Slashdot Mirror

"Random processes"? Any randomly assembled amino acid randomly disassembles as well; even Miller proved that.

The randomly assembled amino acid does randomly disassemble as well, but that is not what it must do. An amino acid may stay the same, disassemble, or it it may form a more complex molecule.

Here is a little demonstration of "randomly" assembling complexity in behavior. I have given each entity the ability to sense the left and rightness and ahead and behindness of 'energy' dots and their nearest peer. They also get a sense of their relative energy vs their peer. The inputs can affect two thrusters which operate like "tank treads". However, their minds are blank. They don't know what to do with the inputs or how they map to the outputs. The genetic program introduces random errors as copies runs of a genome from one parent then the other switching back and forth randomly. The selection pressure simply favors those with the most energy at the end of each generation by granting a higher chance to breed. Use the up/dn keys to change the sim speed, and click the entities to see a visualization of their simple neural network. The top left two neurons sense nearest food distance, the right two sense nearest entity, middle top is the relative energy difference of nearest peer. Note that randomness is constantly introduced, and yet their behaviors do not revert to randomness or inaction, they converge on a better solution for finding energy in their environment.

There is no pre-programed strategy for survival. Mutations occur randomly, and they are selected against, just as in nature. Given the same starting point In different runs / populations different behaviors for survival will emerge. Some may start spinning and steering incrementally towards the food, others may steer more efficiently after first just moving in a straighter path to cover the most ground (they have no visual or movement penalty for backwards, so backwards movement is 50% likely). As their n.net complexity grows their behaviors will change. Movement will tend towards more efficient methods. Some populations may become more careful instead of faster, some employ a hybrid approach by racing forwards then reversing and steering carefully after the energy/food is passed. Some entities will emerge avoidance of each other to conserve energy. Some populations will bump into each other to share energy among like minded (genetically similar) peers. Some will even switch between these strategies depending on their own energy level.

Where do all these complex behaviors come from? I didn't program them, I didn't even program in that more complex behaviors should be more favorable than less complex, and yet they emerged naturally as a product of the environment due to selection pressure upon it. Just because I can set the axon weights manually and program a behavior favorable for n.nets to solve the problem, doesn't mean randomness can't yield solutions as well. Today we can watch evolution happen right on a computer, or in the laboratory. All of this complexity came from a simple simulation of 32 neurons arranged in a simple single hidden layer neural net, with 5 simple scalar sensors and the minimal 2 movement outputs, with a simple single selection pressure. Each time you run the sim it produces different results, but all meeting the same ends, collect energy, reproduce. Just imagine what nature can do with its far more complex simulation and selection pressures... You don't have to imagine, you can look around and see for yourself.

In other more complex simulations I allow the structure of the n.nets and form of sensors to be randomly introduced and selection pressure applied. In larger simulations I allow the breeding and death of generations to occur continuously across wider areas and speciation will occur. Entities will develop specialized adaptations for a given problem space of the environment. I have created simulatio

I write code in Machine Code with a bootable hex editor (446 bytes, fits in a HDD boot sector). It's the easiest way to bootstrap an OS from scratch now that MoBos don't have boot from serial port anymore...
Here, run it in a VM: "qemu-system-i386 hexboot.img ", if you want.
Or, "dd if=hexboot.img of=/dev/sda bs=1 count=446 conv=notrunc", if you want to preserve the partition table on a bootable drive.
Arrows,PgUp,PgDn,Home,End = navigate; Tab = ASCII/Hex, Esc = jump to segment under cursor, F8 = Run code at the cursor. (this is a real-mode version)
When it boots you'll be looking at the code that booted, there's only two variables that didn't fit into the registers, you can see them changing at the bottom of the code as you stroll around.

That's all you need to create an OS, complier, etc. from scratch. You'll probably destroy your system though if you're not careful, so keep in in the VM if you're a noob; Lock-up is a mild danger, but corrupting the CMOS, etc. can leave your system bricked. You can replace the BIOS too if you know what you're doing. Maybe some day I'll publish a path to go from zero to OS while avoiding the Ken Thompson Compiler Hack... Folks are only just beginning to get interested in having actual system security, so maybe we'll lick the problem some other way. There's still chip microcode to worry about, but programmable hardware may allow us to route that exploit vector out too some day.

Screw your bullshit optimized compiler crap. It's stupid and far slower than you think, esp. since the binaries are bigger (1 cache miss and I've already beaten you in most cases). Besides, Next year or so the system will run twice as fast. My need for speed is tempered by my greater need for security and readable machine code. If I identify a patch of code that needs to be optimized or vectorized, I can do it myself.

Premature optimization is the root of all evil.
- Donald Knuth

I don't care about my lawn, it's just there to keep the dirt intact.

Were you being sarcastic? I have compiled thousands of pieces of code in the last 30 years. None of them have magically transformed into anything other than what I compiled. AI is not voodoo, magic, or anything else.

It's not magic. Neither is cognition. Your big ass-brain is highly inefficient, it's a poor standard to gauge others' sentience against. Did you know the machines are exploring Mars all by themselves now? Curiosity has a machine learning system, for navigation, among other things.

It only takes a few cyberneticists being a bit disenchanted with humanity's forty years of failure to realize the spark of life must spread to the galaxy by another means... I'm getting ahead of myself. It only takes one learning program and a super computer's worth of power and a bit of time to create a learning machine system as complex as your mind is.

Check out my little AI children. (up/down arrow to change sim speed). Click one and you can see the neurons firing. Aren't they cute? It takes about 300 lines of code (mostly boilerplate and environment sim) to create programs (plural) that can learn (there are 20 here, learning). It really only takes 4 neurons to get them to collect dots. However, I added a hidden layer and some extra input about their neighbors energy status and location. Neurons Left to Right: [leftness of food], [forwardness of food], [other's energy - my energy], [leftness of other AI], [rightness of other AI]. There are enough neurons in the hidden layer to allow each input to be considered against all the other inputs. The outputs work like tank treads, or thrusters in space, sans inertia. Their "eye" neurons are like simple directional antennae, with only two neurons required to pick up a full 360 direction AND distance due to fall-off (inverse square of distance law).

This environment applies natural selection to the brains. The only selection criteria is those that have more energy get chosen to breed more often. This results in various strategies for movement in different runs of the sim: slow, fast, forward, backwards, spiraling, aiming just past the target, then stopping and reversing into the target. Different social behaviors: Bumping to share energy among a group of possibly like minded individuals, or avoiding each-other to save energy, sometimes switching between the strategies depending on the neighbor's energy level vs one's own... Their brains start blank, and in only a few generations movement is emerged via selection. Steering towards dots comes next, then avoidance or collision, Usually a hundred or so generations the social status becomes a factor to compete via.

Such variation from so minimal input. Intelligence is an emergent property of complexity, you see. Tailor the complexity such that the information is self reflective, and self improving and you get intelligence. Instincts are basic intelligence encoded in genes, expressed as brain structure (firmware), culture is your software, and evolves much faster. Unfettered from a life cycle of years natural selection can be very powerful, with a bit of guidance it could blow your mind...

So, Just create a problem space, and goal. Connect a few dozen neurons, and without any guided training a good solution can be arrived at given a bit of time. This is how a machine learning system could come up with ideas and solutions. Consider the sim not many smaller AIs but one AI made of 320 neurons solving the problem of most efficiently collecting dots via swarm of bodies.

Each brain is 32 neurons, there are 8bits worth of strengths (weights) for each neuron, so 256 bits in the genome (though note: I could make them evolve to move towards dots with only 32bits in their heads). Machine intelligence is efficient. It can do far more with much less. The barrier for sentience is far lower than you think.

Your brain is 100 billion neurons, but is VERY inefficient, and mostly not concerned

There is no test for obviousness, despite it being a requirement that patents must not cover obvious innovations.

There's a difference between "there is no test for obviousness" and "I don't know what the test for obviousness is," and they should not be used interchangeably when you mean the latter. The Examiner can make a prima facie case that a claimed invention is obvious by showing that one or more pieces of prior art, alone or in combination, teach or suggest each and every element in the claim. So, if a claim recites A+B+C and the Examiner can show one piece of prior art that teaches A+B and another that teaches C, and they can be combined without undue experimentation or changing their functions, then the claim is obvious. If, on the other hand, the Examiner can't find anything that teaches C anywhere, then the claim is not obvious.

Happy to help clarify that for you.

Now, to clear up a few other things you say:

The cost to invalidate a patent is in the millions, but the cost to file and receive a patent is in the hundreds.

It costs about $25k to file and prosecute a patent to issue, not "hundreds". And they can be invalidated for just a few hundred via post grant review.

First to File means that patent application secrecy is needless -- If you're filling a patent it should be public knowledge so we can protest anything that is obvious / file our prior art.

That doesn't change anything, as all patent applications were published under the first-to-invent system, too, and the public could protest by submitting prior art to the Examiner.

Look at the Automotive and Fashion industries -- Neither of which have design patents or copyright, and yet design is their core sales point, and they are very innovative in design.

Here are tens of thousands of design patents in the automotive industry. And here are thousands of design patents on dresses. A search for design patents on "jeans" will turn up a similar number, as will socks, shoes, etc. Your premise is demonstrably false.

Anyone who fears Trade Secrecy / Trade Unions locking up innovation has never met a reverse engineer from this century. We strip layers off of microchips to discover their wiring at the nano scale.

Yes, but without patents, there would be more efforts to protect trade secrets - like you wouldn't actually ever own a physical microchip to strip it apart. You'd merely be renting it under a lease agreement, with massive liquidated damages if you do damage it. Do you really want to have to have a subscription to use your computer? I don't.

We have spectrograms and readily derive the secret recipes for foodstuffs and medicines -- Hence the generics market existing.

Actually, the generics market exists because of patent disclosures and the FDA orange book. Would you buy a "generic" reverse engineered version of a brand new medicine that the FDA hadn't approved, much less tested?

It's not economically feasible for secrets to exist in consumer products today, you can't hide the molecules and machine code from me (and yes, I do read and write fluently in machine code).

Watch out, Slashdot, we've got a badass over here.

Everyone seems to love to jump on the "The patent system is at fault here its terrible, it lets this stuff happen" when in reality there is nothing wrong with the patent system at all.

There is no test for obviousness, despite it being a requirement that patents must not cover obvious innovations. The PTO grants a massive amount of patents that should be invalid, as proven time again in the courts. The cost to invalidate a patent is in the millions, but the cost to file and receive a patent is in the hundreds. They don't take responsibility and readily put forth the fact the standard for granting a patent is far below the stated requirements, so the courts are weary to just invalidate a patent -- believing that the patent office is actually doing it's job well, when in reality they'll let you patent essentially anything: If you want a weak patent that won't hold up in court, they're cool with that so long as you pay your fees, cross your t's, dot your i's, and replace all the terms such that the application won't generate any hits when searched in their patent database. First to File means that patent application secrecy is needless -- If you're filling a patent it should be public knowledge so we can protest anything that is obvious / file our prior art. Yes, it would take longer and be more expensive to grant patents, but just make those filing for the patents pay the bills; This would bring the cost of patents up closer to their actual cost -- No, instead it's basically a government subsidy for legal warheads.

Furthermore: Patents are not required to spurn innovation. Demand for innovation exists regardless of patents, and will generate drive for innovation by basic market forces regardless of granting monopolies. Look at the Automotive and Fashion industries -- Neither of which have design patents or copyright, and yet design is their core sales point, and they are very innovative in design. Patents and Copyright merely enforce artificial scarcity, when the artists and researchers could just as easily make their money without patents: Simply withhold their labor unless agreement to be payed for it is ensured (like car mechanics and home builders do, it's a proven system). If there are those who would not invest without monopoly assurances afterwards then those who will invest in research regardless will out compete them -- As proven by stagnant companies doing poorly in the automotive and fashion design industries compared to those who innovate without promise of monopoly. Anyone who fears Trade Secrecy / Trade Unions locking up innovation has never met a reverse engineer from this century. We strip layers off of microchips to discover their wiring at the nano scale. We have spectrograms and readily derive the secret recipes for foodstuffs and medicines -- Hence the generics market existing. It's not economically feasible for secrets to exist in consumer products today, you can't hide the molecules and machine code from me (and yes, I do read and write fluently in machine code).

We have NO EVIDENCE that the patent system is fulfilling its goals. We don't know if it's harmful or not, but there are indications that point to it being harmful (frivolous patent lawsuits and bogus patents), and there are at least two data points that indicate patents are not required at all. The real problem with the patent system is that we did not do a test to see if it was beneficial. Everyone just assumes it is without ANY evidence to support their claims. We need to do the experiment and abolish patents to see if they were harmful or helpful. We can re-institute the laws if we need to later. Continuing to operate the world's economy on assumptions of untested hypothesis is egregiously intellectually and economically negligent. That patents exist at all with no proof they're beneficial is the problem. That problem can not be fixed until we've abolished patents.

If you assert that there are no problems with the patent system then I must point out: Ignorant people like you, who operate based merely on assumptions without any evidence to back their claims, are the problem.

"Are auto-complete results even useful?"

Well, let's ask Google.
Autocomplete is:
Autocomplete is not a function
Autocomplete is not working
Autocomplete is not working in outlook 2010
...

I'm afraid I have to agree with Google on this one.

::yawn:: Whilst my coffee is brewing I've made you this screenshot of said search, because "pics or it didn't happen". I'm a Northern Yank transplanted to Houston, but I like British comedy, so maybe that's what it's picking up in my case.

Hmm. Assembly you say? "We need to go deeper..." I'll just leave this here.

FTRMF:

WARNING

This software is ammunition for foot snipers. You will be editing the system's memory matrix directly, in real time, as it is running. It is strongly suggested you first use an artificial construct such as a Virtual Machine to familiarize yourself with using Hexabootable.

If you edit a program as it is running a hung CPU is the most likely, but not the worst thing that can happen by far; Editing a working stack is just as dangerous. Your firmware and/or hardware could be seriously damaged if you are not very careful in there...

The first page that appears (address 07C0:0000) contains the editor program that is displaying the text. Although some memory may be seen changing as the view and cursor move near the end of the program, you must resist the urge to edit this live machine code (unless you're prepared to face the consequences).

See the Memory Map for your system, and also this program's memory map which is listed in the source code along with many other details.

One of the other commenters quoted Linus, "Talk is cheap, Show me the Code". I have that on my coffee mug right now :-)

Unfortunately most of my work is in machine code and although it works for me, it isn't well tested since that's not my main concern just yet. When I get to the point of having more stable code in a more readable form than raw op-codes or half-implemented ASM, then I'll be sure to release it. Until then, here's the raw memory bootable hex editor I mentioned: Hexabootable. This one has a couple more features and a Visual interface rather than my first EDLINE like input method; Although it fits in a 512 byte boot sector there's no room for a drive partition table (meh, those are really more of a reccomendation, not an essential rule). You should still be able to boot it from a drive via:
dd if=hexboot.img of=$YOUR_DRIVE
or using a VM (which you should use instead if you value your hardware).

I went through and heavily commented EVERYTHING, so even folks who don't know x86 assembler can follow along. There's 26 bytes left of space in the image, so when I have time I'll see about squeezing in the text input mode I've implemented in machine code on a live instance; Till then it's Hexadecimal all the way.

Happy Hacking!

On yesterday's PCs, I could just write raw machine code in Hex, save it to the 1st sector of a drive, boot the disk and be in full control of my own hardware with my own code. Many new-ish PCs now use EFI. To boot from EFI I have to write my machine code within a FAT (32) container, which means implementing MS's proprietary and patent encumbered File Allocation Table format... Tomorrow's PCs will use UEFI to boot, which requires a cryptographically signed EFI boot process. That means signing my own bootloader and installing my own keys, or paying for a key for each bootable from MS (some UEFI systems allow booting w/o signature via special boot mode, some do not) -- On ARM platforms shipping Win RT, MS has said the option to boot unsigned code or install user specified keys must be removed.

So, you can see how it's slowly gotten a bit harder to play with my own new hardware thanks to the increasingly high hoops I've got to jump through. If Microsoft has their way you won't be able to boot any OS that doesn't fork over the cash to them. In fact, even the Linux Foundation is planning to pay MS for the right to sign a bootloader so you can still boot your own software on UEFI hardware. I think that's horrible. I understand they want to make it easy for users to run free software but IMO, paying MS one red cent to give us back the freedom to use our own software with our own hardware is just vile and disgusting. Instead, I'll buy from vendors that respect my freedom. The subject line say MS + Secure Boot == PC Death, but really Apple, and many other vendors who don't let us unlock our devices to run arbitrary code are equally as evil in my book.

Recently a longing for the good ol' days of unfettered computing led me to creating Hexabootable. It's a 512 byte boot sector that contains a Hex editor. With it you can edit raw memory then execute the memory you just edited. Using only this minimal tool you can extend the program's features (eg: disk I/O), write any other program, even create a whole new Operating System -- Indeed that's exactly what I'm doing.

None of my hardware or software hacking hobbies will be possible if the OEMs get their way and lock us out of our own hardware. It's all under the guise of Security, but that's not really the reason. Think about it: OS code is huge and bug ridden; If there's even one kernel level arbitrary code execution vulnerability then the whole effort is useless. If the OS makers could write secure (read: bug free) OS's they would be just as secure with and without secure boot! If they can't write secure OSs then secure boot is pointless! Truly, I can use known exploit vectors against every modern OS, secure boot or not, to run my own unsigned machine code, and so can malware writers... So it's not a boot for normal end user security, it's just digital shackles. The real reason Secure Boot Chains exists is to keep you from tampering with your own computer.

Now, what I do find hopeful is the cool work in the embedded systems fields. There are several projects that strive to be as transparent to the user as possible, and get their code up and running controlling everything. Unfortunately you don't always get to run plain machine code on all of the hobbyist devices. Open hardware initiatives give me a warm fuzzy feeling -- That's what will save the "PC" (Personal Computer) in my opinion. Protip: If you can't personalize the machine code and/or hardware, then it's really an Impersonal Computer -- An impostor of the worst kind...

Here's a fun aside: Since I write software in machine code, I could release it under the GPL and provide no other "source code" but the binaries :-P
Conversely, if you know Machine Code, every (non encrypted) binary executable is Open Source!

Once upon a cubicle, our office was relocated. Just after the move, IT hadn't installed anything on our computers, and thus Internet access was down too. A coworker mentioned that one, "couldn't, do anything with just a fresh install of Windows!". It would be 4 days before the understaffed IT could get to our systems. In the meantime I made a Snackman game to prove a point, but also to kill time. I added the music and sound effects later, of course.

I haven't finished my the public facing blog yet, but you can also play a Tetris game I made using only the URL bar on my website.
What a nerd will do to win a bet... Yeah, I've posted it a few times before, now you can edit it and see the changes though.

So, I've been down this road. It's doable, but WTF man, don't you have a smartphone? Just use it. That, or fucking quit, that place sounds awful.

P.S. That little paragraph about any creations belonging to the company? I just quietly draw a line through it and initial it. If they ask questions I say: "That's my only hobby. Would you give up TV, Videogames, Movies, Golf, etc. for your job? If you want me to do so, then I'll need a hell of a lot more money."