Monday, December 21, 2015

Fear in Silicon Valley

OpenAI, the Savior of the World

Those of you who have an interest in artificial intelligence may have noticed a recent strange announcement from some of the big movers and shakers in Silicon Valley. We are told that humanity faces an imminent existential threat, the birth of true AI. It seems that the best way to defend ourselves against this potential evil is to make sure that any breakthrough in AI is quickly disseminated to the entire world. So they formed a nonprofit corporation called OpenAI to do just that. And they mean business. Silicon Valley luminaries like Elon Musk, Sam Altman, Peter Thiel and others have already committed one billion dollars to the company and they've brought in known experts from the machine learning research community.

Panic at the Top

In my opinion, the real purpose of OpenAI is not at all what its investors claim. Remember that these are people who are already heavily invested in various other for-profit companies that are hard at work on creating proprietary AI technologies. The conflict of interest is obvious. I think this is a sign of fear among the big dogs. I think OpenAI is just brain bait. Here is what I think happened. The lords of Silicon Valley have come to the realization that true AI may emerge from anywhere and not necessarily from the Googles, Baidus, Microsofts and FaceBooks of the world. It dawned on them that it is quite possible that some maverick genius working in a garage somewhere will figure it out before they do. After all, information is freely available on the net and fast cloud computing is getting cheaper everyday. OpenAI is essentially saying: "Look, we got truckloads of cash and we are the good guys. Come join forces with us."

It ain't gonna work. Why? Because whoever figures out AI will have no need of Silicon Valley's money, let alone their reptilian morality. If necessary, he or she will eat SV for breakfast before they realize what just happened to them. We live in interesting times.

PS. In a few days, I will delete a number of old AI articles from this blog because they are obsolete in certain areas. My understanding of intelligence and the brain has evolved tremendously in the last several years and months and those articles do not fully or accurately reflect my current thinking. I also know that a few among you read what I write to get ideas for your own AI projects. All I can say is that I am sorry if I misled you. When the time comes, all will be revealed.

See Also:

Why Deep Learning Is a Hindrance to Progress Toward True AI
In Spite of the Successes, Mainstream AI is Still Stuck in a Rut

9 comments:

Rick Deckard said...

I have studied a lot of AI and I am not impressed. All they figured out how to do is just program a while loop and perform complicated queries based on data structures they arranged just for that purpose.

In order to create the AI, one must solve the puzzle of life itself.

Rick Deckard said...

How about an article which explains what you were thinking wrong and why?

Louis Savain said...

Rick, thanks for the comments. You are right. OpenAI got nothing. They're still doing GOFAI without realizing it. And if they do realize it, they lie to themselves and others that they are somehow doing something different than what they always did: manipulating symbols.

I disagree that we need to solve the life puzzle before we can solve AI. Intelligence is a mechanical cause-effect phenomenon.

As far as explaining where I was wrong in the past is concerned, I was only wrong in a number of details, not in the grand design of intelligence and the brain. I'll just say this for now:

The brain is a massive timing mechanism. It uses a discrete, spike-driven mechanism to process analog quantities. But these analog quantities are not found in synaptic strengths or dendritic polarizations. They are the temporal intervals between various events (i.e., patterns), both internal and external. The brain essentially learns the deterministic relationships between these intervals.

Rick Deckard said...

Its a good definition of the brain.

The thing I am concerned with is why would it even think? Would we have to teach it and why would we even need to teach it?

A plant has no need for a brain to live, yet it reacts to the environment and adjusts its behavior accordingly. I positioned my research based on this.

I think intelligence or life, to me they are the same thing, is simply the mechanism of duality.

Louis Savain said...

Hi Rick. You asked: The thing I am concerned with is why would it even think? Would we have to teach it and why would we even need to teach it?

It will think because it is programmed to think. It will learn like a child and will have to be raised by humans who will condition its behavior. Once it has been properly conditioned and trained, it will continue to follow its conditioning. Intelligence is always at the service of motivation. We, humans, will give intelligent machines their motivations and they will not depart from it. It can lead to utopia or hell.

A plant has no need for a brain to live, yet it reacts to the environment and adjusts its behavior accordingly. I positioned my research based on this.
I think intelligence or life, to me they are the same thing, is simply the mechanism of duality.


Yes, of course. Both plants and brains are complex biological mechanisms designed and engineered by advanced beings. In this sense, humans, plants and animals are alien technology. Brains just have a more complex set of capabilities than plants. It's all about causes and effects.

Rick Deckard said...

Agreed.

Why is it so impossible for people to accept that humans were here first? I didn't take this idea seriously but I gave it some thought.

Logic has a flaw, its not a direct way towards a progression, which is how most scientific community thinks. Complexity is not evolution. Losing has its reward as seen in nature.

Evolution is adaptation to one's environment and most ignore that fact. People use evolution as some sort of progression, but from what I've read the DNA can not tell the direction of change.

We can only assume that we came from a monkey but why is it so impossible to assume that a plant came from a man? Especially in a universe which constantly "deteriorates".

Emiro Diez Saldarriaga Díez Saldarriaga said...

LOUIS, por que no hay entradas nuevas? Deje la pereza: a muchos nos gusta leer5 sus opiniones.

dashxdr said...

Louis, you ought to acquire and take a look at a copy of Neuroscience, 5th edition, by Purves, Augustine, etc. ISBN-13: 978-0878936465 I found some jaw dropping eye openers in that book. There have been a lot of recent advances in uncovering the low level machinery of the neurons. My reference book had previously been a woefully ancient edition of Understanding Biology by Guttman and Hopkins...

If you could overcome your religious faith (bible as metaphor for AI, it's a secret message from god) and eliminate its hobbling of your thinking you might just figure the thing out. Before I do, that is...

I haven't given up on the necessity for variable synaptic strengths just yet. The importance of timing... I agree completely. Your hints at the progress you've made with mechanistic speech recognition... and your writing in general shows we're speaking the same language.

My handicap is the inescapable distraction of the daily burdens of family life. I think your handicap is your faith in spirituality. If you could just get past that I'm sure you'd move forward faster. But then again how much of your motivation derives from that same faith? Motivation is a constant enigma for me as well...

Keep on truckin'!

Louis Savain said...

dashxdr,

Thanks for the comment and the book recommendation. Believe it or not, my first breakthrough in understanding how the brain works was a neuroscience article on the architecture of the retina that I read back in the early 1990s. That's when I realized that timing was the main key to the AI kingdom. I decided right then and there that intelligence depended on the temporal correlations between discrete sensory signals and that there were only two kinds of correlations: signals were either concurrent or sequential. My second breakthrough was to discover soon afterwards that knowledge representations in memory must be organized hierarchically: the "tree of knowledge."

I made very little progress after this until late 1999. That's when I made my greatest discovery. I noticed that some of the metaphors in the book of Revelation were strangely analogous to the model of the brain that I was beginning to construct. I soon became fully convinced that the entire book of Revelation and parts of the book of Zechariah contained a detailed description of the brain, intelligence and consciousness. Unfortunately, the cares of the world made it impossible for me to fully dedicate myself to my research.

Now, 16 years later, my understanding has grown tremendously. I now know that the probabilistic approach to AI is a waste of time, that the brain expects a fully deterministic world and that the Deep Learning crowd, in spite of all the noise and bragging, is completely out to lunch. I know that the brain uses a winner-take-all approach to deal with uncertainty. I fully understand how the brain handles attention, context and short term memory. I understand how motor learning, motivation and adaptation work. I understand how the cerebellum works and why it's necessary. In sum, I understand enough to know that no single person or organization has a chance to figure out true AI even if they had a thousand years to work on it. It's mainly because the organization of memory is extremely counterintuitive. Needless to say, the mainstream AI community is lost in a lost world and they are still doing GOFAI, claims to the contrary notwithstanding.

The point is that I could not possibly have made the progress that I've made without my faith. I do appreciate the friendly advice but I'm afraid that I will have to decline. Hang in there. All will be revealed in due time. Lately, I've been busy figuring out a way to finance Project COSA. True human-level AI will not become a reality unless the computer (both hardware and software) is reinvented.