The collected works of sgifanatic - Page 5

I've recently been impressed with Makulu Linux. Quite complete and simple.

Have used debian often and would recommend it for older, not ancient, systems.

Mint is ok, but I was not floored by it in any measure. Ubuntu w unity is not my cup of tea, but if you can change the window manager and turn off a few services, it is probably still the most polished linux desktop experience.

The spec is terrific for the Pentium era. When the Pentium came out initially, RAM and HDD sizes were much smaller than what you have in your system. Check out the ad below, from 1994:



I think you're going to have a lot of fun with this box. I've really enjoyed my Gateway 486. Read all about it (and see pics) here:

http://www.vintage-computer.com/vcforum ... -of-images )&highlight=gateway+486

I've been struggling to get a Linux or FreeBSD install working on it, but Windows 3.1 and DOS have been running great. I was thinking about trying OS/2 Warp 4, though the voice recognition capabilities require a better spec... will work for you, however:

OS/2 Warp 4 Processor and Memory Requirements
Minimum 486 33MHz, or higher, with 12MB to 16MB memory
For speech navigation, Pentium* 75MHz, or higher, and 4MB additional memory
For speech navigation and dictation, Pentium 100MHz, or higher, with 8MB to 12MB additional memory recommended
Disk space: 100MB to 300MB free

I've been meaning to make myself a Raspberry Pi handheld, but the designs I've seen on the web so far appear a bit impractical to me. I don't think I would actually be able to use something this top heavy/fat:



Or lacking a keyboard:



Or with a tiny screen, a BOM approaching $100, involving too much work:



http://n-o-d-e.net/post/107977286006/how-to-create-a-handheld-linux-terminal

My approach was much cheaper and lazier. I figured I'd eventually find a used/broken portable DVD player with a working screen that could be hollowed out by removing the DVD/CD drive. I would then place a compact keyboard in the cavity. Something like this:



While at a used electronics store, I picked up a pretty good candidate case/LCD combo in the form of a Philips portable DVD player. It was a model PD700, from the "for parts" bin, for a whopping $10. It looked ok, but was missing a power adapter. Here is what a new one looks like:



It didn't have a composite input, but I figured it was worth a shot. Often, the external jack isn't actually soldered on, but these generic DVD player motherboards still have the relevant pads available on the board. If that didn't turn out to be the case, maybe there would be a composite signal somewhere on the board I could get to. Worth an attempt.

So, once back in the garage, I opened it up and found that the motherboard - sadly - didn't have composite out pads. That would have been too easy. Couldn't really trace a usable signal of any sort. The LCD jumper appeared to have R, G and B lines but mixed in with a bunch of fairly complex timing/sync signals. Not something easily tappable. At least for me. Here is what the board looks like:



The chip right next to the connector that interfaces the DVD mechanism just seemed like it was a motor driver. The giveaways were the thicker pins, location on the board... I didn't bother to google it. But the large chip in the center of the board turned out to be interesting; a Sunplus 8202P, which is a RISC based MPEG decoder/all-in-one solution for portable DVD players. Couldn't find the datasheet for the 8202p, but here's the description of the 8202K from the Sunplus website:



And I also found the 8202S pinouts, here:



http://w3.sunplus.com/products/sphe8202k.asp

[For those interested in this chip, it turns out the source code for it is available online. I briefly browsed through it, but this could be potentially valuable for someone who wants to modify the source with custom features. Here's the link:
http://en.pudn.com/downloads386/sourcec ... 24_en.html ]

I figured, ok, the LCD's obviously being driven by the 8202, so if I isolate the pin delivering the signal, maybe I can piggyback my composite out from the Raspberry Pi and then kill the composite out from the 8202. To test this hypothesis I powered up the DVD player... it wouldn't power on. That explained the "for-parts" and $10 price tag. In order to debug, I was messing around with a multi-meter and realized that the system would start up only if the PSU experienced a momentary short-circuit. So, that solved that problem and the DVD player "booted up".

Post boot, I saw the Philips logo and the menu items were working fine! The LCD was actually pretty decent looking so I figured this would make for a nice handheld display.

Next step was to identify some candidate pins... I got the raspberry pi running, and then soldered a couple of probes to the composite out pads on the Pi. I placed these probes on the incredibly narrow and tightly packaged 8202 pins, including the TV_DAC lines and the promisingly named V_COMP line. For a few seconds, I saw the raspberry pi command line appear, very fuzzily, over the Philips DVD menu, in B&W. I got excited and as I was trying the other TV_DAC pins, the probe with the composite signal touched a pin it shouldn't have. I don't know which one. But there's several 3.3V lines in close proximity, as well as bias voltages.

Net-net, after seeing the shell command prompt text appear for a bit, the LCD blanked out and all I could see was a white background with random colored lines. I rebooted the DVD player, tried several times to get it to just show the menu, but no. It was fried. Specifically, "it" probably being the 8202. Damn.

But that wasn't enough, apparently. When I tried connecting the Rpi to another composite monitor, it wouldn't display anything. Yet, I could see the SD card activity LED flash on power-up, so I knew it was working. I tried an HDMI to VGA adapter and, sure enough, the HDMI out still worked. Apparently the Composite out on the Pi had also died.

What I discovered next was probably the worst part of the screw-up. The USB ports on the Pi wouldn't work either. A cryptic power message appears on boot, followed by enumeration errors. I tried a powered hub, wired keyboards of a variety of types, a USB Wifi adapter... nothing works.



What a chain of failures! So I figured, ok, the board still boots, it has an HDMI display. If USB is fried, maybe I can connect to Ethernet (wired) and SSH into it. But no. Ethernet is dead too.

So, the conclusion of this sad tale is that I am back where I started. Two hours, a cheap DVD player and a Raspberry Pi were the casualties in this - on hindsight, perhaps ill-advised - experiment. That said, the LCD itself might still work if I can find a driver board. And the plastic housing of the DVD player is a pretty good option for whenever I regain the courage to try this again.

ivelegacy wrote: I haven't understood what is your will, what do you want to obtain ? a PDA ?

A Raspberry Pi based handheld running Linux.

Why don't you buy a japanese one instead of building one ?

Because it's fun. Since you are participating on a forum where folks derive their enjoyment from resurrecting 25+ year old hardware and porting software to these systems, I am astonished you would expect much in the way of build-vs-buy pragmatism

guardian452 wrote: Maybe try one of the smallish 8-10" TFT monitors, many of which have USB or serial touchscreens. You would have to bolt on a beefy enough battery as most of them aren't exactly designed for energy efficiency, but in keeping with the DIY spirit I expect you would want it to be fairly chunky anyways. Otherwise people will just think it is an ipad. You could 3d-print a thicker back panel to hide the pi and battery, and design some sort of attachment to a keyboard.

Maybe something like this would be a better starting point?

edit: don't know why this URL isn't showing up so I will put it in code blocks...

Code: Select all

http://www.amazon.com/Tontec%C2%AE-Resolution-Raspberry-Rearview-Headrest/dp/B00OOJPAGW/ref=sr_1_1?s=electronics&ie=UTF8&qid=1437945195&sr=1-1&keywords=tft+lcd+hdmi

Thanks for the link, and that's a good idea. There are actually a lot of alternate LCD choices, meant primarily as backup camera displays for cars, that have a composite in. I was going for the DVD player simply because the casing is already perfectly adapted for the LCD it houses. That said, LCD driver boards with a proper composite or hdmi in are a much safer bet. I'll just have to engineer a case from scratch...

pentium wrote: I'm quite baffled how the heck you managed to blow so many components.

I'm equally baffled. It literally happened in under a second. Obviously the probe slipped and touched some other pin. But there were no physical signs of damage on either board. I initially thought I'd just got the system in a weird, unstable state. But after many reboots on both systems, it turned out to be otherwise.

ivelegacy wrote:
Apple Newton's design

I like how the Newton looks, but I am trying to get a keyboard in there. I really took to the blackberry physical keyboards back when blackberry was still a thing...

spiroyster wrote:

sgifanatic wrote: though the voice recognition capabilities require a better spec

For dictation software on the 586 (I think a 486 might struggle)... I used to use something called 'Dragon Dictate' for speech recognition on windows98 (it was on the K6/233 I had with 32Mb RAM!) with no problems at all.. you had to train it by talking to it which I can only assume built up a better phonetic signature for the program to understand... it did get a few words wrong but I do remember it working quite well overall... one had to speak in ones proper english like .... 'goto sleep'

Very familiar with Dragon. Pre-deep-learning neural networks exhibit a restricted ability to learn with the network getting stuck in local error minima during training; this is what caused 90s era ANNs (Artificial Neural Networks3) to never really achieve a high-enough recognition rate. Alternative techniques such as HMMs and GMMs don't rival the accuracies achieved in current DNN (Deep Neural Network) systems... MSFT, Google and FB all use Deep Learning extensively now.

This passage from a Nuance publication specifically addresses how DNNs address the shortcomings of older techniques.

"Another poignant example of an innovation which is not only “hidden” but also based on a rather old concept is Deep Belief Networks which have revolutionized the architecture of ASR (and related systems) over the last few years; This topic was highlighted at the conference by a number of talks and posters on everything from acoustic and language modeling to text-to-speech (TTS) and voice biometrics. As the more recent term “Deep Neural Networks” already indicates, this is the renaissance of the Neural Networks (NNs) of the 1990s. Back then these networks were hugely appealing to AI researchers as they, modeled after the human brain, seemed to have a superior explanatory and modeling power over the Hidden Markov Models/ Gaussian Mixture Models (HMMs/GMMs) combination, which had dominated acoustic modeling in ASR since long before. But yet again it turned out that when applied to “real problems” training methods for NNs were either vulnerable to get stuck in local minima or computationally too consuming to be of practical value. And so HMMs/GMMs continued to rule, bringing error rates down and accuracy up by piling many smaller optimization and (yes) innovations within this framework. Again, the combination of several factors – modern, more powerful hardware including GPUs (chips originally developed for graphical processing, think: computer games) and new ideas about how to train NNs – changed this picture radically over the last few years. Once researchers knew how to make them work, “deep” NNs helped researchers to bring error rates down by 25% or even more with just this one [at its core, rather old] innovation. After this first revolutionary wave swept over ASR and its relatives of TTS and voice biometrics a couple of years ago it was followed again by a phase in which many smaller innovations saw the day of light day, now within the framework of DNNs."

http://whatsnext.nuance.com/in-the-labs ... ence-2014/

GIJoe wrote: well, blackberry makes the 'passport' which still comes with a keyboard. of course it is entirely up to debate if blackberry is still 'a thing'.

I've had an interesting and somewhat complex history with BlackBerry, which goes considerably beyond being a user of their products. Based on those experiences, I can certainly attest to the fact that they truly oozed thingness at one time. Sadly, and I mean this when I say it, not so much anymore. For those interested in BB, I recently got copies of two books, written 5 years apart, that discuss RIM's evolution and eventual tribulations. The books are, "BlackBerry: The Inside Story of Research in Motion" and "Losing the signal' - I have not read them yet, but leafing through them, they appear quite interesting.

GIJoe wrote: i was fond of these for quite a while, too but nowadays i find touchscreen-based ones to work better. no crammed together buttons anymore and when you don't need a keyboard, it just conveniently fades out of the way.

True. I would do a touchscreen, but resistive touchscreens are truly miserable to use, and capacitive overlays are quite pricey. If you remember a pre-netbook handheld computer company called oqo, and their rival, flipstart, that's kind of what I am going for. The flipstart more so, to be specific. Here's what it looks like:

hamei wrote: Maybe take a look at the Neo900 ?

Thank you for that info. I wasn't aware of this project. Glancing briefly through the specs/design, it certainly looks interesting.

And I did look at it in more detail. Interesting, but 1100-1200 euros for a 1ghz/1gb system with a small screen doesn't quite fit with what I had in mind.

The Flipstart form factor, large screen and 1ghz x quad core performance at an affordable price point, i.e. 80-90% lower than the neo900, running Linux, is kind of what I am shooting for. But mine is just a hobby project. It certainly won't have the fit and finish of the neo900.

Welcome! That's a great system to have fun with. Do you have the teal feet for your Indigo2?

A few quick pictures of the cardboard prototype. Plenty of ways to make this thinner, but for now all the electronics, voltage convertors, battery pack and LCD etc are all working well.

I'm in central Texas and have a teal indigo2 I can part with. If you're interested, please pm me and we'll work something out.

Simplified internal cabling and tried a simple plexi case:

Put together w glue for now. Will add side panels and replace glue and dowel spacers with bolts... Didn't have the correct length. Probably paint the plexi also.

For work, or have you done something naughty for which this is punishment?

With Win10 here, the new Skylake processors hitting the shelves along with the Z170 chipset and 1TB SSDs approaching $300 , it's not a bad time to buy a new system or upgrade your existing one.

I don't know if you guys have seen this. Cringeworthy, but memorable.

We've all read about how Steve Jobs ran people into the ground, or how Musk can crack the whip and assault employees mercilessly, so this is neither unexpected nor particularly "new". Nonetheless, reading this NYT piece on Amazon and its culture still rattled me for some reason... we've got to figure out a better way.

http://www.nytimes.com/2015/08/16/techn ... place.html

I did go back and watch a few more episodes. Still hate the drama. Still hate the burning of shipments - actually, the mindlessness of *most* actions by the main characters. It just makes the whole thing stupid and unreal. That said, I am in it to watch the damn computer being built

Have not seen Mr. Robot. Will check it out.

A few changes:

Fears about AI fall into two big buckets:

1) AI will kill us
2) AI will take our jobs

There is a third, somewhat subconscious fear, i.e. if machines out think us, we will no longer be "special".

Re the first one, there is enough autonomy coming to weapons where a very small number of people might be able to deploy and propel a large force of powerful autonomous weapons systems to create havoc at an unprecedented scale. Whether or not the initial intent is supplied by a goal setting AI or a human, at some level, becomes immaterial. Robots as warfighters are going to be a reality. CIWS cannons, autonomous drones, fire and forget missiles and unmanned ground vehicles are just a few examples of weapon systems that make decisions on their own. And the thing about not firing without human approval is a bit of a myth... they are instructed to take on a range of targets, but AI-powered systems do individual target selection and firing on their own today.

Re the second, yes, AI-based systems are taking our jobs and will continue to. Whether its autonomous cars decimating the number one source of employment for US men (truck drivers) or reasonably advanced "agents" that may not entirely obliterate, but will certainly greatly reduce, employment in the number one job category for women in the US (assistants). Other than these, mass providers of employment, such as manufacturing, agriculture etc. are clearly going to be robo-sourced. Yes, not every job will be done by robots/AI in the near future, but this doesn't matter. Enough will be done so that some pretty fundamental socio-economic assumptions will have to be re-examined.

Re the third, everyone is special in their own way

AI is here and its footprint will continue to grow. Whether robots are "sentient" or have "consciousness" are almost peripheral questions when considering the impact of AI with regards to all three of the fears cited above.

hamei wrote:

sgifanatic wrote: AI is here

Well, not really. That's not intelligence. That's "following instructions". You can set up a row of dominoes, then push the first one so they all fall down in a row. Is that "intelligence" ?

The question you are asking above appears to be rhetorical. I don't know if you are familiar with layer at a time learning, for example, but when exposed to images the system begins to learn and abstract the same kinds of basic features humans would, without being told to do so, or given any knowledge of what a line is or an edge is. There is no first-order "if-then-else" involved. AI systems today are performing their own feature selection, feature evolution and so on. The abstractions they learn - and the rules they evolve - are not provided by a human. In most cases, humans are unaware of them.

The theory of relativity and the sistine chapel were not 'following instructions'. That's intelligence.

Chess was the ultimate exemplar of human intelligence until Deep Blue.

Computers have invented important things like super efficient 3D junctions, without being explicitly told how to do this. These results were achieved by programs that could "re-write" their own rules and discover new rules not considered by their human creators... "advancing knowledge". The 3D junction I alluded to is just one example of an unanticipated discovery made by AI software that did not result only from following explicit instructions provided by a human. There are many other such examples.

Very early on, AI software was already capable of chancing upon important theorems and mathematical discoveries. For example, an AI program re-discovered Goldbach's conjecture. And within the domain of mathematics, automated theorem provers have been providing increasingly more impressive results. Their discoveries will become better known in time. A likely outcome within the next century is that much of the new mathematics being discovered, will in fact be machine discovered. This doesn't require machine sentience, by the way. Just faster computers and the same types of GA/GP techniques that have proven themselves over and over in this category.

It's interesting that you used the theory of relativity as an example of human intelligence. Particularly in physics, many human discoveries have been attributed to symbolic manipulation; i.e. not "seeing the truth" in some structural sense, but arriving at it by following a mathematical chain, constrained by the pre-defined rules of mathematical operations. When human mathematical symbolic manipulation leads to a new "truth" it is accepted universally as evidence of intelligence.

Whether robots are "sentient" or have "consciousness" are almost peripheral questions when considering the impact of AI with regards to all three of the fears cited above.

Dancing on the head of a pin again. The question itself is irrelevant.

I agree. As I said, both these questions are peripheral when considering the impact of AI. AI is having a massive impact already and the debate over whether software can be sentient is thus becoming less interesting.

Not for too much longer. When there's no food in another fifty years,

I'll start stocking up!

commodorejohn wrote: But, but, but hamei, it's The Future now! Or if it isn't, we can make it be The Future by just insisting hard enough! What are you, some kind of filthy Luddite , that you don't believe in The Future!?

I can only speak for myself. And while I am sure some folks may limit their interest in this space to rhetorical insistence, I do, in fact, do a little more than that

hamei wrote:

sgifanatic wrote: Chess was the ultimate exemplar of human intelligence until Deep Blue.

Where did you get this idea ? Chess is problem-solving but it's not intelligence. Never has been.

I got this idea from human history. You may have always known that Chess was merely algorithmic problem solving, but most of humanity did not. Quotes regarding adroit chess-playing requiring intelligence are legion.

A modern quote: Chess is a unique cognitive nexus, a place where art and science come together in the human mind and are then refined and improved by experience. - Kasparov

An ancient quote: Al- Mas‘ūdī writes likewise that the king "Balhit", who is said to have codified the game of chess, gave it preference over nerd, a game of chance, because in the former intelligence always has the upper hand over ignorance.

This stuff you refer to as "intelligence" is merely "being really good at solving predetermined puzzles."

It ain't the same.

Until it is. This is the history of AI. The moment an undertaking previously needing human intelligence is solved by a computer, that problem is redefined as something that never needed intelligence to begin with. Until recently, driving a car and pattern recognition were considered beyond the real of algorithmic problem solving. Look again.

You miss the point. The intelligent part was not calculating the math. The intelligent part was wondering why the universe is made the way it is.

Actually, no, I didn't miss the point. Algorithms can have a sense of wonderment not unlike our own. The current issue is scope. When we are faced with something that is unknown, we question it and use cognitive tools - more recently, the scientific method - to explore. The tree of all possible knowledge is indeed vast, and wonderment is a sometimes inexplicable desire to climb up the tree. Man may be further up its limbs in many ways, but the exploration of this tree is not intrinsically exclusive to man or biological intelligence. Much consideration has been given to how we set "goals" for ourselves, and there are interesting theories regarding the "one goal to rule them all"'; the core goal that gives life to all our "drive" and "curiosity". All of this may not be as magical as it seems. And dismissing the possibility of increasingly intelligent AI systems that can explore more and more of the tree of knowledge, is, in my opinion, not particularly intelligent.

Sorry. Stepford wives ain't never gonna cut the mustard.

I don't get the reference. But let me ask Google if it is relevant to our discussion in any way.

Probably the largest collection of these systems anywhere?

http://www.ebay.com/itm/322483980652

Following up on the two recent BeBox auctions where 66Mhz systems were sold between $1000-1300, now NeXTStations also appear to be increasing in value... working systems listed at $699, $799 and $1200 at the moment. And the $800 unit doesn't even boot into an OS.

$1200
http://www.ebay.com/itm/Next-NeXTstatio ... SwdjNZAqrV

$800
http://www.ebay.com/itm/VINTAGE-1992-NE ... Sw0hlZN0ef

$700
http://www.ebay.com/itm/Museum-Quality- ... Sw~y9ZOAx7

robespierre wrote: (You should also do readers a favor and remove the tracking tags from your links.)

Not sure what you mean.

What "tracking links" and what favor would I be doing them?

jan-jaap wrote: Those systems are all gone now. The rest is a matter of supply and demand I guess.

Exactly! This is a broad trend. Apple II prices have gone up tremendously, as have BeBox prices... NeXTStations and Indys, as you say, are heading in the same direction.

Dodoid wrote: Hoard your SGIs I suppose!

bifo wrote: those prices aren't aimed at hobbyists like us, they're aimed at people with corporate budgets who need to replace dying legacy hardware which can't be upgraded (at least in theory). its the same reason you see $200+ prices on 20mb

That may be the case for SGIs, which were widely used in commercial settings, but harder to believe for BeBoxes, Apple IIs and NeXTStations.

necron2600 wrote: Prices are aimed at anyone who will buy.. including hobbyists. I don't have a BeBox yet.. what are my choices if I did want one? Only today's high prices. Supply and Demand as jan-jaap said.

Spot on. And for the historical significance of the BeBox, combined with the near-zero supply and low production numbers, they should really go for even higher than they do.