We have a problem.....

[SYZ]

Q:   "Which is the larger number: infinity plus one, or infinity plus two?

A:   ∞+1=∞
   ∞+2=∞

Simple algebra then tells me that 1 = 2
assuming ∞ is a constant.

[Mathilda]

Current AI is not intelligent. It can only give the illusion of intelligence because it is trained on data generated by intelligent humans. It cannot autonomously adapt without being retrained. It does not understand any more than auto-complete does. It also relies on a significant amount of human reinforcement learning.

So far, LOL!  Beware of gifts bearing Greeks.  

I didn't intend to reply for more than that, but just before I hit "reply", a thought occurred to me.  When does a major aspect of human (or other) evolution occur?    One day, no hominid uses a sharp stone, and the next day one does.  One day, no hominid thinks of drawing on a cave wall, but the next day one does.  And the same thing about creating a bow&arrow.  Why?  What happened literally overnight genetically or socially?

Maybe one day no AI will be self-aware, but the next day one will.  I'm not trying to act scary about this.  But advances seem to come rather suddenly sometimes.  It makes me wonder.

What is commonly called AI today is one of many subfields of AI, but one that has grabbed all the attention and funding. It's not a route to strong AI, as in actually being intelligent and it never will be. There needs to be a paradigm shift for that to occur and that will only happen once it becomes clear that the failings of the current paradigm can't be overcome no matter what is tried and no matter how much it's scaled up. Normally that heralds in what we call an AI winter, or in this case, when the tech bubble bursts.

How do I know a paradigm shift is required? Machine learning 'AI' needs to be trained. It does not learn autonomously or in real time. How can something be called intelligent if it isn't capable of learning by itself? What is meant by training? In terms of AI it means opening up the black box and changing the weights connecting the neurons. Compare this to an animal being trained for example. Sensory input comes in and the brain will adapt accordingly. You don't need to open up your dog's skull and start probing inside to teach it to fetch your slippers for example.

[Mathilda]

The weakness of purported AI can easily be shown
by asking an easy question of ChatGPT...

Q:   Is it healthier to eat two avocados or four bananas?

A:   Ultimately, both options have their own health benefits!
      It might also be worth considering how they fit into your
      overall diet and lifestyle.

So... this is, effectively, a non answer which tells me nothing, and
in reality throws my question back at me to answer.  What a wank!

My favourite one was when an LLM was asked to explain the health benefits of eating crushed glass and it did so. Anyone with a body that has ever been cut by a shard can imagine the effect of eating crushed glass on their intestines and anus but an LLM isn't embodied and has no sense of identity.

[brewerb]

Q:   "Which is the larger number: infinity plus one, or infinity plus two?

A:   ∞+1=∞
   ∞+2=∞

Simple algebra then tells me that 1 = 2
assuming ∞ is a constant.

Or 1 and 2 = 0.

(oh shit, I'm 1)

[AutisticWill]

The weakness of purported AI can easily be shown
by asking an easy question of ChatGPT...

Q:   Is it healthier to eat two avocados or four bananas?

A:   Ultimately, both options have their own health benefits!
      It might also be worth considering how they fit into your
      overall diet and lifestyle.

So... this is, effectively, a non answer which tells me nothing, and
in reality throws my question back at me to answer.  What a wank!

My favourite one was when an LLM was asked to explain the health benefits of eating crushed glass and it did so. Anyone with a body that has ever been cut by a shard can imagine the effect of eating crushed glass on their intestines and anus but an LLM isn't embodied and has no sense of identity.

"Oh, my! I'm afraid that steamed carrot was too spicy for me."

[Rhythmcs]

I didn't intend to reply for more than that, but just before I hit "reply", a thought occurred to me.  When does a major aspect of human (or other) evolution occur?    One day, no hominid uses a sharp stone, and the next day one does.  One day, no hominid thinks of drawing on a cave wall, but the next day one does.  And the same thing about creating a bow&arrow.  Why?  What happened literally overnight genetically or socially?

In a word, alot.  Especially with a bow and arrow.  None of these things are believed to have happened over night or been the product of a single stroke of brilliance or any individual persons work.  Based on contemporary human (and other primate) behavior we suspect that we were making improvised tools for quite some time beforehand - that knowledge being the foundation of later stone work.  Then you need specialist tools for making bows, and for making arrows.  An understanding of the right kind of lumber and proper cordage.  Some idea of how to store it all and..above all else, a need for the thing.  We seem to have transitioned to bows, for example, because the game changed.  We may have known how to make them or something like them (bow drills) for quite some time but they're not much good for hunting megafauna - so we see spears and spear throwers and pits.  It's only when the forest closes in, and the game is relatively small and quick turning that a bow is even the right tool for the job.

Alot of the industry there is genuinely lost to time.  Try knapping or building a selfbow and when you're done look at the pile of stuff - is there anything there that would be recognizable as industry.  Look at the end product.  Will it still be there for a future archaeologist to find 1.3 million years from now?  Even if it were, would it be recognizable as such or even as anything? If we compare it to ai, and say that today is ai's hh phase...then it seems like we may have a way to go yet between hh ai and hss ai. Maybe it does the work quicker because we've already done it and we're teaching it, and the same appears to be true for different branches of our family tree...we borrowed our earliest tools. They were of other hominids design. I think it's more an issue of continual refinement. We've been equipped with the brains and bodies to go to the moon for 50k years at least. Still took us about that long to get there. We started out with tooth and nail. We can throw stuff. Then we found sharp things and chewed sharp things. Eventually we discover that fire hardens wood. That we can seperate and reorganize plant and animal fibers. That we can make spring traps with branches. Slings with rope. That we can shave growth rings with our stone tools. That we can grind tips, drill nocks, steam out branches, split feathers, boil bones down to glue to fit them better. Mix pigments with oils to stain our arrows for retrieval. Then...then..we get creative for no utilitarian benefit..and make things pretty like we like it.

One of my favorite tools and processes from prehistory is a particular type of club. Obviously you can just find a log or a root and carve it...but you can also dig up a young suitable tree, bind it's roots into a ball, and hang it in the river bank. Let it grow into a club. It's such a collaborative and caring and patient process to make what we assume was mostly for killing people. It's also a pretty direct potential answer to the essence of the question. What changed? Well, the root did. We watched it do that, and we saw it's potential. So we repeated the process, with improvements for the use we had in mind.

[Cavebear]

I didn't intend to reply for more than that, but just before I hit "reply", a thought occurred to me.  When does a major aspect of human (or other) evolution occur?    One day, no hominid uses a sharp stone, and the next day one does.  One day, no hominid thinks of drawing on a cave wall, but the next day one does.  And the same thing about creating a bow&arrow.  Why?  What happened literally overnight genetically or socially?

In a word, alot.  Especially with a bow and arrow.  None of these things are believed to have happened over night or been the product of a single stroke of brilliance or any individual persons work.  Based on contemporary human (and other primate) behavior we suspect that we were making improvised tools for quite some time beforehand - that knowledge being the foundation of later stone work.  Then you need specialist tools for making bows, and for making arrows.  An understanding of the right kind of lumber and proper cordage.  Some idea of how to store it all and..above all else, a need for the thing.  We seem to have transitioned to bows, for example, because the game changed.  We may have known how to make them or something like them (bow drills) for quite some time but they're not much good for hunting megafauna - so we see spears and spear throwers and pits.  It's only when the forest closes in, and the game is relatively small and quick turning that a bow is even the right tool for the job.

Alot of the industry there is genuinely lost to time.  Try knapping or building a selfbow and when you're done look at the pile of stuff - is there anything there that would be recognizable as industry.  Look at the end product.  Will it still be there for a future archaeologist to find 1.3 million years from now?  Even if it were, would it be recognizable as such or even as anything?  If we compare it to ai, and say that today is ai's hh phase...then it seems like we may have a way to go yet between hh ai and hss ai.  Maybe it does the work quicker because we've already done it and we're teaching it, and the same appears to be true for different branches of our family tree...we borrowed our earliest tools.  They were of other hominids design.  I think it's more an issue of continual refinement.  We've been equipped with the brains and bodies to go to the moon for 50k years at least.  Still took us about that long to get there.  We started out with tooth and nail.  We can throw stuff.  Then we found sharp things and chewed sharp things.  Eventually we discover that fire hardens wood.  That we can seperate and reorganize plant and animal fibers.  That we can make spring traps with branches.  Slings with rope.  That we can shave growth rings with our stone tools.  That we can grind tips, drill nocks, steam out branches, split feathers, boil bones down to glue to fit them better.  Mix pigments with oils to stain our arrows for retrieval.  Then...then..we get creative for no utilitarian benefit..and make things pretty like we like it.

One of my favorite tools and processes from prehistory is a particular type of club.  Obviously you can just find a log or a root and carve it...but you can also dig up a young suitable tree, bind it's roots into a ball, and hang it in the river bank.  Let it grow into a club.  It's such a collaborative and caring and patient process to make what we assume was mostly for killing people.  It's also a pretty direct potential answer to the essence of the question.  What changed?  Well, the root did.  We watched it do that, and we saw it's potential.  So we repeated the process, with improvements for the use we had in mind.

You make some very good points. Most human advances come gradually, small bits at a time. Accumulations of small changes.

However, I chose the several I mentioned only because they seemed to come rather suddenly and perhaps individually. Some individual hominid probably realized a smashed stone delivered a sharp edge. Some individual got the idea of blowing a handprint or trying to show an animal on a cave wall.

The bow seems one of the oddest inventions. Some individual got the idea of bending wood with some sinew from tip to tip and the idea of a "small spear" launched from it with useful force. And even then, it required some concept of a nocked "small spear" bcause you can't launch an arrow otherwise.

I'll grant credit to your idea of it coming from a bow drill, though. I was thinking the weapon came first (mostly because a bow drill has to have a loose string). Either might have come first and I wasn't thinking of a bow-drill when I posted. But either is a rather sudden invention.

Thank you for that post!

[Rhythmcs]

If you use bow drills often you see the beginnings of a bow on your scarred palm. Under the tension you apply to get good twist, the friction rod has a tendency to jump out (or up, depending on the type) and bite you. They end up getting nocked to solve this problem, the nock being a part of another type of weapon or tool - the atlatl. That double nock, btw, et voila, we've invented the spindle, which is crucial to the metal lathe, the foundation of all modern industry - though we wont see a lathe, let alone a metal one, for many many thousands of years. If we reference the remaining hunter gatherers on earth we'll notice that a bow isn't necessarily used as a "long range" weapon. The hazda, iirc, jog their prey down to exhaustion first, and then the coup de grace is delivered from near point blank range by a very simple bow that they also use to start fires with. Like a captive bolt with a few extra feet on it made by zippo. Until the invention of metal tools, atlatls had more range and killing power than bows. The "spear" on these spear throwers is in fact no such thing. It's a flexible dart that releases spring force. The idea for the body of the bow (whether a fire drill or as a weapon) probably came from the dart if it didn't come directly from the tree - similar situation with the arrow.

Bows are weird. We discover them, lose them, rediscover them. Multiple instances of independent construction across continents oceans and eras. They might be at the edge of what I'd call naturally realizable weapons. We only need to get hit in the face with a limb to grasp spring. We only need to cut ourselves on a rock to grasp an edge. We only need to see a vine to understand cordage. Watch a bird to understand feather control. There are simpler and fully usable versions of every piece of a bow - the only thing you couldn't put together properly in just one afternoon of nature watching is the system as a whole itself and the tools required to produce a passable version. Couldn't automagically come up with the right kind of tree either, I guess.

Ceramics are another one, imo. Sure, all the material is out there in nature, and the earliest forms look a hell of alot like gourds we think (sometimes know) were already in use...but as soon as you start to make -alot- of ceramic stuff you'll hit on a bunch of process improvements (and waste products) that aren't intuitive. So....bricks and bowls, sure....but producing iron ampules from contaminated creek water while firing bricks ...we probably never saw that happen "in the wild". We're crafty, though...people tend to imagine prehistoric life as a chaotic shitshow of nightmare scenarios back to back, but we know we spent more time sitting around trying to make shit..than using shit. So it becomes a chicken and egg thing. Did we become crafty because we made tools, or did we make tools because we wanted to be crafty? It wouldn't surprise me if some things of the chain of discovery for complex items boiled down to trying to make a smoother dildo.

Which we did...alot. We have more and older and better constructed dildos than ancient bows. I'd caution against the notion that the bow appears suddenly or individually. They don't preserve well, and neither do their parts, when we recognize them -as- parts. Like with boats...a preserved item doesn't and can't imply singular brilliance, it implies industry. That there were enough of them floating around for one of them to have survived..and survived in the spot we happen to be looking, no less. We never find the first one (even in these artificially constrained chains of custody, ala "first one in this valley"), and there are alot of versions in between whatever that was and the earliest one we find.

Here's a simple q to illustrate the problem. The oldest known bow dates back to about 7kbc. Who..then, sailed across the ocean to tell the americans about them? Why were western north american self bows so incredibly similar to mesolithic european bows?

[pattylt]

Thank you…I love getting an education!

[Rhythmcs]

The whole tool thing is the greatest story every told, at least about us anyway, imo. If alien archaeologists ever come down to a long dead earth they're going to have a hell of a fun time figuring us out.

*another example of the above, that I'd feel like a failure for failing to mentioning. Agriculture goes back to about 11kbce in the old world. Here again..who took the time to tell the americans about it?

[Cavebear]

Q:   "Which is the larger number: infinity plus one, or infinity plus two?

A:   ∞+1=∞
   ∞+2=∞

Simple algebra then tells me that 1 = 2
assuming ∞ is a constant.

Infinity is not a number.  Infinity plus 1 is the same as infinity times infinity as is a google of infinities.  I know you know that, just had to say it.  

When I was in 9th grade, the teacher offerred a A+ to anyone who could prove 1+1=2.  I spent nights at it before I realized that numbers are just concepts.  I might have 1 finger and another finger but that didn't define  what "2" meant.  "Counting" might have been the origin of math, but physical experience isn't about math.  Or rather math isn't about counting.

In 10th grade, the geometry teacher made the same offer about trisecting a line of uncertain length without a ruler or compass.  Needless to say, I failed at that, too.

But I sure tried.  I learned about maths.  Not that I am any good at them, but I learned a lot and I'm sure that was the point.  Well, OK, I was really good at plane geometry (it was all visual to me) but I never pursued the subject further.  Algebra just hurt my head (I hate variables).  I failed calculus even though I understood the concept.  I hate "e" and logarithms offend my sense of geometry.  

Oh what the hell, I'll post this...

[mordant]

The weakness of purported AI can easily be shown
by asking an easy question of ChatGPT...

Q:   Is it healthier to eat two avocados or four bananas?

A:   Ultimately, both options have their own health benefits!
      It might also be worth considering how they fit into your
      overall diet and lifestyle.

So... this is, effectively, a non answer which tells me nothing, and
in reality throws my question back at me to answer.  What a wank!

My favourite one was when an LLM was asked to explain the health benefits of eating crushed glass and it did so. Anyone with a body that has ever been cut by a shard can imagine the effect of eating crushed glass on their intestines and anus but an LLM isn't embodied and has no sense of identity.

Researchers at Apple recently concluded that LLMs are too unstable to do math. Answers can be changed simply by substituting a synonyms for certain words. Current benchmarks don't adequately take this into account. In their judgment no amount of fiddling or scaling up will change this.

Every use I've made of the LLM now embedded in my software development environment tells me the damned thing is faking it to make it but has no actual understanding. Since most people today have few to no reasoning skills, what LLMs do impresses them. In practice all I use it for is to speed up writing simple boilerplate code. Its only other value is to give me a good laugh when it makes spectacularly bad guesses.