Dr. Jim in Robot Magazine
mallred
Posts: 122
For those of you interested, Dr. Jim has been published in the current (July/August) edition of Robot Magazine. The article is entitled "Machine Intelligence".
This article may answer some of your questions regarding Dr. Jim's approach to machine intelligence.
It may also open up new questions.
Thanks,
Mark Allred
This article may answer some of your questions regarding Dr. Jim's approach to machine intelligence.
It may also open up new questions.
Thanks,
Mark Allred
Comments
Well, I'm a sucker for slowing down to view auto crashes, so I will reluctantly watch this train wreck.
Maybe I'll be pleasently surprised. I dunno.
But I remain convinced that Gouge's first priority has to be firing Mark Allred.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
JMH
Post Edited (James Michael Huselton) : 5/23/2009 4:20:17 PM GMT
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
For me, the past is not over yet.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
JMH
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
For me, the past is not over yet.
Never Mind...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
JMH
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us are now in stock. propmod_1x1 arrive on 26th. Only $30
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
Yes, you're correct. I am being snyde and petty. My behavior is not professional and unbecoming of a gentleman. The man deserves better.
Mallred, I apologize.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
JMH
mallred, thanks for posting.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
For me, the past is not over yet.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
JMH
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cheers,
Simon
www.norfolkhelicopterclub.com
“Before you criticize someone, you should walk a mile in their shoes. That way when you criticize them, you are a mile away from them and you have their shoes.” - Jack Handey.
Pretty much just caused more questions for me... Since it was just an article on how to wire up a propeller PCB to his expansion device. But maybe others will understand it better...
http://www.botmag.com
Direct Link to article...
www.botmag.com/issue17/Processing_Platform.pdf
Post Edited (bmb!) : 5/27/2009 3:44:33 PM GMT
"In the following example of self modifying code, the WRLONG instruction would not execute properly on the Propeller without inserting a NOP instruction between the MOVD instruction and the WRLONG instruction that it modifies. This NOP could be removed when the code is run on a simulator and the routine would still execute properly. A simulator would not reveal this or other subtle timing issues which are the most difficult to troubleshoot. Using a simulator in a case such as this would serve to confuse the developer (that would be you) because the code would execute just fine without the NOP instruction on the simulator but would not execute properly on the Propeller".
"Since a simulator can’ t provide the timing environment of the target machine, it can not be used to develop timing sensitive code".
Really ? Guess I know jack then.
It did tell me a lot more about his memory expansion board... 4 control lines, 8 data lines... looks like he has three-byte addresses, two latched from the data bus, and the lowest eight bits latched into pre-settable counters, probably two 74<flavor>163's, whose numbers have been blacked out. A decent and quite obvious circuit, I considered using one like it before.
Why he is using 74LS series chips is beyond me, they need 5V, and he does not appear to be taking any precautions with interfacing to the 3.3V IO on the propeller.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Largos - a nano operating system for the Propeller
www.mikronauts.com - a new blog about microcontrollers
However, with the Propeller and its pipeline and its dependence on self modifying code to do anything interesting there is no way you can get away with a simulator that does not take these features into account.
Has anyone dared to publish a prop simulator/emulator that does not? It would get flagged as a bug almost immediately.
By the way does anyone know what happens if you ignore the 4 stage pipeline in an 8086 emulator when it is executing self modifying code? Presumably this problem exists on all pipelined processor architecture. Its just that normally self modifying code is taboo.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
For me, the past is not over yet.
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
Post Edited (Leon) : 5/27/2009 7:01:30 PM GMT
The types of errors cited so far are typical when working in isolation, without the support of a community like the Parallax Forum. Hopefully, Dr. Gouge will, at some point, see fit to join our vigorous fray as a full participant, where he can contribute to, and be guided by, the collective knowledge available here.
-Phil
It's entirely possible to create a simulator which does run code which will not run on the Propeller but it's not much of a simulator if it does, simply because it will also not run all code correctly which does run on a Propeller.
It may be true to say some simulators may not detect this issue, but it's completely false to say they cannot.
At least there's some good news - "As an intended side effect, this modified Proto board will also be prepared to attach to the memory expansion board which will be constructed and explained in the second article". So hopefully how the memory works will be revealed.
I believe someone here is doing exactly that for the Prop. Not to mention how the Prop was developed in the first place.
I wish I had the time to think about creating a COG in VHDL which could be run under the open source GHDL compiler.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
For me, the past is not over yet.
In the end, I believe that the percentage of participants who do come here will be relieved by what the community has to offer and will themselves contribute. It seems to me at this point that the more magazine articles that are available to promote Propeller and the ability-wealth of the community, the better things will be. You always hear about AvrFreaks.com as a strong, helpful community. Parallax clearly understands the positives of this community. I honestly think we should be a little less zealous around here though.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
--Steve
Propalyzer: Propeller PC Logic Analyzer
http://forums.parallax.com/showthread.php?p=788230
Even if mallred's post WAS a shameless self-promotion, how many times have so many of us done this very thing? How many times have we been super proud of some project, or publication we did, and wanted to show it off? I know I have. [noparse];)[/noparse]
To bad mouth something like this, IMO, can only give bad publicity and hurt the image of the Propeller in general - even if it is self promotion.
I can't say that I've noticed any other cases where interest in the Propeller has generated such negativity, and one has to ask why that is. It seems to me to be basically a culture clash. Their arrogance, unwillingness to answer questions, and 'storming off in huff' got my back up. Their published view on simulators is quite disrespectful to people on the forum who have done what they claim cannot be done. They don't appear to have come to the forum to become members of this community to engage and enhance things but to solely self-promote and sell. Hopefully that will change, and this is just my opinion. How you come through the doors at a party sets how you'll be judged. And that's all I'm going to say on the matter.
YOU NEVER Modify the next line of code just below you.
That line of code could been pre-fetched already.
Any good debugger or simulator should take this pre-fetching in to account.
The robot brain project highlights the need for more memory. One can never have enough memory!
Re using LS and the voltages, I wonder if they could be substituted with HC? And use the memory chips that can run at 3V, eg the ones that Cluso is using? That would just be a matter of specifiying different chips (which would be around the same price).
I look forward to the next article.
"The KISS Debugger(Keep it Simple Stupid) is designed to use the smallest possible portion of Propeller
resources and therefore much of the usual fluff has been omitted"
Later in the article is a screenshot of the used resources (see attachement 1). They use the TV_Terminal
with graphics.spin, no wonder that most of the RAM is used only for the debugger.
11 pages describe this KISS debugger, but it is not clear how to set up a code for debugging. Is this debugger
for Assembly code or Spin code, or both? Or is it only a Memory HexViewer with some commands?
Compared with POD, PASD, Cluso's debugger or ViewPort this KISS seems to be too simple.
The second attachment shows the footprint of PASD, and this debugger has a GUI with Breakpoints, Single Stepping
on source code level, and is free.
POD has an integrated disassembler, and is free.
Cluso's debugger needs no cog ram to debug assembly code, and is free.
ViewPort has a nice PC frontend, can debug Spin code at source level and much more.
All are real time debuggers which run on the propeller, no simulators. All are released long before KISS.
All were ignored by Dr.Jim and Mr.Allred, because they don't study the Propeller scene...
Andy
It had a release date of May 26th if you look at the Robot magazine website under Robot Schedule. It should be in stores now.
Thanks again for your interest.
Mark Allred
Post Edited (mallred) : 5/30/2009 1:58:21 AM GMT
FYI, and despite all the wild theorizing, Dr. Gouge and Mark Allred are not the same person: they're father- and son-in-law. I had my suspicions, based on this page, and found confirmation here. Not that it matters one bit, but at least we can put that speculation to rest (before the National Equirer starts calling) and concentrate on more substantive issues.
Also, of note, some of their prices have come down substantially. This would indicate that they're not oblivous to constructive criticism and were simply unaware of market realities. Let's hope that their integration into the Propeller community at large continues!
Now. Jim and Mark. Next step: ditch the wirewrap (it's so 1960s) and invest in some PCB layout software (or use the free PCB CAD programs that are out there). Printed circuit boards are cheap to have made, even in small quantity (rf: www.4pcb.com) and make circuit assembly so much easier!
-Phil
Post Edited (Phil Pilgrim (PhiPi)) : 5/30/2009 3:40:11 AM GMT
That is a good suggestion re boards. I am using the $125 version of Eagle and that makes a 160x100mm board and OurPCB can have them back to me in about a week for $100 for 10, with fixed costs of about $70 and board costs as low as $3 for multiple quantities. The costs are now low enough for hobbyists. On a good night I can draw up a schematic and a PCB in a couple of evenings. I reckon many others here can do it faster than that.
Soldering PCBs can be quick if all the parts are to hand. Less than half an hour, which would be a lot faster than wire wrapping (even though I like wire wrapping too).
I'd love to brainstorm machine intelligence with Dr Gouge and Mr Allred. I'd done neural network simulations in the past (for bionic ears) and with Z80 technology it was possible (being brutally honest here) to (crudely) replicate the nervous system and behaviour of a slug. With a propeller, maybe it could go up to an Ant. With multiple propellers, ?? could it be possible to get to a Bee.
An ant is still a very complex organism. Just coordinating all those legs is not trivial, and Machine Intelligence seem to be on the right track with the ability to drive lots of servos. But there are other aspects of ant behaviour that are also very interesting, such as the demarcation of tasks in a colony, switching from 'worker' to 'soldier' behaviour, and the way ants lay down pheromones and then ant trails end up taking the shortest path from A to B. Some insect behaviour is learnt, but much is hardwired and insects need to hardwire their programs in efficient ways. A flying insect for instance can devote 30% of its weight to vision and vision processing, and I'll bet that there is not one wasted neuron in the vision system of an insect.
I'll be interested to see where these Robot magazine articles take us.
Post Edited (Dr_Acula (James Moxham)) : 5/30/2009 3:28:05 AM GMT