Are Large Language Models Really AI?

How do you think the brain works when tasked? It looks for patterns in how you have solved similar or connected problems in the past. It recognises patterns of what succeeded and what did not. There is nothing special about the human brain that won't be able to be emulated by AI one day. The brain is just electrochemical reactions powering a complex neural network.

The current path AI is on, is not going to lead to real AI in future. The research and development path is based on fooling people into giving good responses.

There is nothing actually there, it's a chatbot. It doesn't do anything if you fall asleep while talking to it, it's just coded to respond. It's a really good chat bot, but it's still a chatbot.

Thinking investing a ton of money into muhAI will lead to real AI, is like thinking investing 100 billion dollars into a magic shop will lead to gandalf.

You need to be able to read to take an IQ test. Not sure what you don't understand here.

Tell me you've never taken an IQ test without telling me you've never taken an IQ test.

I'm pretty sure "Real AI" isn't even possible. True intelligence and sentience are beyond a computer's capabilities. The ability to reason comes from not just learned experiences, but personal emotions and feelings. No matter how complex you make its algorithm, that AI will never "feel" sad, or happy, or sorry for you, or angry. Concepts such as pain, pleasure, prosperity, desperation, sorrow, or elation are completely foreign to a machine. AI is only capable of interpreting and rendering abstract concepts when it has an existing reference to go off of, and it often does it very poorly.

AI is very good at networking a response or generating an image based off of many similar responses and images that are already available for it to reference. And even then the output AI gives you doesn't always make sense. AI generated stories have been popping up more frequently, and some of them are just bizarre, or painful to listen to, either repeating the same few things over and over or trying to build up to some important point only to never actually tell you what it is.

Don't get me wrong, AI is a fun tool to play with and I can see some practical uses for it. But to call it "intelligence" is really far fetched.

Fundamentally there are 2 types of knowledge:
A priori - those things that a definitionally true (all reasoning, mathematics, language itself)
a posteriori - those things which are observationally true (science, history, current events)

This of course ignores things like opinions and morality.

AI cannot (and I would likely argue never will) be able to process the a priori logic. Even simple regression models will produce negative numbers where none should be possible. Programmers must add stuff to the model to make the results palatable.(I would argue its not AI at this point)

Boolean logic is not pattern recognition, its definitionally true and a fundamental aspect of intelligence. Correlation is not causation (100% of people who drink water die!). Actual intelligence can infer how/why on some level which goes beyond pattern recognition (or multi layered pattern recognition)

AI is great for stuff where an exact answer is not needed (or perhaps where even humans cannot say for certain based on available information). Its terrible for things on which you need a correct answer 100% of the time.

Noone defines intelligence as "knowing". Go read scientific definitions and get back to me.

Noone defines intelligence as "knowing". Go read scientific definitions and get back to me.

Computer based AI has been following the machine learning model since its inception. All that has changed is processing power and how quickly the training of these models can be done.

Most experts cannot even agree what constitutes intelligence in humans, let alone machines. By that I mean is human intelligence just a by product of the number of neurons we have and their potential connections? Does every mamal with an equivalent size brain then have human-like intelligence?

You could argue yes to a point, but we still do not have mammals that display human levels of intelligence. Closest is probably dolphins, again experts do not agree.

Point is since we do not know what causes emergent intelligence, the ability to reason, think and come up with novel ideas, we cannot simulate it in software either.

There has been tons of research into this as you can imagine but even models that simulated every neuron in a relatively simple brain did not end up with independent thinking.

Most experts agree that if an AI actually achieved intelligence, it is a given we then become irrelevant since that AI can now improve itself without our help. Ie, we're f0cked.

Irony would be that I've long assumed "artificial" in AI would refer to the source of said intelligence. (non-sentient/non-organic) but I suppose you hit on the fact that its a complete misnomer and not intelligent at all.

Even in cases where AI can sift through massive amounts of data and give you a correct answer, it could never possibly explain "why" said answer is correct. Intelligence is something more than pattern recognition. Even if its say playing chess, where it basically the same thing (pattern analysis combined with some odds) AI gets there deductively, a human inductively, these are different (humans can of course use deduction, but I don't think AI could ever use induction).

I suppose at some level its a bit esoteric and I'm not sure I can explain it better but I don't think its the same or multilayered abstraction of pattern recognition and if you still disagree so be it.

Part of the difficulty explaining it is because there's a philosophical layer here that we as humans don't even fully understand ourselves. What is Self Awareness? What constitutes sentience? What makes a human capable of rationalization and reasoning? What is a soul? Even if you completely ignore any religious connotations, these are questions that don't have clear answers despite the advances we've made in neuroscience.

AI is at its most basic form just another program running off of digital 1's and 0's. "Intelligence" is something deeper than that, and I don't believe machines can ever be made to replicate it.

Some food for thought for you.

What is the difference between a human brain in a human flesh and blood body, and the same human brain suspended in liquid in a jar with nutrients to keep it alive and connections that let the brain communicate with a computer.

Are they the same person?

What about the same human brain, simulated 1 for 1 in a computer down to the gluons that make up the atoms that make up the neurons of the brain. Same identical thoughts, different substrate - software instead of flesh and blood.

If these are identical, has that software then achieved intelligence?

I don't believe a software of 0 and 1 can accomplish "identical thought" as myself.

This is because my decision making model is constantly changing based on the situation and environment that I am in. It's never static.

I use completely different thought models based on my current role in the society. My thought process when I am an employee is different from when I am a shareholder. My short term and long term personal goal changes based on previous outcome of an action, as well as environment change.

So how can a software model all that without a body, real life experience, desire and survival need?

A software doesn't need to worry about social status, survival need, money, personal goal, friends and family. They can only emulate it based on data from the past, but emulation can't be perfect 1:1 if all of the above factor change all the time. Put the software in a new environment or situation then they'll need new data which may not exist from the past.

Then this software may proceed to make decisions on their own without updated data, but is that the real "me" if the whole decision results ended up different?

You are describing the ability to reflect and reason, which is an effect of emergent intelligence. Your thought process is you.

The software is not modelling anything, it is but a substrate, where your brain lives. The same way blood vessels and neurons are your brain's current substrate.

Do blood vessels think? Do neurons? Does the fact there is blood - 70% water, minerals and your body's vessels - running through your brain make you think and have a consciousness?

You care about your body because that is where your brain is. If your brain was in a jar, you would care about your jar being clean and not your discarded body.

I don't believe they are or can be identical. I strongly believe the flesh and blood substrate as you describe it is an important part of what allows us as humans to be what we are. The human body is made up of a complex weave of chemicals and biological material. They afford us sensory inputs and enable complex emotions that a software program can't truly understand or replicate. Software cannot feel, it cannot taste, it cannot see, it cannot hear, and it cannot smell. While it is possible to write a program to interface with audiovisual equipment to create digital records of sounds and sights, I don't believe these are the same at all. I don't believe the software is processing these inputs the same way a human is. And I certainly don't believe it is interpreting and associating meanings and values with them like a human does.

There is a very important distinction between the brain in the jar connected to the computer and the one in the human body. The brain in the jar requires the computer to provide it input and is only capable of interfacing with the world around it through the aide of technology. It does not conjure up thoughts or images on its own, and if you turn that technology off, there is nothing. I don't believe the brain is sentient at this point anymore. Conversely, the brain when connected to the human body is interfacing with all of the biological systems and subsystems, creating sensory input and regulations that are not possible in an isolated environment. And doing it in a different way...a biological way.

Perhaps that is one of the greatest distinctions I think there is to make between AI and humans. We are mortal. We experience sensory input in a different way than a digital entity could. We are connected to the world in a manner that is biological. Our bodies eventually wear out and we eventually die. Our bodies are the vessels through which we achieve sentience, and separated from them I believe sentience and self awareness is no longer possible. And no manner of digital code can ever replicate that.

Yeah, my thought process is me. And this thought process is constantly changing based on the environment and people interaction around me. I never use one single thought process in my entire life. My preference, social role, personality and personal goal always changes because environment and situation changes. A "me"(in terms of thought process) in 2010 isn't the same "me" in 2025 because time, life experience and environment is different, even though we have the same brain.

Environment and social variables on decision making outcome is something so complicated that it can't be emulated easily with only 0 and 1 imo.

That is definitely one school of thought.


The brain does this association. If you could somehow transplant a brain into another body, it would still be the same brain, no? Same person too just in a different body, no?

Does this then not follow if you could brain transplant not into another body but into.. a cupboard, a jar, a bookcase, a swimming pool, anything at all.. is that also not the same person?


Ah, this is where most experts disagree, and so do I. Brain transplants could happen, for one. Also there are conditions where a person has locked-in syndrome for example, where they can perceive everything around them but not move their body or communicate in any way. They are, of course, still sentient despite the inability to communicate.

There are also conditions that deprive you of your senses, one or more. Those people are still human and very much sentient.


This is all true, and you are right it is a distinction. No I don't think a machine can "replicate" it, in terms of having that same connection and social life. Those are all effects of having a flesh and blood body though, are they not?

Would you care about what you wear if you existed purely on the internet? Some humans do exist purely on the internet, and end up looking.. not very human like (hi discord mods).

It's more of a philosophical question. If you, your sense of self, existed only as software, you'd want backups in case stuff happened and probably some clone bodies to move around in physically, if you wanted to be physical.

Take it a step further, if you had the technology in Altered Carbon for example, downloading a human brain into any other body, that is the same brain in different flesh and blood.. moving vehicles of sort.

Are you not still you in that case just because you are using a different body than the one you were born in?

The Altered Carbon TV show is great, by the by. The books even better.

Brain is just reacting to a series of biological inputs to create a response. It stands to reason that sufficiently advanced technology could simulate the entirety of the inputs and reactions, or even understand them well enough to interface directly with a brain kept functional outside of the body.

What we're currently doing with AI isn't that and has no chance of becoming that, though. Same goes for Neuralink.

Edit: Think in terms of physics simulator rather than thought simulator. If the goal is to create the closest replica of human thought, you calculate in the same way the human brain does. The sum compute available to use is certainly adequate to do that, if our modeling of the brain reached sufficient understanding to design the simulation. Most likely, a middle ground will be reached far before then that manages to produce passable intelligence without simulating all the waste, but it sets a baseline for how artificial intelligence is absolutely possible.

WTB
10 Gandalfs for 1T

The biggest current flaw with AI is how the memory works, there is no long term memory system. It is currently all chat log based, and even with the systems that have very large token ratings that is still laughably weak and unimpressive.

The biggest issue is the more people use the tech, the less impressive it becomes. Cause you see all the flaws.

You can't see something as being real, if you understand it has no long term memory system.

Blindness, deafness, mute, etc... are all unfortunate conditions that occur when the biological systems responsible for our senses are either damaged or malfunctioning. An object penetrating your retina can cause irreparable tissue damage that can never be undone. A severe blow to the spinal cord can cause permanent paralysis. Traumatic impacts or very loud sounds, disease, or simply aging can cause permanent deafness out of one or both ears. Damage to the taste buds can result in loss of sensitivity to one or more flavors, or in the worst cases a complete loss of pallet entirely.

But even when they're damaged to the point of non functionality, these systems are still there. Even when we lose a limb or have portions of our bodies amputated, the brain is still connected to the body. There are still inputs and biological processes occurring. I don't think any one individual sense, or even a combination of the 5 together alone are what's responsible for self awareness. And I don't have a definitive answer to what makes people sentient. Our bodies are complex, and despite all the research into functionality, there is still a lot we don't understand. But for a machine, those systems are never even there to begin with.

And with that I'm going to bridge into a topic for you to think about. There is a well known neuologiucal/mental condition that causes progressive loss of cognitive function. I'm talking about Alzheimer's disease and dementia here. I had a relative decades past who came down with Alzheimer's disease. Watching his condition deteriorate as the disease progressed was extremely painful for all those around the man. Eventually his wife made the very painful decision to cut off life support because he deteriorated to the point he was no longer aware of his surroundings. Didn't recognize his friends or family, couldn't remember things that happened less than 5 minutes ago, didn't know who he was, couldn't even tell you his own name.

We know a fair bit about what causes alzheimer's and dementia, although we still have no cure for it. It is a progressive illness that on the most basic level it is the breakdown of biological processes; neural degradation of the brain, leading to eventual death. It is the very embodiment of the slow loss of sentience, to the point people consider the person afflicted in a state of statis or "a vegetable". It's a terrifying illness that only manifests BECAUSE of biological interactions within the body. If that doesn't demonstratably, connect sentience in some way to the biological interactions in our body, I don't know what does.

People have differing opinions of what intelligence is, so you really need a consensus on terminology for specific behaviors to have a meaningful discussion about it.

Anything else is just arguing about what color the dress is.

If the body fails the mind won't be happy either, that's to be expected.

All the more reason to invest in mind transfer technology research.

Considering my role is mostly as an employer than employee, I kinda agree with the sentence that AI is causing layoffs. It is not that an AI is replacing actual people. That is far from true, because the portion that an AI can automate things is only a fraction of the total work.

BUT the thing with AI is… it is a strong bad for comparisons.

Lately when reviewing some outsorce personal work, that soend weeks to generate tests, colleagues were saying: “hey look. This guy spend two weeks to write the tests, when copilot can write then in minutes”.

So, that is the major danger that AI present to employees in general. It wont take your job.

It will only make evident how your work is bad.

So, last few profiles we are writing to hire here, Ive been saying the candidate must have AI assisted tool knowledge.

Simply because it is only fair asking for someone to have that skill of we are planning to demand it to go over what an AI can do.

So, the basic requirement that could become the impicit norm i. The future is that the human must at least produce the same quality work of a AI assisted tool.

That could be the tabula rasa for developers

In people let alone machines.

And THAT is a large part of why I started this thread.

You couldn't be a kid in the mid century and not take IQ tests. I took many. Now please tell me what IQ tests measure OTHER than your ability to take IQ tests.

Hint: If you say IQ you are indulging in circular reasoning.

It is subjective when you say something is subjective.

I could objectively say why this is objective, but you could objectively reply with a subjective argument to snatch an objective win, even being subjectively wrong.

I actually think the large part of why you started this thread is simply because you like the convo.

Depends which IQ test you take. Tell me which you're referring to and I can reply. Seems pretty clear you don't get this though, even though I already outlined the difference between two tests for you.

Wechsler-Bellevue and Stanford-Binet.

Both probably obsolite today. But still ....

There are non verbal IQ tests out there ...