Summary
The discussion revolves around comparing artificial intelligence, plant intelligence, and human "natural stupidity," emphasizing the limitations of AI in achieving "true intelligence" and the inherent brilliance of nature. Yann LeCun critiques current AI systems for lacking common sense and fundamental traits like reasoning and hierarchical planning. The conversation humorously contrasts AI's computational needs with plants' efficiency in sustaining life through photosynthesis, highlighting the intelligence embedded in nature and humanity's often destructive tendencies.
Key Words
- *Artificial Intelligence (AI)*: Limitations, hallucinations, reasoning, and common sense.
- *Plant Intelligence*: Photosynthesis, survival substrate, and natural efficiency.
- *True Intelligence*: Traits like reasoning, memory, and adaptability.
- *Natural Stupidity*: Human greed and environmental destruction.
- *Comparative Intelligence*: AI vs. plants vs. humans.
[20/11, 08:51]hu3: In the first paragraph of the story it says "will never achieve true intelligence". What is true intelligence? Is it what LeCun have said or made up by TOI? Lot of *hallucinations* in the story.
[20/11, 08:55]hu4: LeCun, Chief AI Scientist at Meta, argues that current AI systems are fundamentally limited because they are primarily trained on language and lack common sense. He points out that even a cat or dog has more common sense about the physical world than the most advanced AI today.
[20/11, 08:55]hu2: Are those journalistic or aistic hallucinations?
[20/11, 08:55]hu4: He identifies four key traits necessary for true intelligence:
*Understanding the physical world*: Learning "world models"—internal models of how the world works—simply by observing it, much like a baby does in its first few months of life.
*Persistent memory*: The ability to retain information and experiences between conversations and interactions, a feature largely absent in current LLMs.
*Reasoning:* The capacity to think through problems in a structured way, potentially over a long time, rather than just generating text one token at a time.
*Planning complex actions hierarchically:* The ability to set an objective and plan a sequence of actions and sub-goals to achieve that objective.
[20/11, 08:57]hu2: The term true intelligence is fraught with epistemic and hermeneutic uncertainty though 😅👇
[20/11, 09:03]hu4: And then there is *Natural Stupidity* 🥴
[20/11, 10:12]hu2: Exemplified by trees who exhibit the highest form of intelligence by their ability to capture photonic data and process it to provide energy for the entire planet!
And a large majority of these morons for which trees provide food will not think twice before cutting them down wholesale to make way for a concrete food mall!
What can be more naturally stupider than that? 😅
[20/11, 10:14]hu5: With the lack of full sensory and cognitive experience, such as, smell, taste, etc. the LLMs need to be treated as 'entities with special needs'. Then while imitating humans, fantasizing, we begin to describe them as 'hallucinating'.
[20/11, 10:22]hu2: Humans unlike LLMs have been blessed with the curse of dimensionality.
Certain humans tried to mitigate the curse long back by reducing our world to two dimensional academic flatlands that began with asynchronous communication mediated through cave paintings aka primordial AI.
Now with upcoming tech singularity humans are finally going to realise their dream relief from that age old curse?
[20/11, 11:09]hu3: One critical trait humans possess is "metacognition", which is missing in animals.
[20/11, 11:17]hu2: Missing in other animals?
Well human animals have proven beyond doubt that they have meta cognitive abilities to their own kind using linguistic modes but they may not have proven it to other animals?
Hence other animals don't know that we have metacognition similar to how we really don't know if other animals have meta cognition and it's currently just a human communication problem with other animals?
[20/11, 11:19]hu3: Don't disagree. 😊
[20/11, 20:31]hu6: Common Sense Problem was an issue addressed by Prof. Marvin Minsky. It still is a challenge for machine to learn. The current AI effort is *inductive* and is by design an approximation. Not designed for the whole truth.
The _*traditional AI*_ was designed and driven by *deductive reasoning* thus it pursued the solitary goal of *truth*.
With a different set of foundations and assumptions the expectations will vary. Accept it. The article is written by LLM apps and accept the dose of hallucinations.
Study REASONING as step one before you start your AI Journey.
[20/11, 20:41]hu3: Fundamental question - why you need an AI engine to have common sense when the human race in general does lack common sense?
[20/11, 20:44]hu3: I won't be surprised if we start debating why AI needs GW of data center power? Why cannot it have a few chapaties or a plate of rice and work for an entire day.
[20/11, 20:45]hu4: Perhaps to augment the cumulative benefits.
[20/11, 20:46]hu4: *Neuralink's* implantable N1 chip is designed for extremely low power consumption, with the System-on-a-Chip (SoC) consuming a total of 24.7 mW in a typical configuration. This efficiency is achieved through a combination of low-power analog processing and digital circuits, and the implant relies on a small, inductively charged battery for power.
[20/11, 22:52]hu2: Is neuralink currently the one with the least chapati consumption?
[20/11, 22:57]hu4: In most likelihood
Other conversations on animal intelligence to plant intelligence:
[12/11, 08:45]hu2: A short film on how to engineer education assessment toward climbing the learning
ladder:
The only explanation that this short film clip didn't get an Oscar yet is perhaps because it was released just 6 days back although it already has 3.7M views!
Hate to give away the climax as a hook but here goes: the final breakthrough in this student's education came when they found a teacher who thought so much out of the box that for him the box didn't exist!
What made the student reach up to the next level from a learner to a performer was an "out of the box" education assessment technology designed by this "out of the box" engineer teacher and the the team named that tech :
"education progress bar elevator!"
More here 👇
[12/11, 08:58]hu5: Fancy operant conditioning
Sorry! But just hype
A parrot can recite Tagore - does it compose too?
[12/11, 09:01]hu5: My version of the experiment would design it like something that's "natural" for the creature
[12/11, 09:03]hu5: The music is a reward context. Play the music and get food
[12/11, 09:11]hu2: Yes like if you were to teach a plant to dance slowly you may design a dark room with slits that can switch on and off the sunlight toward plant taxis!
[12/11, 09:11]hu5: That would be slow. Low performance value
[12/11, 09:12]hu2: Let's try to design something where an octopus can perform naturally
[12/11, 09:13]hu5: Think of this Synthetiforms keeping Bioforms as pets.
They would operant condition us to produce modem sounds... No matter how slow the bit rate
[12/11, 09:13]hu2: Plants have low performance value in terms of locomotion but when it comes to delivering even when standstill there are no current matches globally?
[12/11, 09:14]hu5: Art value. That makes sense
Hence the distinction - art vs performance
[12/11, 09:14]hu2: So plants have no other value than art value?
[12/11, 09:14]hu5: Our survival substrate
[12/11, 09:17]hu2: So plants are more intelligent than Tagore when it comes to providing food for survival rather than just food for thought?
Also plants are dumber than the dumbest human animal as they don't even whimper a protest while they are taken down to make way for human greed?
They are intelligent enough to have already planted their embedded intelligent systems before dumb human animals can totally finish them off?
[12/11, 09:18]hu5: Bioform.South Asian.Male
[12/11, 09:18]hu2: Bioform.South Asian.Male human animal
[12/11, 09:19]hu5: Bioform implies animal."so called" human
[12/11, 09:19] Rakesh Biswas: Why not plants?
[12/11, 09:20]hu5: Nope
That's just a physical phenomena... Producing food for the ecosystem
Second... They do "talk" ... Actually they talk a lot
[12/11, 09:21]hu5: I use the terms Bioform and Synthetiform in terms of TxHx (transhumanist) human like sentience
Thematic Analysis
- *AI's Limitations*: AI systems are critiqued for lacking the foundational traits of true intelligence, such as understanding the physical world and reasoning.
- *Plant Intelligence*: Plants are highlighted as efficient, self-sustaining systems that provide for the ecosystem, contrasting with human behaviors that often harm the environment.
- *Human Shortcomings*: The discussion humorously points out humanity's "natural stupidity" in exploiting nature despite its dependence on it.
- *Philosophical Exploration*: The conversation explores what intelligence truly means, using humor and analogies to question human-centric definitions.
- *Interdisciplinary Analogies*: Drawing parallels between AI, nature, and human systems underscores the need for balance, adaptability, and sustainable thinking.
No comments:
Post a Comment