Saturday, February 24, 2024

AI Artbots: Nobody Home

 




1. They are NOT Artificial Intelligence. They are Large Language Models – they have NO intelligence and cannot reason or think. What they do is learn patterns of numbers and repeat those patterns.

The reason they are being called “Artificial Intelligence” is marketing: the spruikers are trying to drum up investor cash to keep their new companies funded until they come up with something that actually makes money.

2. Artbots will not replace human artists. Well, not in the near future. You still need someone to tell the artbot what to “draw” and apply quality checks to what it does.

Yes you can get the artbot to draw nice images that could be photos but they can’t be anything really “original”.Whatever that is.

3. Artbots can’t make content with consistent characters and sets. Yet. Maybe in future. 

If you want a one-shot image of something, fine – but if you want to make a full movie of a character’s adventures the best you can get is that the Artbot will do it all in one go: Reshoots with the same scenes and actors are going to be a problem. This is mainly a problem of the nature of Artbots: it’s all about processing power and memory. In theory if you had enough computing power to devote to this task you could churn out movies directly from a terminal just by typing in commands – but I doubt that they would be very good.

Big computing power costs big money and time on big systems and that means big studios will be the only ones trying that. Maybe the big studios will try to do this but then they won’t be “studios” any more, just server farms. Big ones.

Even if you got the scenes to all be consistent and the actors to be consistent you still need a script.

See below.

4. Artbots can’t write. Well to be exact they don’t write. They can absorb a library of books by human authors and assemble something from, say, five or six of them into a script but once again There is nobody home: you will need a human writer and/or director to read the script and check it to see if it even makes sense let alone makes a good story.

If we are talking movie making it needs to be interesting enough to get people to watch it which is another thing again.

If you want a whole lot of sequels it is going to work just fine but if you want something new, with a new style and new ideas . . . .  forget it. 

Sounds like Hollywood already, right?


So there it is: I predict that big studios will pour zillions into “AI” movie machines and probably tell us all that it will bring “A new age of movies” or something like that, and no doubt there is will be some successes – but in the end they will still be dreived from previous movies and media and not really anything original . . . . and you will still be able to see hints of the artifical nature of them in places.

Small Independent movie makers will pop up doing their best to provide real live action works that don’t go big on special effects and have real actors who you can see in other media and people will always connect with that because we need real identities to identify with in stories.

People want real people in their stories. Unreal people can still be attractive but they don’t have a real existence so they don’t have opinions or families or get in car accidents or have cousins or do dumb things that tell us they are real fallible humans.

As stated above, AI is a sales pitch, not a real thing.

Getting robot electronic brains and hardware to work well enough to replace humans in menial tasks (General Purpose AI) is a big, big, BIG challenge and our hardware and software is still not up to it.

Nowhere near it.

Robots that operate in pre-prepared situations and do limited operations such as the robots on a production line are plentiful and cheap because they are very simple.

General Artificial Intelligence is not near and when it is eventually reached (if ever) it will be running on huge server systems because the requirement is so huge that nothing on a desktop will have enough processing power and memory to do it.






Tuesday, January 16, 2024

A MATTER OF GRAVITY

 

 

(image credit: ABC Australia)
 

Take a look at fossilised dinosaur bones that have been dug up and put on display in museums around the world. These are reconstructed to portray dinosaurs that stand five, six, ten and up to eighteen metres high and theoretically could weigh up to 60 tons. Wow. Watch the movie “Jurassic Park” for recreations to see just how huge these things were. We also have fossils of giant insects and plants from the distant millenia.

Now consider the elephant. You know how big they get and how slow they must move because they weigh so much and they are pretty much at the limit of size for a land animal because bone and muscle just can’t hold up anything bigger.

Yes, you can argue this point : go ahead, prove me wrong, but I bet you can’t. Bone and muscle obey the same rules of engineering that steel and electric motors do.

Are you seeing what I see? Either our ideas of the strength of bone and muscle are very wrong or these dinosaurs could not have even walked on this planet under the current 9.8 M/S sq gravity.

I don’t think the archaeologists are wrong and that the dinosaurs were not that big or anything like that, I have a much simpler answer as to how these giant creatures walked the Earth:

The gravity was much lower back then.

This then brings the question of “What is gravity and how could it increase over time?”

I like to think of gravity as “Coactive mass”. The mass of Earth attracts us because all masses attract each other and it is millions of times bigger than we are.

If the force of gravity is the result of the total mass of the planet then as long as mass of Earth remains constant, gravity will remain the same.


Except it doesn’t.


Every day an average of three tons of solid matter in the form of asteroids, dust, rocks etc. falls from space and adds to the Earth’s mass. 3 tons a day over a million years is 365 million tons.

Yes, some gasous matter constantly leaks off the edge of the atmosphere in the solar wind but that is still well below 3 tons a day in total mass, and it’s gas, not solid.

Don’t take my word for this, you can check up on it yourself.


If a really big lump of matter suddenly hits the planet, the total mass will jump up and a lot of animals and plants will need to adapt or die. This event would also cause huge geological upsets as the forces and masses would sudddenly change.

Sounds familiar huh? Maybe those huge dinosaurs didn’t die because they starved in darkness, but because they could no longer walk. Much smaller creatures would be less affected by this of course.

If there were humanoids in the distant past they would have grown taller than modern man simply because the force of gravity was less. Did we have bones of a much taller race of men who came before us that are being hidden or destroyed? Some people are quite sure of this and there are a lot of records that suggest this - but it seems that someone has been busy getting rid of all of those giant skeletons. 


This brings me to the last big question about all of this. 

How come I seem to be the only person who has figured this out? 

Why is this not being spoken of out loud?

Maybe it is because when another large asteroid hits Earth, according to the ideas set forth here, it will not only cause environmental catastrophe but if it is big enough, all of our cities, infrastructure and systems will fail because they are not designed to stand under more than one gravity. 

Or then again, maybe people just don't like the idea that gravity might not be as consistent and constant as they like to imagine it.    




Wednesday, January 10, 2024

THE CNC PROJECT PART 5 - THE CABINET

 

The idea was to make the cabinet from 2020 aluminium extrusion since I had a lot of it and it would permit putting clear plastic sheet in the slots for windows to see what the machine is doing.

I could not find anyone else who had done this - all of this is my own idea.

 

Normally with 2020 I used internal corner brackets to join the parts and everything bolts together using bolts or nuts that fit the slots. This is not possible for most of the cabinet though as I need the slots you would normally use for internal corner brackets for the panels. The one exception is the back wall which needs to be removable after the frame is assembled so the panels are fixed to the back rather than in the slots. You do this using the “T” nuts and bolts. The neat trick with “T” nuts is that you can slip them into the slots and then as you rotate the bolt, the nut rotates 90 degrees until it locks in the slots and then grips. This is okay but in practice it can be a pain when you have a panel with six T nuts and you need to get them all to lock: if any of the bolt holes are even slightly off, the nuts may jam and not rotate.

 


The other option was cast metal corner blocks. They have fingers that slide into four slots and grub screws that dig into the metal in the slots to hold the bar ends in. I did the first assembly of the cabinet using these.

This may be fine for some things but for this project the grub screws just don't have enough grip. The front of the cabinet has two doors that hinge up and they need to be strong just to open without problems so I had to come up with something better.

I made two jigs: one to drill the holes in the ends of the bars with the corner block fitted and one to thread the hole in the centre of the 2002 bar.

The drilling jig was just a clamp in the drill press with a spot to allow the finger of the corner block to stick down.


To make a frame with window in it in 2020 extrusion:

1. I cut the 2020 lengths to fit the corner blocks. I used a drop saw with a fine toothed blade, makes a nice clean cut.

2. I rough assemble the frame and mark the fingers that need to be cut off: you only need three and there are four. Chop ONE of the flat protruding ones.

3. Glue the corner blocks to the end of the 2020 with JB Weld. Make sure you put enough JB Weld in there and let it set properly or things might not go according to plan. JB Weld is amazing, never seen a glue that sticks metal so well. Of course, the usual gluing rules apply: roughen surfaces and clean first.

A prepared end with the block glued on and the hole drilled: (X)

Note that the finger points down. It is a great help for aligning the other 2020 bar for assembly.

4. Mount the ends in the jig and mark, centrepunch and drill the bolt hole, then countersink it so the bolt is set into the corner.

The bolts are M5 x 20 countersunk Stainless (natch).

The Tapping Jig:


 

yup, cheep plastic but I don’t care, it holds the tap in line and that’s all that matters.


I used 3mm plexiglass and added a strip of “D” profile sealing strip on the edges on one side. The slots are 6mm so the foam strips fill the gap and as they are outside, if something does fly off inside and hit the plex the foam strips will absorb some of the impact.

The solid panels are all a special panel that I stumbled on – it is arylic plastic with a thin layer of aluminium on both sides. Not sure what the original purpose of it was though: I had some lying around so it went into the project.

 

The Hinges

I have two big doors that hinge upward so they have one external hinge and one internal. The external was easy as I had some plastic hinges designed to fit the 2020 system but the internal was a very different story.

They needed to go inside the cabinet and not collide with the clear panels – there was nothing around that would do this, so I bought some much bigger hinges and cut them down to fit. This was quite involved and the hinge pins turned out to be sleeved in plastic which melted when I cut the hinge plates with an abrasive disc cutter.

I ended up replacing this with some heatshrink which seems to have done the job just fine.


Here are the external hinges on top of the cabinet:

 

You can also see the Stainless Tee plates I used to fix the top crossbar on. The top ones are doubled for stiffness and this is one of those places where being able to loosen the nuts and slide the whole thing in the slots is a big advantage – it really helped get the doors to line up.


Now on to the internal hinges:



  The bolt hole arrangement tells you that you won’t buy any hinge to fit there.

The bumpers keep the doors apart when they are open.

I also added light bars to the outside of the cabinet and once again the JB weld saved me a lot of trouble: the LED strips came embedded in an alu extrusion and I was at first stumped by how to attach them – but then I saw someone on youtube called Project Farm. The guy tests all sorts of things with measuring equipment and bar charted results, and he reckoned JB Weld was good. Yes, his voice is a little bit high but that’s not his faultThanks, Project Farm.

I put in a lot of work into making my own limit switch mounts for the X and Y axes but the Z had me stumped and besides it looks like getting them to work with the controller would be a pain so I am going to get the Elite Upgrade kit which includes a new controller, servo motors, drag chains and limit switches. I am not pleased by junking stuff I just finished but there are standards here that really should be met.

I might try and use my drag chain system and some other parts too such as side fences to keep the y rails and crews cleaner.

Here is one of them:


 You can also see the plugs that seal off a lot of the holes in the mounting brackets which I bought off etsy in a set. A good idea to save lots of cleaning up later. 

Here is one of my 3D printed limit switch mounts :


 the blue bit is the reed switch. 

That's all for now,  gotta wait for my upgrade kit to arrive. 

The bulge in the cabinet still has me stumped though.

Saturday, November 4, 2023

THE CNC PROJECT

 

Part 4


THE DUST BOOT

The milling spindle of the CNC spins at thousands of RPM’s and it spits out dust that goes everywhere if you don’t have some way of collecting it: this is the job of the Dust Boot.

The standard one that you can buy for the OneFinity has a socket for the suction hose in front of the spindle. This is not workable for me because it not only blocks vision of what the mill is doing, it also sticks forward and that won’t fit in my cabinet – there just isn’t room.

In my trials assembling the bench to mount the CNC on top of, I made the benchtop 1220 deep as that was the width of plywood available.

I thought about patching another bit of ply on the end but that would have been very hard to clamp on straight while the glue dried and even then I had doubts about the strength and flatness of such a patched panel - so 1220 it is.

I needed something that sucked up the dust from the back where there is a lot more room due to the design of the 1F.

 


I looked around and PWN CNC made one that looked pretty good. It wasn’t cheap and it is mostly 3D printed parts but that’s okay – the variety of parts make it flexible and it can be set up to fit behind the drag chain, something I really needed.


The Suction Hose

The back end of it runs up through a clear tube and then joins the first hose piece through a magnetic clip which is pretty neat. You can pull off the hose and use it to clean up any overspill inside the enclosure.



 

This then runs through the flexible hose to the back right corner of the bench and down underneath to connect to the shop vac.

This is still in the trial and error phase of things – the middle of the hose hangs from an arm that is supposed to stop it fouling the head as it moves around but in manual testing it is not yet working properly yet: it jams or the magnetic fitting pops off.


Fixing the shoe mounting

The back end of the shoe is dragging on the deck. This partly the mounting arms and partly the load on it from all of the hoses and fittings.



At “A” you can see the bolt head where the two parts of the mounting arms are fixed together: this join is loose which contributes to the back of the shoe dragging on the deck. “B” is the slots where the shoe slides fit in and they are loose too.

You can see the sag at “C” and the duct tape. That’s temporary.


I bought a pair of mini drawer rails to mount the arms to the carriage.

This is still not ideal but it is better than it was.

I want dust shielding for the mini rails. Not sure exactly how but I will work something out.


 



mini rails test fitted to the carriage.



They don’t need to be perfect since the rails will not be moving when the machine is operating but it is still important.


Here is a diagram of my idea for mounting the dust shoe on the rails.

The arms need some sort of locking mechanism that allows adjustment.

I would like to have springs in there too – or some sort of friction brake so that when it hits something it will move up and not break. Well, that’s the theory.

This is all good but there is also a serious space limitation that it needs to fit in since adding to the width of the carriage will cut down on the working area of the machine.


 



I might 3D model the whole carriage so that I can design the rail covers and other bits since it is a narrow area and there is a lot of bits to fit in.

Thought some more about shoe drag and came up with this: the arm that holds up the hose should be sprung so that there is almost no weight on the back of the shoe: then the shoe will be just fine and the adjustment problem disappears. This hose arm also needs to provide pivoting for the hose but I think both those things can be achieved in one device.

I still need to finish the shoe mount and add a position setting widget but this will solve a lot of the problems.


The hose is nearly touching the back wall at most rear position so the hose will need to curve forward before it goes back: moving directly back with the carriage at the far right will cause maximum curving of the hose.


----------------------------------


A change of plans

I changed my plans for the dust shoe mounting as they were just too complex and as soon as things get dirty those drawer slides would have gummed up.

As this part of the project had stalled I dropped all of the fancy stuff - and since I now have a really good 3D printer (Plug here for the BAMBU LAB X1 Carbon!) I designed a new mounting arm and got it right on the second iteration ( I’m getting better at it now, the first couple of 3D printed parts I designed took five or six tries )

Printed and test fitted them, yup, all looking good. YAY!!!! Finally.

Also printed new knobs for the Dust Shoe mount and fitted them to the new mount arms. All is good so far, next up is making a new part to extend the tube running back from the shoe to the vertical tube.


This is the CAD pic of the new shoe mount in Blender. This is the inside face, the outside is flat.


 



I have no idea how the old ones were supposed to work - with the 80mm spindle mount installed they won’t fit. Mine has a cutout to allow the spindle mount to get past, the bolt holes are correctly aligned and the knobs on the bolts are shaped to avoid hitting too. The are designed to fit on Hex head M4 SS bolts. I used the same magnetic locking system as both Onefinity and PWN – 6mm dia x 4mm magnets. I used four bolts rather than two to make things stronger since the big issue here is the load on the the back of the dust shoe.


---------------------------------------


Here is the new setup. I used the clamping plates ( metal plates that run in the back of the vertical slots) from both the original and the PWN CNC mountings so that there are four bolts holding it on.


The shoe slides bolted to the PWN shoe itself that slot into the arms were too narrow and slopped around so I glued some ABS sheet on and filed it down until it was tight in the shoe mounts.

Below is the pic of the shoe installed showing the mounting arms and knobs.

You can also see in the second pic below my 3D printed side screen mounts with 3mm clear plex screens, the Estop button fitted and the white box under the bench is the power meter so I can see what sort of current the whole thing will pull.

There is also a small LCD above the Estop that gives a temperature reading for the cooling water tank.


 




 


 . . . . . . and this is where I got to just about the same time Onefinity announced the Elite Series upgrade kit. 


 The control panel is finished, the lights work, most of the fabrication is done but there are still a few things to finish such as getting all of the limit sensors mounted - I was making my own because I really wanted them and they didn't come in the box. . .

 . . . but I was still having problems - mounting the limit switches on the Z axis was one, but the software was a much bigger one: I do not want to get into coding.  The whole point of this machine is to be able to use it to make parts for other projects.  It was not supposed to be a major project itself. Yeah, I know. 

So I'm getting the Elite upgrade kit. It solves a lot of my concerns with the original unit namely the lack of limit switches and step counting - and perhaps more importantly, software improvements. . . . . .

I hope to use a lot of my exisitng parts but we'll see about that. The dust collecting bits should not change and if you want the STL for the better mounting arms just email me. 

The biggest problem I can see with installing the Elite kit in my setup is that the X servo will stick out past the wall of my cabinet : I will need to make a new right side wall with a bulge in it for that.  I have already figured out a way to fit the Y servos by adding a "bump" but they are not going to move. That bulge is not so easy.

 I originally made a plywood cabinet to put the control boxes in but then learned from the Onefinity forums of troubles other folks were having due to RF interference, particularly with VFDs and unsheilded cables so I dropped that and bought a PC cabinet which I shored up with some sheetmetal to make a well shielded box.

So there it is, looks like it will be early next year before I get it running but I am not in a big rush.  Better to do things right even if it takes longer.

I was going to do a Part 5 with all of the wiring but that wil need to be redone with the new system and cables.  




 

Monday, August 21, 2023

The Management Problem

 A long time ago I saw how there seemed to be people in positions of power who really should never be in power.  I called it "The Management Problem" because I think it is the biggest problem for humans ever.

This video talks about it in detail and offers suggestions for how we can solve a lot of it.  It is long but very worth watching.  Thankyou Brian Klaas, I wish that more people can learn from your work.



Monday, May 8, 2023

 

THE CNC PROJECT


Part 3


DRAG CHAINS


What are drag chains? 

 They are a sort of hollow chain that is used to carry the cables from the fixed part of a machine to the moving part while protecting the cables and stopping them from getting tangled up.

The OneFinity I bought did not come with any and I wanted to add a milling spindle with water cooling so I bought some and made my own mountings to fit them to the 1F. 

I first looked at what other people had made on the 1F forums and one person was selling 3D printed drag chain fittings - but when I looked at them I was not happy.

There are three parts to such a set: the Top Mount, the X End Bracket and the X Bend Mount that curves all of the cables 90 degrees and then runs them into the second drag chain which goes to the benchtop.

This is his Top Mount:



I may not be a qualified engineer but I am pretty sure bending cables back 180 degrees is not a good idea.

Here is his X End Bracket and X Bend Mount:



Not bad but I think I can do a lot better. I also want to make sure that my cooling pipes from the spindle are smoothly routed as any sharp bend will restrict the flow.

I started with cardboard and sticky tape: I stuck the bits onto the machine with masking tape and marked where things went and fiddled with things until it all worked right, then drew the shapes carefully with compass and rulers.

Here is my pattern for the most important part, the X Bend Mount:


The idea is that the fluid pipes will go round the outside of the curve with the wires on the inside. The side walls are to be built up in two layers of 5mm thick ABS sheet while the bottom is 2mm ABS sheet.

The mountings use “L” shaped bits of the same 2mm sheet and three are used – two for the Bend mount and one for the end mount that is only there to support the aluminium rails that the drag chain rests in.

Here you can see the patterns and the parts ready to be cut from the sheet:



On the left are the side wall patterns and below them the bracket pattern.

The parts were first roughed out and then the sidewalls were glued together to final thickness before lots and lots of cutting, filing and sanding.

Note that the glue faces must be sanded first and cleaned to get a good gluing. I used “superglue” - I have some experience with it.

Note also the use of masking tape to allow drawing on black plastic sheet, an otherwise impossible task.

The mounting brackets needed two bolt holes in them so I drilled these and used the holes to bolt all three together for easier shaping.

During the production some details of the design chainged but that is normal and the main parts were checked repeatedly to make sure they would fit as planned.

This took some time but I think it all came together quite well.


Here are the finished parts: On the left, the X End Bracket, right is the X Bend Mount and below is the Top bracket.



My brackets have a step for the drag chains to mount on so that the cables run smoothly on and off the chains.

The Top bracket has a magnetic fitting under it where a clip for the dust boot pipe fixes on – you can see the magnet in the picture.


Here is how they look fitted to the machine: you can see the rails that the drag chain rests in on the X axis. Yes, I know it’s not all polished but I don’t care: it’s not worth the trouble for a piece of industrial machinery.


When the pic was taken I was getting the rails for the Y chain ready to be fitted. They are “L” profile alu extrusion screwed to the benchtop.

You can also see back end of the dust shoe I bought to run the suction back and behind the drag chain rails. More about that in part 4.

Here is another view of it, this time with the cooling pipes fitted and . .

a rope.



How do you get the cables and pipes into the drag chains?

This is my answer.

Get a rope and a driver bit. I used a phillips bit as it was less likely to catch on anything. It had a narrow collar near the back and I used a small bit of thin iron wire to fix it to the end of the rope and then covered it all with duct tape to make a smooth outside.




The last part is a good strong magnet.

I threaded the end into the drag chain and then use the magnet to pull the driver bit along inside the chain from outside. When I got to the bend I went right up to it, then moved the carriage so the bend went past the driver bit then carried on pulling it with the magnet to the end. 

 Now the rope is all the way through and you can fix your cable to end of the rope and pull it gently through. Note that this really only works well with bare ended wires – you need to add the plugs etc. after threading.

Here is the same X Bend Mount in a more finished state with the cable cover plates fitted. The X motor cable runs over top and through the small white clip.


Finally, here is where the cables and pipes run through the benchtop:

I bought a lot of grommets and plugs for this project and here you can see some of them.

 See you in part 4