Wednesday, September 16, 2015

No Way Out


I am torn between two options: releasing the results of my speech learning research to the public or erasing everything that I have worked on these many years. The world is in such a precarious state that I cannot think of any scenario in which the introduction of artificial intelligence will not do great harm to humanity. The problem with AI is not that intelligent machines might rebel against us (that prediction is just stupid materialist superstition), but that the powers that be will use the technology to gain even more power and impose their twisted will on humanity. The world could quickly turn into a hell much worse than anything George Orwell could have imagined. I would rather die than live with the knowledge that I might have contributed to it.


Unknown said...

By releasing it into the public domain the world and the people have a better chance to use it for good. Otherwise, someone else will anyway (sooner or later) arrive at the same results as you, probably working within the private or goverment sector.

Louis Savain said...

"By releasing it into the public domain the world and the people have a better chance to use it for good."

There is no doubt it can be used for great things that we can't even begin to imagine. But there is a lot at stake. If it falls into the wrong hands and they cease the opportunity to use it to gain power, we're done for. History has not shown humans to be kind to one another.

"Otherwise, someone else will anyway (sooner or later) arrive at the same results as you, probably working within the private or goverment sector."

Maybe but I doubt it. Knowing what I know about intelligence and the brain, I don't see AI coming from either governments or the private sector. The big deep learning craze is a major will turn out too be a distraction, one that is forcing their best minds to pay attention to something that has little to do with intelligence.

Louis Savain said...

Sorry, let me rephrase that last sentence:

The big deep learning craze will turn out to have been a major distraction, one that is forcing their best minds to pay attention to something that has little to do with intelligence.

Bill said...

All of the great scientists and inventors have struggled with this issue.

I suspect that if you don't go forward with your AI project, someone else will invent it:

There are enough smart people working on adjacent ideas that someone is bound to stumble onto it.

You should just accept your destiny as a visionary and claim priority for your ideas.

Louis Savain said...

Hi Bill,

I agree that there are many smart people working on the problem. But I am afraid it will take more than smarts to crack this nut anytime soon. The only person I know of who is on the right track is Jeff Hawkins and even he is still far from the goal. Hawkins is now stuck in a rut of his own making. The biggest flaw with his hierarchical temporal memory (HTM) is that it does not do unsupervised pattern learning correctly. Otherwise, it would have kicked everything else out of the water. And pattern learning is the easiest of the big problems of intelligence, IMO. The deep learning crowd, on the other hand, is not even in the starting gate.

I am not a visionary. I was lucky enough to stumble on an important source of knowledge that others are blind to and have rejected. Lately, I have come to realize that this AI thing is not mine to do as I please. My paranoia about its being used for bad things has only grown in the last year. The human species cannot be trusted with this knowledge because it can only lead to full self destruction. I cannot even trust myself. After all, I, too, am human.

Chris said...

Honestly, I really want you to do it, if only so that I can finally be made aware of what you have been working on these past years. The curiosity kills me, I just want to see if it works. I have to agree with Unknown and Bill above though. Either someone will come by it later, or people in the FOSS community will have the opportunity to make it a good thing. Mostly I want to satiate my curiosity.

Louis Savain said...

Chris, thanks for the comment. True AI will come out sooner or later. I don't know exactly when but certainly within our lifetimes and probably much sooner than many expect. It will be much grander and have a much greater impact (good and bad) on the world than most of us suppose. True AI is like an enormous lever, a multiplier of everything that is good and bad. Rebel Speech incorporates a significant part of it, IMO. And this is why I'm afraid.

John Driscoll said...

I am curious if you agree with the main points of this video:

Louis Savain said...

Hi John,

I only watched half of it and I disagree with what I heard. Sorry, I can't watch all of it. The differences in the way computers and the brain process information are irrelevant. Given enough computational power, a computer program can simulate the workings of the brain's neurons. It's a flawed presentation and the arguments are erroneous.

Alex909 said...

Louis, I admire you for having a conscience! In 20 years of living in Silicon Valley I have met virtually no reflection on the societal impact of radical new technologies. Technologists like Hawkins pay lip service in their book and talks, while trying to implement the very thing that you fear most:

I wholeheartedly agree with your concerns regarding AI and the possibility of a total command and control society. (In some sense we are already close, as the Snowden affair and the new laws in France and UK have shown. What is "bulk collection" if not the collection of the haystack which can be used to define "normal behavior", the pattern of life which, should you deviate from it, makes you suspicious and worthy of investigation.) The current technology is child's play compared to what happen if our elites would get access to unsupervised learning technologies!

The fact that you are wrestling with yourself speaks volumes about yourself! It makes me regain faith in my own Christian upbringing and beliefs.

Louis Savain said...

Alex, thanks for your comment. I agree with you that even with the current state of the art in AI technology, governments can easily multiply and abuse their powers. And they do.

Unsupervised sensory learning is the holy grail of perceptual learning. Nobody knows how to do it right; not Hawkins, and certainly not the deep learning crowd. They're not even in the ballpark. So I'm not worried that any of those guys will suddenly find the correct solution anytime soon.

I've come to the conclusion that I've bitten more than I can chew. I can't deal with the moral responsibility that goes with AI research. I'm not equipped for it. I'm just an internet kook and a nerd.

jeanpaul said...

Hi Louis

Been reading this blog for years now (lattice propulsion) and if I'm not mistaken, you never revealed the results of that either. I'm still waiting for that floating elephant :) I do like reading your work but at some point a bit of 'proof' that you are on the right track would be nice. I understand your issue with sharing your knowledge but if everyone thought like that, we'd still be in the stone age.


Louis Savain said...

Hi Jean Paul,

I'm sorry for all that. The lattice hypothesis is scarier than than AI, if that were possible. I deliberately stopped researching it and even stopped thinking about it years ago. The consequences of that knowledge would be disastrous. In a way, I'd prefer that we were still in the stone age. Then I would not know what I know and I would not be so worried.

Now, we are rapidly hurtling toward what some have called an "extinction event". It's very depressing. Humans are a species of assholes but I'm one of them and I love them. I cannot in good conscience contribute to our own demise.

Spruce said...

Hi Louis,

I'm with Jean Paul on this. I find it hard to imagine you didn't consider possible consequences before you started sharing your thoughts. So why even bother? It seems to be a pattern that you trickle enough to keep some readers interested with a promise of a big reveal that ultimately does not materialize because, you know, conscience.

If you do end up moving on to another interest, I look forward to hearing about it. So don't go silent. =)

Louis Savain said...

Hi Spruce,

There is more to it than just morality and conscience. There is a hidden but crucial variable in this equation that I am not at liberty to discuss.

There is no question in my mind that I will eventually release all my work to the public. However, what I have come to understand is that it is not my call to make. There are hidden but powerful players in this game that I cannot talk about at this time.

This being said, maybe it's time to shift gears a little. I think I've figured out a way to finance Project COSA. Computers and software engineering must be reinvented if we are to see the full promise of truly intelligent machines. I'll keep you informed.

Maksym Diachenko said...

Current progress in AI field is more than enough to build autonomous killing machines. There is no need to find something else, some other approach to it.

Malik Muqtadir said...

Keren banget infonya gan.. berkat artikelnya saya telah bisa menyelasaikan karya ilmiahku yang berjudul "Kapa Kita mempercayai Pengetahuan Kita" dalam blog terima kasih bro

Unknown said...

It will be fine, just like any invention... if you are afraid of humanity then you are afraid of intelligence... there are always evil and good forces at play and that is what makes intelligence possible in the first place, so to be paranoid about something that is beautiful is to only deny yourself your existence...

If its not you, it will be I who releases it... I am almost done with the engine, just a few more months...

Louis Savain said...

Good luck on that. Let me know how it turns out.

John Driscoll said...

As far as project COSA is concerned, it sounds fascinating. However, I am not sure I fully understand it. The way I see it, COSA is based on the following premises:

The computer was originally designed and optimized for applications that are much different than what it is being used for today. The original computer was essentially intended to be a machine for solving mathematical problems in a straitforward, step-by-step approach. It was never intended for things like video games, artificial intelligence, etc.

The transistor is an incredibly powerful device. However, by choosing the algorithm as the basis through which humans interact with computer hardware, we have so far been unable to fully realize the full potential of transistors. The transistor is perfectly suited for carrying out parallel processes. There is no reason to restrict it's parallel processing potential by using it for linear processing.

Algorithmic software construction forces the programmer to conceive of the entire program from start to finish, before even starting. This precludes any trial and error approach to software construction. Without any chance for experimentation, the programmer has no sure fire of knowing his program will work upon completion. Many hours of development time can be lost on a program that was doomed from the start.

Because of the enormous amount of advanced planning required to create a working, functional program, software construction is unavailable to those without such gifted mental ability. This cuts out potential candidates who, despite not being able to "think several moves ahead," could still offer valuable contributions to computer science given they had access to a less restrictive software construction environment.

Linguistic programming is archaic, inefficient, and unnecessary.

Louis Savain said...

Hi John,

Thanks for the comment. You pretty much got it right. The essence of Project COSA is that what was done for electronics, physics and biology (gates, particles and DNA sequences) should be done for software engineering. Once we figure out the fundamental building blocks of software and how all programs can be built built by connecting blocks together, we will be halfway there. But for that to happen, we must have something that blows everything else out of the water. This takes money.

I am in the process of raising funds for project COSA (wish me luck). My goal is to reinvent the computer, both hardware and software. The idea of a central processor executing lists of instructions is archaic, cumbersome and error-prone. It will have to be replaced sooner or later. Hang in there.

Rick Deckard said...

Good luck!

Rick Deckard said...

One think that puzzles me, how does COSA compare to other patents from the 80s that you mentioned. I am not sure I comprehend the difference. It seems like all you have to do is create a reactive system and that is the only difference between an algorithmic system. All the ideas have been available for quite some time, but I congratulate you on the complication and I think perhaps is the gem here. I would think it would be difficult to raise money for just a simple concept without having any proof it works. Sounds like catch 22.

SantaClaws said...

The problem is that you are like you said, not a visionary. A visionary used to mean, someone who could literally see all of it, and draw it, and explain it to people, because they correlated the contents of their head, they could review it like they would a book, even as time continues to pass, they can thumb through themselves. If you have a gigantic blog and endless paragraphs of writing, and it doesn't look like its gonna stop, maybe you are not really the teaching type, not really an idea man at all. Like you said, you are frustrated that you have to concieve the entire program at once. If you don't have angels flying around your head presenting you with your own little personal sphere of knowledge that you witnessed in your lifetime crystalizing like fine old wine, you haven't quite 'hypotized' yourself with the Word of God. It's that which makes you feel like you can accomplish things the whole world hasn't yet, but before you get the end of that process which ends conclusively, you people tend to yammer on and get flustered and get delusions of granduer which are even more petty than what you might really accomplish.

As a non visionary, who is all by himself as a lone rebel, you can't predict the consequences of your actions and reasonably be held accountable for them, you could do anything and be seen as an innocent little baby. You could aspire to control yourself and not fear 'what might have been if I had continued to double down on my hunches". I remember reading how you were gonna save your wife, now its the whole worlds fault. Regardless of the facts it doesn't look good. The world is a terrible place full of people who sell drugs and computers they haven't really figured out if they work but they sell them and say buyer beware! People like you, trying to solve problems they think they see, they stumble onto a niche, and they think its their responsibility to cultivate that niche, that without them it'll be opportunity lost. That's the Sindrome Man. You could let the demon out of the bottle, like the guy who invented fascism or the atomb bomb.

bitcuration said...

I recently read Pedro Domingos book "The Master Algorithm", it's quit an eye opener. It's also amazing what CNN has achieved lately. Except there is a fundamental problem that implies this is not the right algorithm or how our brain works, it requires large amount of data. However, the success of DL in vision and auditory may simply means it has captured human brain's low level sensory function which does expose to huge amount information stream since a new born baby comes to the world.

What appears far fetching to DL is the logical and analogical reasoning ability that seem so natural to human (with minimum training though), may beg for new advance in AI. By that, I don't believe symbolic has a chance, but then this book represented a novel attempt.

Rick Deckard said...

SantaClaws you must worship Elon Musk. His anti AI campaign is just a way to raise money.

He is wrong about AI, it won't hurt us and no demon. We have 7 billion most intelligent beings on the planet and the reason they don't all go crazy at the same time is because they are intelligent.

You ever see chaotic traffic in India or Thailand? organized chaos. Do you know why people don't bump into each other? Because they consider the risks of their own survival.

pobri19 said...

My previous comment was deleted? I will repeat. You've been saying this for years now. If you had anything to prove your claims you would provide it. You're an obvious snake-oil salesman. Why you do this, I have no idea, do you like the attention?

Louis Savain said...


Actually, your previous comment was deleted by mistake. Sorry.

Think about what you are saying. I'm not afraid of being seen by almost everyone as a crackpot. Do you think it bothers me that you think I'm a snake oil salesman?

Besides, how can I be a snake oil salesman if I'm not selling anything? My blog does not even carry commercial ads. I don't remember ever asking you or anybody for money for my AI stuff. I don't need it as I'm provided for. But I have worked hard on this and I don't work for free. When I do deliver, I will expect to get paid for my work. As the Master himself once said, a worker is worthy of his wages.

My advice is: be patient. It will happen when the right time comes and when you least expect it.

Adam Becker said...

Hi kid.

If you wanna go "rebel", you'd better do some real harm, instead of seeking attention childishly.

Despite your crackpot behavior set, we do think you've got some potential.

We are working on a spiking neural network aided, AGI capable botnet C&C, its learning mechanism is working but still got some problems. If you ever played a python game called "endgame singularity"(no it does not come from the church of singularity despite its name), you would know what I'm talking about.

Our bot can discover 0day exploits at an alarming rate, via human level thinking. We can already hack the shit out of the Internet. However we've got much more plan than hacking the network.

We could use some of your help if you've got the spare time.

It you'd like more details, or wanna join us, feel free to contact me.

Contact me only on tor, my jabber You may reply here but I may not be able to see or reply further.

Best wishes.

Louis Savain said...


Thanks for the comment and the invitation. I must decline for several reasons. First, I work alone. Second, I have ceased all AI work for the time being because I think it's too dangerous, IMO. Third, I'm focusing on Project COSA, an old idea that is important to me. Fourth, I'm not a hacker. Finally, I'm not a kid. Only at heart.

Sorry and good luck.