#moravec's paradox

[diracsea]

TIL: Moravec's Paradox

Moravec's paradox is a phenomenon observed by robotics researcher Hans Moravec, in which tasks that are easy for humans to perform (eg, motor or social skills) are difficult for machines to replicate, whereas tasks that are difficult for humans (eg, performing mathematical calculations or large-scale data analysis) are relatively easy for machines to accomplish.

source

[metamatar]

Moravec's paradox is the observation by artificial intelligence and robotics researchers that, contrary to traditional assumptions, reasoning requires very little computation, but sensorimotor and perception skills require enormous computational resources. The principle was articulated by Hans Moravec, Rodney Brooks, Marvin Minsky and others in the 1980s. Moravec wrote in 1988, "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.

[youzicha]

Moravec's paradox says that it’s easy to write computer programs to do the things humans find hard, like chess and symbolic algebra, but difficult to do the things humans find easy, like visual perception or walking.

Isn’t there a mini-Moravec in the recent success of deep learning? The jobs that seem on the line to be automated first are illustration and translation, which both are quite difficult for humans, you have to study several years to get to a professional level. Meanwhile, easy jobs like programmer (go to a three-month boot camp and you’ll be economically productive) or “office temp” (no training required) remain unaffected for now. 

[videoworm]

oh i just remembered something cool i learned recently:

its called Moravec's paradox

Contrary to popular belief, high cognitive functions like reasoning take wayyy less resources than basic function like movement and sensing. The latter are the hard things for a brain to do. And for any artificial brain.

Not surprising that the frontal cortex is only a small fraction of the brain. Reasoning is so simplistic that we can easily simulate it with computers.

(there are AI-models that generate mathematical proofs, and noone beats a computer in chess (unless the program is modified to give the player a chance lol) (the chess-thing is 'good old fashioned AI' aka no artificial neural networks. which used to be the hot shit in the 80s but nowadays its starting to drift away from whats even considered 'Intelligence'. The same process happens with Artificial Life (newer research field). We don't really know what Intelligence and Life is, so the definitions are gonna change with further research)

god i cannot write a lineal text WHERE was i

the animal kingdom naturally evolved bottom-up: simple reactions to stimuli like touch or no touch evolved into clusters of neurons and then into brains [cool paper about anthropod brain evolution], these brains kept evolving and now we have some mammals with prefrontal cortexes. and can consciously do stuff like... counting to 100,000. Generally speaking, what we do with Artificial Intelligence reseach is exactly the opposit (top-down).

Back in the days scientist thought that Consciousness resides in this 'high intelligence'. But brains do not work in a vacuum. We need stimuli from the outside and we dont get or process those stimuli with our forntal cortex. Ones cognition is highly influenced by our whole body (something something link this to neurodivergence) [Embodied Cognition].

Artificial Intelligence is some Process we invented which is unique. It's like body-less math. It exist in an almost vacuum and any input it gets are one-dimensional bit-streams. (alright, that last sentence is more peotry than science)

Aritficial Life does more of a bottom-up approach (thats why i like it :^) ) (also it does not get blasted through and misrepresented in pop media like AI currently is). But keep in mind we couldnt even realistically simulate a single neuron with all its molecules and chemistry and atoms. real life is just too complex

[carvalhais]

AI is harder than we think, because we are largely unconscious of the complexity of our own thought processes. Hans Moravec explains his paradox this way: “Encoded in the large, highly evolved sensory and motor portions of the human brain is a billion years of experience about the nature of the world and how to survive in it. The deliberate process we call reasoning is, I believe, the thinnest veneer of human thought, effective only because it is supported by this much older and much more powerful, though usually unconscious, sensorimotor knowledge. We are all prodigious Olympians in perceptual and motor areas, so good that we make the difficult look easy”. Or more succinctly, Marvin Minsky notes, “In general, we’re least aware of what our minds do best”. Melanie Mitchell, 2021. Why AI is Harder Than We Think. arXiv:2104.12871.

[cyanocoraxx]

wrote this at 1.11 am because i couldn’t sleep. just a little drabble about neo and their siblings because i felt existential and was reminiscing about how writing the og fanfic made me feel

& ao3 link

-

moravec's paradox

generally speaking, it is comparatively easy to make computers exhibit adult-level intelligence on intelligence tests or playing chess, and difficult to give them the skills of a human infant when it comes to perception and mobility. this observation is known as moravec's paradox.  

the obvious implications of this are that, although our technologies can be skilled at performing specific tasks beyond our own capabilities like flying at great heights and speed, for instance, or processing massive quantities of data - they're quite bad at doing things that most organic creatures do without thinking, like walking and picking up objects and opening doors, things that are really extraordinarily complex and demanding. an impossible combination of signals from the brain to the body, too difficult to replicate perfectly in another body. but perhaps this is self-centered thinking. perhaps it is possible, just not in the way we believe.

neo was painfully, but almost comically aware of this paradox as they simply walked down the hallway of their home. they were not made for walking - their feet were heavy and clunky, their movements rigid and although mechanically even, almost off-balance at the same time. their joints and fans whirred along with their precise movements.

one thing that organic beings have over machines like himself was evolution. all organic skills have a biological basis, machinery designed through natural selection. natural selection preserved design optimizations and improvements - and the older a skill is, the more time natural selection has to improve it. compared to a skill like walking, abstract thought developed a lot more recently. consequently, its implementation should not have been particularly efficient.

take a chess-playing AI, for example. because a chess-playing AI would be driven entirely by the maximization of its utility function - that being, to win at a game of chess - any scenario in which it might be shut down is one that it would be motivated to avoid, given that being turned off would drastically reduce its specific function. when a chess-playing robot is destroyed, it never plays chess again. there is no second round. such outcomes have a very, very low utility - and this machine would do whatever it could to avoid it, to avoid losing. so, you could build this chess-playing robot, thinking you can just turn it off if something goes awry, but you find that it strenuously resists your attempts to shut it down. this basis for survival is not akin to abstract thought, but rather, an almost instinctive desire to survive.

neo ruminated on this simple thought as he turned the corner to meet his siblings in the living room. he folded his arms, a very human gesture, as he leaned with one shoulder against the doorframe. silver sonic mk ii and mecha sonic, the quintessence of machines built for one task, enjoying a video game together. he watched them with a swell of pride and his optics glowed just a fraction brighter.

“you modded, mecha!”

“i did not. we are engaged in the exact same game as we speak.”

“you’re a stinky liar, mecha. you clipped through that wall. you’re a fake gamer and you know it.”

“that was merely a glitch in the game’s coding, brother. are you certain you are not glitching? must i run diagnostics on your processor again?”

“what?”

mecha stared at him without speaking.

“ever.” silver finished, poking his tongue out at his sibling to rile him.

“i see. you are not glitching. your personality is simply defective.”

“rude. i’m a beacon of likable personality.”

“likeable is subjective.”

“your mom is subjective.”

“... what.”

neo was no stranger to emotion. no, emotion was very much fuel for his motivations. and he loved his siblings. he did. love, something once only thought to be reserved for humans and some other specific organic creatures - he had come to embrace it. he relaxed, gently, into the doorframe as he watched his brothers play a video game - a paradox in itself - a machine working a machine - tuned his auditory receptors as his youngest brother laughed loudly amongst their chatter. he was clearly trying to rile his stoic sibling to the best of his ability - and chaos, he was good at it.

what a piece of work was life, neo thought; what a paragon of dust.

silver sonic mk ii was sent tumbling gently into a row of cans by the wall, and he screamed in delighted outrage, before climbing up again. although created for nothing more than murder, there was a distinct gentleness - a deep brotherly bond - between the two machines. neo cast a look down to his sharp, steel claws - glinting in the light of the television - and pondered on that. looking at his design, some would feel an instinctive terror of predation, perhaps compounded with the knowledge that robots like him had been created with the intent to kill by one of the world's most powerful human beings.

and yet, there was softness. there was beauty in their power, in what they chose to be. chess-playing robots playing the game of life. creatures surviving in their environment through social bonds. dagger digits learned to close around objects gingerly to avoid causing undue damage to them. learned to move with precision to re-wire broken connections in his siblings’ parts when they needed it. yes, the soft animal bodies of organics had the benefit of evolution through millions of years of life - but artificial intelligence like them had the benefit of optimization through love. a love for what they did, for the bonds they created. an instinctive desire to survive through social bonds, through a love not created by organic oxytocin - rather, created through the perception of what the world was to them. what they deemed life to be.

none of this, neo felt, could be rendered in code. none of this, he thought, could be run on any other substrate. their beauty was bodily, in the most profound sense, in the most wonderful sense. he never loved his brothers more, he realized, when he thought of them as animals. he dragged himself, his animal body from the doorway to join them.

[rthidden]

Why AI Struggles with Simple Tasks

Explore why AI excels at complex tasks but struggles with basic human skills.

1. Moravec's Paradox

AI can master chess but can't wash dishes due to Moravec's Paradox.

In 1988, Hans Moravec noted that while AI handles high-level reasoning well, it struggles with sensory motor skills. These skills are deeply ingrained in humans through millions of years of evolution.

2. Sensory Motor Skills vs. High-Level Reasoning

Human brains excel at tasks like perception and mobility due to evolutionary reinforcement.

High-level reasoning skills, such as calculus or chess, are newer and easier to teach machines since they are well-understood and documented by humans.

3. Unconscious Human Skills

Oldest human skills operate unconsciously, making them hard to decode and teach to AI.

Steven Pinker highlighted this in 1994, stating that easy problems for humans are hard for AI, while hard problems for humans are easy for AI.

4. Examples of Moravec's Paradox

Tasks like language translation and playing chess are easy for AI but hard for humans.

Conversely, tasks like walking or recognizing objects are difficult for AI but second nature to humans.

Understanding these distinctions helps set realistic expectations for AI capabilities. Start considering how you can leverage AI's strengths in your business today!

[appventurez]

Explore the divergence between robotics and AI through Moravec's Paradox, which posits that tasks easy for humans are challenging for machines, and vice versa. Delve into the complexities of this paradox as we navigate the gap between robotics and AI, deciphering why certain capabilities lag despite advancements in artificial intelligence.

[alshlyapin]

Moravec's paradox

I recently came across this: https://en.m.wikipedia.org/wiki/Moravec%27s_paradox. It claims that, contrary to popular belief, intellectual labor will be replaced by AI first, followed by manual labor. And their logic really makes sense. For example, ChatGPT can already code at a junior level and write texts at a professional level, whereas autonomous vehicles are not yet fully developed (although Waymo has already launched a taxi service in San Francisco). It turns out that a programmer's job is easier for AI than a taxi driver's job. This led me to think. I always believed that a programmer's job (and other intellectual work) would be replaced after non-intellectual work, but it turns out to be the opposite. I've been thinking about this lately but haven't yet figured out how to change my life considering this new information. Should I go into coal mining? Become a cleaner? Work in a factory? Become a waiter? It doesn't seem like a logical solution. In the end, everything remains as it is, and I stay as an LLM engineer.

[gerdfeed]

Indeed, Moravec’s paradox, a principle conceived in the 1980s, states that cognitive skills that are harder for humans, like math and logic, are easier for a computer to handle than things that are typically easier for humans, like motor and sensory skills.

Analysis | What Hollywood gets right — and wrong — about real-life AI

[plugchain]

Solution to PlugChain’s Breakthrough: AI+Oracle=Metaverse

Introduction: At the end of March, Musk and scientists from the Future of Life Institute signed an open letter calling for a halt to the development of advanced AI systems. Musk has long warned that people should be cautious when facing artificial intelligence, and he is joined by physicist giant Hawking. Hawking predicted in 2014 that “artificial intelligence could end humanity.”

With the application of large-scale language models, one can not help but worry whether sci-fi stories like “Black Mirror” are just the creators’ groundless fears or nightmares that will come true.

Moravec’s paradox, a famous paradox in the field of artificial intelligence, refers to the fact that AI can produce excellent results in areas where humans are not good at, such as complex mathematical operations, but can not easily perform tasks that humans can easily accomplish, such as going out to buy groceries or tidying up a room.

In recent years, “weak AI” that focuses on solving single tasks has made some progress, including playing chess, mastering knowledge, and identifying medical conditions. This year, generative AI has exploded, although it has not yet reached the level of the “strong AI” or artificial general intelligence (AGI) that is an all-rounder, it has already made people very excited.

Similarly, the hot concept of AI has also spread to blockchain technology. Or, to be more precise, at the beginning, AI set its sights on the blockchain field!

• Oracle+AI: Unleashing the “Value” of Data

Currently, with the continuous development of blockchain technology, the oracle has become an indispensable part of the blockchain ecosystem. The oracle is a technology that can input data from the real world into blockchain. It can ensure that blockchain applications obtain accurate and real-time data, thereby expanding the application scenarios of blockchain.

However, the shortcomings of the oracle itself also limit its development in the blockchain ecosystem. At this time, the emergence of artificial intelligence (AI) technology provides new development opportunities for oracles. Obviously, the combination of oracles and AI will trigger a new technological revolution.

Firstly, AI can provide more efficient and accurate data analysis for oracles. AI technology can classify, filter, and analyze a large amount of data and process and model the data to better provide data for oracles. AI can also automate the process of data collection and processing, reducing the cost and time of manual intervention. In this way, oracles can obtain more accurate and real-time data from a wider range of data sources and provide data to blockchain applications at a faster speed.

Secondly, the combination of oracles and AI can expand the application scenarios of blockchain applications. Since oracles can obtain data from the real world, they can be used in many fields, such as supply chain management, finance, and healthcare. By using AI technology, oracles can better process and model the data in these fields, providing higher-quality data for applications and expanding the application scenarios.

Finally, the combination of oracles and AI can also provide a more efficient and secure data verification mechanism. Data verification in blockchain applications is an important task that usually requires multiple nodes to verify the data. Using AI technology, oracles can analyze and model the data and judge the verification results to improve the efficiency of data verification. In addition, AI can also provide a more secure data verification mechanism by identifying potential data tampering and fraudulent behavior through machine learning and model prediction technology.

Overall, oracles provide a channel for data to move from Web 2.0 to Web 3.0, and the technology of AI makes the application scenarios of data more extensive. Therefore, oracles and AI will play an increasingly important role in the development of the blockchain ecosystem.

• Oracle: The “Cradle” of Blockchain Data

The concept of AI may be familiar to everyone, and there are also many mature products on the market, such as the previously famous AlphaGo and the currently popular ChatGPT. However, people’s understanding of language machines seems to be very limited, and some may ask what exactly is an oracle?

As its name suggests, an oracle is not a prediction tool, but a “bridge” that maintains data and information communication between blockchain networks, the Internet, and other blockchain networks. Especially in decentralized applications (DApps) such as DeFi smart contracts, developers can use oracles to call various external data resources, including market prices, to connect DApps to the external real-world data environment.

The reliable support for the price feeding function or data source that we encounter in our daily lives is only in its early stages. In the long run, the oracle will serve as a collection of various elements such as real-world data, information, credit, and assets. It provides the correctness of data or information, the decentralization level of technical implementation, and the intelligence level of module scripts. It can also be combined with multiple blockchain fields, acting as a material that can be used to build credit with DID, combined with stablecoins to ensure the value match between stablecoins and the US dollar at the current exchange rate, and combined with insurance to ensure the accuracy of claims information, etc.

Therefore, the oracle is almost the data cradle of all DeFi protocols and an indispensable part of the blockchain field.

So how does an oracle work?

First, an oracle generally interacts with the execution engine as an independent module of the blockchain or a third-party service. The oracle is only responsible for obtaining trustworthy data and does not directly participate in the execution of transactions. Users initiate oracle service requests through contract calls (or other methods such as special API interface services), informing the blockchain execution engine that they want to execute a transaction that includes an oracle service by calling a certain built-in contract interface.

Second, during the execution of the execution engine, when a service request to the oracle is detected, it is forwarded to the oracle module through an internal communication component. This request will contain some information about requesting external data sources, such as a web data request, which will include common information such as URL and HTTP headers.

Third, after receiving the service request, the oracle initiates a data acquisition request to the external data source, obtains the data, uses a transaction generator to generate a new internal callback transaction, and signs it (this process will use hardware technologies such as TEE to ensure security and immutability).

Finally, the oracle sends this callback transaction to the execution engine, which performs a series of operations such as organizing, managing, and storing the obtained data. This completes the entire process of executing a blockchain transaction that includes an oracle service.

• PlugChain Takes the Lead in the Race

With the rise of DeFi, the popularity of oracle has also been ignited. In this field, Chainlink, PlugChain, NEST, Augur and others have emerged successively, providing more diversified options for DeFi projects. Among them, Chainlink and PlugChain both uphold the decentralized principles of blockchain for their decentralized oracles. They usually use multi-signature or distributed algorithms to ensure data accuracy and consistency without introducing third-party institutions, but this approach is more difficult to implement and performance may become a bottleneck.

As a PlugChain , it has established a decentralized data network, where each oracle is a node in the network, and its architecture is divided into on-chain components and off-chain components. The on-chain components are responsible for interacting with users, collecting and responding to user requests, while the off-chain components are the aforementioned data network, used for processing data acquisition and transmission, greatly improving the efficiency of on-chain data processing.

In addition, the on-chain AI system is an important component of the PlugChain oracle network. By using AI algorithms and data analysis technology, PlugChain can automate the execution of smart contracts, as well as predict and analyze data and market trends.

These predictions and analyses can be converted into real-time market data and used to support various applications and services in the PlugChain ecosystem. For example, AI technology can be used to establish automated trading strategies and trading robots, providing efficient trading and investment support for users. At the same time, the more efficient and intelligent interaction of on-chain data makes it possible to combine AI with oracle to create a Metaverse.

Based on AI and oracle technology, once PlugChain constructs a decentralized Metaverse, the ecosystem will support various applications and services, covering GameFi, SocialFi, DeFi, Swap, DEX, DAO and many other application scenarios. Users can trade in Metaverse using “value data” and “cryptocurrency”, thereby realizing asset appreciation and data trading income.

Obviously, among the many oracle contenders, PlugChain, with the support of on-chain AI, has opened up a new path and is ready to catch up with the oracle leader, Chainlink!

In conclusion：it is not difficult to foresee that interacting with the outside world will be the next logical step for oracle. From a technical point of view, Oracle solves the problem of trusted data connectivity inside and outside the blockchain, but there is still the issue of “how to use” data. The emergence of AI is like a data service provider, endowing Web3’s data with liquidity and truly releasing its value, completely solving the problem of “where to use” data.

In the future, as the scale of blockchain in the financial, insurance, and IoT industries expands, the ecological value of AI+oracle is even more worth looking forward to!

[metamatar]

RUO Meme: Robotics

the thing is. ultimately. computers are coolest when they're used to do robotics. that's the point. sure, you can be a computer toucher. but getting the computer to touch you your world? that's where its at.

like its one thing to do abstractions on the computer and use the computer to talk to the other computer to give you real brainworms but when you use the computer to run a forklift that's when you feel that adrenaline of seeing a new world edging in the most visceral way. more controversially i think all intelligence is necessarily embodied – materially. that's the real answer to moravecs paradox.

[sexofword]

AI Paradox

"It is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility"

Hans Moravec

[a-clockwork-justice]

Hot chocolate is so good, just, goddamn, why does no one talk about hot chocolate more?

Aside from the fact that it tastes amazing and even better with whipped cream and marshmallows melting and especially around the holidays where they have a million weird and wonderful flavours, it’s such a great love language? It’s pretty much a scientific fact that food tastes better when someone’s made or even bought it for you and that applies so much to hot drinks too like ... you’re sick, hurt, stressed, tired, upset or any combination of the above, and someone close to you brings you a hot chocolate. Consider the possibilities:

You’re lying on the couch after a bad breakup/rejection/disappointment/life getting you down. Hot chocolate = “I’m here for you. You can tell me everything.”

You’re sick and achy and too tired to get out of bed or off the couch but not too nauseous to eat and drink in small amounts. Hot chocolate = “Hope this makes you feel better.” Even better if they sit next to you and let you rest your head on their shoulder as the hot chocolate warms your cold hands.

Even just coming home from a long day at work or school and you find a mug of hot chocolate being kept warm by a mug warmer. Hot chocolate = “You’ve earned it.”

You’re extremely emotionally vulnerable, for whatever reason, and you’re so weighed down by all your pain that you can’t do much more than be still or sob your heart out all over your confidante, who may or may not know the reason for your tears. They don’t pressure you to talk, but hot chocolate = “You need this, you deserve to not hurt this badly.”

Coming home from work is one thing, but hot chocolate in the workplace just hits differently. In an environment that doesn’t always allow for deep, emotional conversations and complaining is discouraged, someone brings you a hot chocolate right to your desk. Hot chocolate = “Keep on marching, soldier. I’ve got your back.”

Hot chocolate is a love language.

(Feel free to add more!)