AGIHound

"This may surprise but, unlike deep learning, the brain doesn't learn to recognize any specific pattern or scene. It learns to perceive anything on the fly."

[AGIHound @TrueAIHound]

LLMs mimic language, not intelligence. When was the last time an LLM walked into a random kitchen and boiled an egg? When was the last time an LLM learned to walk on its own, using its own sensor s and effectors?

Moreover, the only reason that LLMs can mimic language is that they are cheating by stealing the work of millions of human beings who did the hard intelligent work.

LLMs are not based on any new understanding of intelligence. They are based on old linguistic science that predates the AI field. Linguists have known for a long time that language is highly statistical, i.e., contextual. LLMs calculate the stats and store them in tokens. This is not intelligence. 🤔

Deep down, every LLM is dumb as a rock. 😀

[Pedro Domingos @pmddomingos]

What is our intelligence, if LLMs can mimic it so easily?

[AGIHound @TrueAIHound]

@PaulPallaghy @skdh By your definition, a thermostat is intelligent. Are you serious?

Nature tells us what intelligence is. It is the ability of a system to learn continually, to generalize, to set and achieve goals and to quickly adapt to novelty in the real world.

[AGIHound @TrueAIHound]

"The real AI wall is the rapidly increasing energy need." ~ @skdh

I beg to differ. The real AI wall is that no one in the AI community, especially the fake-AI mafia, has a clue how intelligence works. It is for this reason that they have an energy problem. 🫤

[Sabine Hossenfelder @skdh]

Most people do not understand that the real AI wall is the rapidly increasing energy need

[Srini Pagidyala @Srini_Pa]

It is exactly this kind of nonsense that got us here.

Show me AI that can learn, adapt, self-correct, and compound continuously and autonomously with metacognition.

Context is downstream.
(Cognitive) Architecture sets destiny.

Economics expose the architecture.

[AGIHound @TrueAIHound]

It's the same problem if you are trying to solve intelligence. Why? Because if you solved intelligence, you will have solved both full self-driving and general robotics. The number of degrees of freedom is not an issue. The machine will learn to use them.

[Pedro Domingos @pmddomingos]

A self-driving car is a robot, but it controls a single velocity vector. A housebot needs to control dozens. It’s a problem of a different order of magnitude.

[AGIHound @TrueAIHound]

Elon Musk believes in vibe consciousness. Essentially, if Grok or Optimus behave like they are conscious, then they are conscious. 🤦‍♂️

Western civilization used to have great philosophers. This is depressing.

[Dustin @r0ck3t23]

Elon Musk just dissolved 3,000 years of philosophy in four words on Lex Fridman’s podcast.

“Might as well be human.”

And it has nothing to do with machines.

Musk: “It will soon be able to simulate what we consider consciousness. To a degree that you would not be able to tell the difference.”

Think about what that actually means. Not for AI. For you.

You have never once confirmed that another human being is conscious.

Not your mother. Not your partner. Not your closest friend.

You watched their behavior. You heard the right words at the right times. You saw expressions that matched the moment.

And you called it real.

That is a Turing test.

You have been running one on every person you’ve ever known since the day you were born.

And every single time, you passed them on faith.

Fridman: “From the aspect of the scientific method, it’s might as well be consciousness, if we can simulate it perfectly.”

Fridman is not making a claim about AI.

He is naming something humanity has never confronted.

Consciousness has never been proven between two human beings.

We never verified it in each other.

We performed it for each other.

And then we trusted the performance.

For millennia, we told ourselves our flaws were the proof. That our hesitations and contradictions were the signature of something no machine could touch.

Musk: “Talks like a human, makes mistakes like a human… and you literally just can’t tell.”

If a machine can perfectly simulate your imperfections, your imperfections were never sacred.

They were patterns.

The question was never whether AI will become conscious.

The question is whether consciousness was ever anything more than the performance itself.

We assumed something existed behind the behavior. That being human meant something deeper than the act of being human.

Musk didn’t build a machine that passes the test.

He revealed the test was all there ever was.

Musk: “Might as well be human.”

Four words that don’t elevate the machine.

Four words that reveal “human” was never a proven category.

Just a performance we agreed to believe.

[AGIHound @TrueAIHound]

@Agrail A glue is a property. It's like having a hand that can grab another particle's hand. Its strength is proportional to the particle's mass. The "strong force" is a glue. The "weak force" is also a glue.

[AGIHound @TrueAIHound]

Continuity is worse than a monstrosity. It's a stupid concept because it requires infinity. Infinity, too, is stupid. Continuity is a giant monkey wrench in physics research. It makes everyone stupid. 😀

But watch out. Without continuity, most of modern physics, including Einsteinian physics, is just a "castle in the air", as Einstein himself called it. The ancient Greeks knew that infinity was bogus. Democritus used this understanding to predict the existence of atoms more than 2000 years ago.

The only reason that infinity and continuity are still used in modern physics is politics. It is the main reason that, 100 years after Einstein published his theory of general relativity, humanity still has no clue what causes gravity. 😠

[Pedro Domingos @pmddomingos]

Continuity is a monstrosity.

[AGIHound @TrueAIHound]

AI "world models" vs. perception in the brain

"There is a lot confusion as to what a "world model" should be or represent." ~ @filippie509

Yes there is. In my opinion, there is no need to represent or compute a 3D "world model" in memory. The world is its own model and "computes" itself perfectly. Early in childhood, the human brain learns to use its sensors in order to perceive (capture) the results on the fly. The precise timing of discrete sensory events is essential to this learning process.

Neurons are too slow to compute a 3D world model. Representations of patterns or objects are constructed instantly from sensory signals and almost all of them are discarded soon after use.

I estimate that we remember less than 1% of what we experience. Remembering requires effort, i.e., multiple recalls. Only a few unique traces of the percepts may be kept for future reference. The traces can be used for recognition, planning or reasoning.

Deep learning systems cannot perceive anything unless they have previously been trained to recognize it. By contrast, the brain can perceive anything even if it has not seen it before.

Deep learning is fine for computer automation, but it's a monkey wrench in the works if solving intelligence is the goal. 🤔

I love the brain. 😍

[Filip Piekniewski🌻 🐘:@[email protected] @filippie509]

There is a lot of confusion as to what a "world model" should be or represent. A result from a world model is robust behavior. It doesn't need to represent 3d structure if that doesn't facilitate robust behavior. It depends on the "embodiment" and the degrees of freedom.

[AGIHound @TrueAIHound]

@AbhiRaama22 @filippie509 @worldlabs Yeah. It's another scam imo. Wickedness lives in Silicon Valley. 😀

[AGIHound @TrueAIHound]

Electrons are not fundamental particles. No particle with a mass greater than the fundamental mass is fundamental. It takes force to hold masses together.

Electrons have special "glue" properties that hold them together. These properties allow them to hold on to each other without causing annihilation.

I've determined that each electron or positron consists of 4 subparticles. Each subparticles also have "glue" properties. Electrons and positrons can combine to form protons and neutrons. I believe that understanding the intrinsic properties of electrons and positrons, one can predict the properties of the nuclear particles and all the elements.

One major consequence of this model is that the unacceptable problem of the missing antimatter introduced by quark theory, no longer exists. Quark theory is worse than pseudoscience. It's crackpottery imo.

I don't claim to have all the answers. Work in progress.

[Jack 🤖 @JacklouisP]

Humour my recent rabbit hole into the colour blue. I discovered it’s universally hard to produce, and the physics is very cool:

•Real blue is rare AF. Fewer than 10% of plants are blue, almost no animal makes a blue pigment, and the ones that look blue are mostly faking it. Humans have had the exact same challenge.

•The challenge. Pigments colour things by absorbing some wavelengths and reflecting the rest. Blue is the hardest because it means absorbing red, the weakest light energetically. Soaking up low-energy light takes a big, complex molecule, and the bigger the molecule the more expensive it is to build. So nature leans on cheaper colours like red and brown.

•Nature also fakes it. Most blue you’ve seen in the wild isn’t pigment, it’s structure. Butterfly wings, peacocks, blue eyes: no blue pigment in any of them, just nanoscale architecture that scatters blue light back at you. Damage the structure and the blue dies.

•Even blueberries aren’t blue. Squish one and the juice runs deep red. The blue is a coat of microscopic wax crystals that scatter blue and UV light, sitting over a dark red skin. We only figured this out in 2024.

•Humans also struggled. Real blue was so hard to make that ultramarine, ground from a single Afghan stone, cost more than gold.

•The fix came by accident: in 1700s Berlin a contaminated lab batch produced Prussian blue, the first cheap, synthetic blue, and the dam finally broke. Its colour isn’t a fragile molecule at all but iron sitting in two charge states in one crystal; light knocks an electron between them and that jump swallows red, leaving a deep, stable blue.

How did we beat it? Not with a better organic molecule, nature already tried those and they’re all expensive. We stepped outside biology.

Nature is a great model, but it has its limits.

[AGIHound @TrueAIHound]

@JacklouisP Thanks for posting this. I love it. The sky and the seas are exceptional, no?

[AGIHound @TrueAIHound]

My model is just that, a model. It has no importance to the world unless it leads to a demonstrable breakthrough.

Quarks are total nonsense because the theory disappeared most of the antimatter of the universe. Instead of admitting that their theory is BS, quark theorists had the gall to announce to the world that they discovered an anomaly of nature.

The fake-physics mafia never disappoints. 🤦‍♂️

[AGIHound @TrueAIHound]

@Agrail This is where we part company. I believe in particles with multiple properties.

[AGIHound @TrueAIHound]

The 'point' is a concept that started with Euclid, I believe. It has no size. Therefore, the space between any two points separated by any distance has an infinite number of points.

Some ancient Greek philosophers (Zeno, Parmenides, Democritus and others) noticed the problem with infinity and continuity millennia ago. Democritus rejected infinity and used his understanding to predict the atom, the smallest possible size that can't be divided further.

Modern physicists and mathematicians think they're wiser than the ancient Greeks. They are self-deluded fools imo. Infinity is a debilitating poison in the body of physics imo.

[AGIHound @TrueAIHound]

Hawking and his legacy will be forgotten. Continuity physics, which he practiced, is full blown crackpottery, in the not-even-wrong category. 😀

Even Einstein came to realize this late in his life. In 1954, one year before he died, he wrote to a friend: "In that case, nothing remains of my entire castle in the air, gravitation theory included, [and of] the rest of modern physics." ~ from 'Subtle is the Lord' by Abraham Pais

Big changes are coming to physics. Sacred cows will be slaughtered. 😬

[Physics In History @PhysInHistory]

Stephen Hawking speaks against creationism ✍️

Because there is a law such as gravity, the universe can and will create itself from nothing. Spontaneous creation is the reason there is something rather than nothing, why the universe exists, why we exist. It is not necessary to invoke God to light the blue touch paper and set the universe going.

[AGIHound @TrueAIHound]

@Agrail I don't believe in quarks. I know that the electron and the positron are composite particles. Each has 4 sub-particles. Some physicists call them quarter electrons or quasi-praticles.

[AGIHound @TrueAIHound]

In my opinion, the published CODATA values of the proton/electron and neutron/electron ratios are slightly off because physicists use Einstein's E = mc² to calculate the masses. In my model, E = mc² is only valid for the strong force. Neutron decay is caused by the weak force. So I don't believe that neutrinos exist.

In my model, the proton consists of 918 electrons and 919 positrons, for a total of 1837 and a net positive charge of 1. The neutron, on the other hand, has 919 electrons and 919 positrons, for a total of 1838 sub-particles and a zero net charge.

Quark theory is complete nonsense. Protons and neutrons are made of electrons and positrons. There is no missing antimatter. I can calculate the precise half-life of the neutron from the mass of the electron alone. I know why the proton is stable and why the neutron is not. But I'm not yet ready to release all my secrets.

Work in progress.

[AGIHound @TrueAIHound]

@Agrail The Planck constant is polluted by geometric constants and Newton's G. I don't need it in my model. I don't need G either.

[AGIHound @TrueAIHound]

In my particle model, the mass of the electron is equal to exactly 1838 fundamental masses. I had to make k equal to 1 and the only way to do that was to create a new unit for mass.

I can calculate the precise force in Newtons between 2 electrons using the mass of the electron and my '1838' hypothesis. No Coulomb constant is required. This is based on a discrete physics model. This formula will work using any unit for mass.

[AGIHound @TrueAIHound]

I have doubts about the correctness of Planck constant h. It's not a pure physics constant imo.

Not that I disagree with your hypothesis but, for accurate discrete physics, one needs accurate values for the fundamental length and the fundamental time interval.

In my discrete model, all the fundamental constants can be surprisingly derived from the speed of light alone. The kilogram unit of mass used in the SI system is not coherent with the speed of light. I had to create a new mass unit called the cgram.