observed in rats, who display vicarious trial and error while navigating mazes, playing out different paths before making a decision. Primates, with their developed neocortex, exhibit theory of mind, imitation learning, and anticipating future needs. This suggests that the neocortex's core adaptive value lies in its capacity for simulating one's own inner simulation.
Theory of mind, the ability to infer the mental states of others, emerges with the primate brain's granular prefrontal cortex. Primates, including non-human primates, demonstrate theory of mind by predicting others' intentions and understanding their knowledge. This facilitates imitation learning, where primates learn from observing others' actions and inferring the intent behind those actions. The imitation learning process is enhanced by the primate's simulation of thinking about different things, enabling them to engage in counterfactual learning and anticipate future needs.
While biological brains have a complex and non-linear architecture, current AI systems have a more linear and differentiable structure due to the constraints of backpropagation. However, the differences in architecture between biological and AI systems may lead to challenges in understanding and interacting with AI systems. Humans' ability to relate and understand each other relies on shared reward structures and similar brain mechanisms, which may not be present in AI systems. Consequently, developing effective communication and understanding mechanisms with AI systems that have different architectures might require novel approaches and considerations.
AI simulation and counterfactual ability are crucial aspects in the field of artificial intelligence. It is believed that the neocortex plays a significant role in enabling simulation, which is learned through trial and error. While AI has made notable progress in reinforcement learning, there is still room for improvement in terms of sample efficiency. Language models have shown potential in simulating and evaluating different paths, although this approach is considered somewhat of a hack compared to biological systems. The challenge lies in understanding how mammalian brains enter the counterfactual mode and generate inputs for hypotheticals.
Planning, or simulating possible futures, is a concept that has been explored for decades. The challenge lies in finding a practical approach to planning that works effectively. Research has focused on developing systems like the tree of thought, but the question remains whether these architectural tweaks are sufficient or if there are fundamental aspects of planning that still need to be uncovered. Mammal brains exhibit a unique ability to switch between different behavioral modes, pausing and thinking when necessary. Understanding how mammalian brains decide when to pause and think, how they prune the search space, and how they evaluate possible outcomes can provide valuable insights for AI systems. Continual learning and the integration of new information without catastrophic forgetting are additional key areas of research that can enhance AI's ability to simulate and model the world.
In this episode, Nathan sits down with Max Bennett, author of the book: A Brief History of Intelligence. They discuss the five major breakthroughs in evolution that resulted in human brains and cognition. They delve deeper into the workings of each breakthrough, the evolution of intelligence, and how AI can benefit from understanding these principles. If you need an ecommerce platform, check out our sponsor Shopify: https://shopify.com/cognitive for a $1/month trial period.
We're hiring across the board at Turpentine and for Erik's personal team on other projects he's incubating. He's hiring a Chief of Staff, EA, Head of Special Projects, Investment Associate, and more. For a list of JDs, check out: eriktorenberg.com.
LINKS:
A Brief History of Intelligence: https://a.co/d/6VPWTf7
SPONSORS:
The Brave search API can be used to assemble a data set to train your AI models and help with retrieval augmentation at the time of inference. All while remaining affordable with developer first pricing, integrating the Brave search API into your workflow translates to more ethical data sourcing and more human representative data sets. Try the Brave search API for free for up to 2000 queries per month at https://brave.com/api
Omneky is an omnichannel creative generation platform that lets you launch hundreds of thousands of ad iterations that actually work customized across all platforms, with a click of a button. Omneky combines generative AI and real-time advertising data. Mention "Cog Rev" for 10% off www.omneky.com
NetSuite has 25 years of providing financial software for all your business needs. More than 36,000 businesses have already upgraded to NetSuite by Oracle, gaining visibility and control over their financials, inventory, HR, eCommerce, and more. If you're looking for an ERP platform ✅ head to NetSuite: http://netsuite.com/cognitive and download your own customized KPI checklist.
X/SOCIAL:
@labenz (Nathan)
@maxsbennett (Max)
@CogRev_Podcast
TIMESTAMPS:
(00:00) Intro
(06:00) Max's background on commercializing AI in technology
(09:53) AI has taught us a lot of our intuitions are wrong
(11:22) Why we're better off avoiding defining intelligence
(14:38) The Five Breakthrough Model - steering, reinforcing, simulating, mentalizing, and speaking
(15:43) Sponsor - Brave Search API | Shopify
(18:20) The story of the first breakthrough - Steering
(21:40) Evolution doesn't have forethought and every generation has to survive
(23:05) The study of nematodes - how did this first brain emerge?
(27:10) yet in bilaterians, but the idea is you classify things in the world into good and bad.
(33:00) Netsuite by Oracle | Omneky
(36:16) The story of the second breakthrough - Reinforcing
(1:32:44) The story of the fourth breakthrough - Mentalizing
(1:45:57) The story of the fourth breakthrough - Speaking (pls listen to the audio book / read the book)
(1:46:46) How evolutionary paradigms will shape up in the future
(1:55:03) A unified theory of value or moral patienthood that can be applied to the five breakthroughs
(1:59:21) Wrap + Sponsor | Omneky
This show is produced by Turpentine: a network of podcasts, newsletters, and more, covering technology, business, and culture — all from the perspective of industry insiders and experts. We’re launching new shows every week, and we’re looking for industry-leading sponsors — if you think that might be you and your company, email us at erik@turpentine.co.
Listen to all your favourite podcasts with AI-powered features
Listen to the best highlights from the podcasts you love and dive into the full episode
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
Listen to all your favourite podcasts with AI-powered features
Listen to the best highlights from the podcasts you love and dive into the full episode