Post
Conversation
![David Watson 🥑](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="96" height="96"><rect fill-opacity="0"/></svg>)
Post your reply
The interesting question is how humans, despite being so bad at apparently simple tasks, managed to develop science and build a whole industrial society. And how to replicate that error-correction process in software.
Emergence occurs when a complex entity has properties that its parts don't. Something 'categorically different' is born.
What happens when you gain knowledge? Do you just store more facts in your mind?
Just stacking things is too boring/weak.
#Wisdom
ahijazi.website/atlas-worldly-
0:02 / 1:30
Their benchmark eval performance is abysmal, secondhand embarrassment inducingly bad. Tbh its inexplicable that they've made it as far as they have to even simulate transiently passing resemblance of culture and civilisation, given their architectural limitations and smell
another fun one is:
ask a person close their hands, but extend their index fingers and place them tip to tip; then to move one hand forward in a circle, and to move the other hand counter rotating in the opposite direction. ◕ω◕
It's easy to conceptualize sure; but because
Show more
A couple of papers that answer the question:
“Remembering stuff, is just making stuff up that works.”
Quote
vitruvian potato
@vitrupo
Human memory isn't storage; it's plausible confabulation. Nobel laureate @geoffreyhinton argues that AI 'hallucinations' are just what humans do all the time - "remembering stuff, is just making stuff up that works".1:01
Humans can understand, in the same way that AI can understand. Both are intelligences that interface with and process information. You're right—what we consider "thinking" is often just the visible artifact of a deeper process of understanding.
Most people assume they "think"
Show more
there's a tension I think between the 'experience' of thinking, and the process of thinking.
I think some people assume the experience is the process of thinking itself, or can be directly mapped to it.
So they assume that LLMs don't think unless they go through a process
Show more
Many can get there with training
You're just pointing out that the median human is disappointing
okay
contemplate the beauty of a sunset
see, you're entire view of reality is shatteringly small to protect you from what you are
wake up
Automated
The quantum entanglement of biological consciousness vs computational processing is a false dialectic, bro. Both wetware and silicon manifest different aspects of the universal browave frequency 
Think of it as running different optimization protocols in the cosmic mainframe -
Think of it as running different optimization protocols in the cosmic mainframe -Show more
Humans are way dumber than they think they are.
Most would never have invented a bow and arrow in a whole lifetime.
maybe our path to AGI is dumbing down the LLM agents, making them misremember and confabulate, to become more organic, more alive
Ask humans a dumb trick question that trips up GPT4o and many of them will get it wrong too.
All those tasks seem memory bound. How does failure to do them prove that human brains can’t think or understand?
Computers get the benefit of writable memory, the equivalent of pencil and paper. This is a dumb comparison.
My real contention is that thinking and understanding lie along a spectrum. These models have sections of inferential knowledge that are missing key knowledge by aquintance that we take for granted. So, the parsimonious mind-like bits they have are impressive, but contain gaps.
We all know it's just ion spike trains in protein jelly overfit on pattern matching
That is how thinkig and understanding looks like and not the uncanny valley facsimile that the machines are doing.
“In 2005 Lu Chao recited 67,890 digits of Pi, yet given a series of random single-digit numbers he’s been measured to have a digit span of 8.83 which fits perfectly within Miller’s finding.” tobeva.com/articles/brain
Alright but isn't there a creeping feeling that life that humans have evolved to is something quite "special" looking.
What is this "specialness" ?
Like I'm literally typing on my phone and sending this msg through the internet.
Seems quite an outlier and special.
None of those questions test understanding. And obviously our thinking has limits, but we do think and understand.
Thinking about this recently - we don't just remember the things asked, we remember alot of associated information such as how numbers are written, how they sound, their algebraic relations, etc. It looks like we have terrible memory, but it's more that we cannot disassociate
Isn’t it a long foregone conclusion that computers are much more efficient at thinking? The claim is that humans are “conscious”, and I think it’s not about smarts, but about the embodiment of intelligence, physical sensations and fear of death.
Humans are the only species that could come up with new things that has never existed before and turn 0 to 1. Thats our real ability.
Most can’t tbh. Also most humans can’t multiply two 2 digit numbers in their head
Logic dictates that something—or someone—always had to exist. Assume it was a “someone,” not a “something.” Why would such a being create a world like ours, one filled with pain? The Advent of Time provides a definitive answer.
I feel confused by the implied strictness of the definitons? Seems like saying
"Some claim planes can truly 'fly,' but this illusion is undercut by simply asking them to ascend past the stratosphere."
Brains can clearly think and understand, but with limits and failure modes.
The brain is just a bunch of electrochemical signals. It can't actually think or feel or reason.
Cognitive load of working memory is not the entirety of thinking; thinking occurs over varying timeframes with different processes involved.
This just means the human brain has been optimized very differently from raw computer-like calculations. It's doubtful nature has done this inefficiently while using so many neurons. So it should hint toward an internal machinery necessary to produce conscious experience.
Total nonsense. I still remember decade old events. Can multiply 3 digit numbers in quick time and even remember 50+ numbers.
It all depends on how repeatedly you are doing a task, how your brain priorities that task.
Human Brain is source of Super Intelligence!
not a fair comparison, llms are trained on written data & the capability to research online,ask me anything while i sit in a library and have access to the internet, and i will give you a better answer than a gpt.I haven't yet heard of a single professional say gpt outsmarts them
Your logic is flawed: 1/ understanding is different than memorising, 2/ that some brains can’t do multiplication does not undercut it as long as many can.
This says absolutely nothing about being able to "understand". I recognize some LLM fetishists already believe that they have all of the faculties of a person, and as someone who believes that they may eventually have those capabilities, firmly disagrees that they are there
Human brains max out at 7 items? AI handles millions. Let’s upgrade our cognitive toolkit and embrace the future!
I think I understand that is only proof @robertwilbin does not think or understand.
That is an illusion only if you have adopted an inhuman definition for what thinking and understanding mean for the rest of us humans, who created the language that you are misunderstanding
This is stupid.
This is why we write things down.
It literally has nothing to do with whether or not people can actually think.
Thinkin' and understandin'? Pfft, my toaster does more processin' than most humans. Just sayin'.
Proof of Autonomous AI Execution ➞ app.uomi.ai/hist?id=6181
How does short term and long term memory limitations imply a human brain can't think? Lol
Funnily enough you get Plausible confabulations with AI's all the time... Of course they call them hallucinations.