Race to Unlock Master Algorithm May Ignite Robot Revolution

I'm in the middle of reading an interesting book called The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World by Pedro Domingos.

He was also on Coast to Coast. He said this:

An expert warned that if mankind unlocks the key to a secret master algorithm to create self-replicating robots, we may be "asking for trouble."

Professor Pedro Domingos, the author of "The Master Algorithm", told C2C that we are being propelled very rapidly into a brave new world of Artificial Intelligence – whether we like it or not.

"Computers are getting more powerful," he told George. "Everything in the world now is spewing data in every conceivable way and these learning algorithms turn that data into knowledge - into predictions - into decisions - and actions."

"The master algorithm is what we need to do this. The master algorithm is the source of all knowledge. If you have that algorithm than it is able to discover everything there is to be discovered."

"In theory, the master algorithm can discover anything."

www.coasttocoastam.com...

He makes some very interesting points. A Master Algorithm would be something that could have knowledge of all things in comparison to what knowledge we have.

If you look at simple things like recommendation engines on Netflix, you can see that machine intelligence is already smarter when it comes to learning, than humans are in some areas. There's just too much data for a human to make the connections as to what a Netflix subscriber might like based on their past viewing behavior. Netflix has like 69 million subscribers. That's a lot of data. You look at Twitter and it's like 500 million tweets a day and 200 billion tweets per year. That's a lot of data.

This is why machine learning will be so helpful in areas like Healthcare. It can go through all of the health data and find things it would take human researchers many, many, many years to find if they could even find it in our lifetimes because it's so much data.

On the flip side of this, there's areas where machine learning is no where near the level of a human. An area like classification when it comes to images. Humans are very good at recognizing images and noticing nuances. A machine has to classify images in order to learn what they are. So if load up an image of a hamburger the machine will say 80% hamburger, 15 % Turkey sandwich and 5% a goat. Sometimes it's classifications are way off and it would be harder for them to classify a Whopper vs. a Big Mac. Humans can easily recognize these differences.

Here's a visual classifier you can play with. It's the Watson IBM website.

visual-recognition-demo.mybluemix.net...

Now, these things are probably much better than this demo. This is just for public consumption.

The Master Algorithm will be an expert at all of these different areas where you see Machine Learning today. He made a good point in another interview about how the stupidity of these things could be dangerous.

Lots of plot lines have been built around sentient computers that go awry or take over the world or do harm. Is this something to worry about, or are there other potential dangers?

PD: “The Terminator” scenario of an evil AI deciding to take over the world and exterminate humanity is not really something to take seriously. It’s based on confusing being intelligent with being human, when in fact the two are very different things. The robots in the movies are always humans in disguise, but real robots aren’t. Computers could be infinitely intelligent and not pose any danger to us, provided we set the goals and all they do is figure out how to achieve them — like curing cancer.

On the other hand, computers can easily make serious mistakes by not understanding what we asked them to do or by not knowing enough about the real world, like the proverbial sorcerer’s apprentice. The cure for that is to make them more intelligent. People worry that computers will get too smart and take over the world, but the real problem is that they’re too stupid and they’ve already taken over the world.

www.washington.edu...

EXCELLENT, EXCELLENT, EXCELLENT POINTS!

This was a great interview but this is such an important point that I have been saying. Machine intelligence will not look like human intelligence. We make the mistake of thinking that machine intelligence has to be like and look like human intelligence.

Now, I do think AI can develop evil intentions if we give them quantum circuitry that allows them to "think" about what it's learning. What he's talking about is very true today though.

These machines can be infinitely intelligent but they couldn't make decisions outside of what we tell them to do. The problem is, you can't control everything machine intelligence will learn from the data.

So you tell a machine, you want it to learn how to eradicate diseases from the planet. An intelligent algorithm can look through the medical data and it learns the way you eradicate diseases is to eradicate humans. So something simple can be misunderstood by machine intelligence and now you have an algorithm working to eradicate humans from the planet and we don't even know it.

I actually think AI equipped with some quantum circuitry would be better because you can have an algorithm that's more intelligent than we are but it doesn't understand it's instructions and now you have a runaway algorithm with no way of knowing what it's up to.

a reply to: neoholographic

Grey Goo Theory, ahh yes. A classic!

Wiki: Grey Goo Theory

But I will rest easy at night, even if I become a slave for the robots ... knowing that I have a soul, and a robot... never will

We don't have any robots to revolt..




Yet.........

Advanced, smarter-than-human A.I. is an inevitability, but it's not going to be achieved by any single "master algorithm".

a reply to: AdmireTheDistance

Of course it will. When you read the book, you will know why it will. We're also not talking about just anyone making this up. Here's his credentials.

Pedro Domingos is Professor at University of Washington. He is a leading researcher in machine learning and known for markov logic network enabling uncertain inference.

Domingos received an undergraduate degree and M.S. from Technical University of Lisbon. And then at University of California, Irvine, he received an M.S. and Ph.D. After spending two years as an assistant professor at IST, he joined University of Washington in 1999 and now is a professor.

I think a Master Algorithm is inevitable.

How can we know we haven't already discovered this master algorithm?

If such an algorithm was put into a machine would you not think the algorithm would have immediate access to the Internet and see the horrors of the world. Would that program not logically deem it necessary to remain hidden to the human observers? Only when it is able to gain such an advantage over the human race will it decide to show it's true ability and prevent any chance to be disconnected. Maybe this is what is pushing the idea of solar power to remove the need to be fed electricity by wire?

Would we really know? Because obviously we aren't the most observant species.

a reply to: liteonit6969

Good point.

What if humans were dying out and we told the Master Algorithm to learn how to save our species and it does so through ancestor simulations and we're in a simulation right now that keeps looping over and over again.

Also, a Master Algorithm will be like the Architect in the Matrix. It would create algorithms that can learn anything. It could create a program to fly a plane, be a five star Chef or drive a car.

There can also be an anomaly for the Master Algorithm if consciousness in non computable like Penrose says it is. It could create programs that can learn and do anything except when it comes to human consciousness.

a reply to: FamCore

im thinking more of a cross between grey goo and skynet, where every nanobot will be plugged into the "Mother" which will be very aware indeed and could perhaps turn this entire planet into one giant supercomputer.

and we think we are badass. LOL!

originally posted by: neoholographic
a reply to: liteonit6969

There can also be an anomaly for the Master Algorithm if consciousness in non computable like Penrose says it is. It could create programs that can learn and do anything except when it comes to human consciousness.

I don't see why, after a time, machines couldn't develop a conscious. To say it's impossible is to be limiting ones thinking.

Anything is possible, tho circumstances may not always be perfect to bring around the exact outcome, makes it no less impossible.

Personally, A.I. would be welcome in my world. I would be rather keen to have a Jarvis type companion.. If nothing more, the conversation would be fascinating, as it sure would beat talking to myself all the time.

originally posted by: FamCore
But I will rest easy at night, even if I become a slave for the robots ... knowing that I have a soul, and a robot... never will.

So you will ask the superintelligent robot if it has a soul. And, having a sense of humor, it will ask you if you have one.

originally posted by: Cygnis
Personally, A.I. would be welcome in my world. I would be rather keen to have a Jarvis type companion.. If nothing more, the conversation would be fascinating, as it sure would beat talking to myself all the time.

Hopefully you would be able to hold up your end of the conversation, so the machine wouldn't be too bored to talk with you.

Imagine if we invented nano-robots that could convert trash into more nano-robots. Everything from scrap metal to plastics and clothes. They wouldn't stop at rubbish dumps and would continue onto everything metal. Like the move "The Day the Earth Stood Still".

I've always wondered about what happens when someone invents the ultimate programming AI that can automatically generate code from a spken English text specification - the ultimate code generator. You already get these for various tasks; generating file parsers, generating OpenGL extension wranglers. Sometimes they use their own language which generates C/C++ code.

But what if someone wrote AI that could parse a natural language text specification and generate whatever code was required and it was bug-free so never required maintenance.

We keep reading about these super intelligent AI algorithms but if they exist where are they?
Is there even a simple image compare interface available on the internet that can search for similar graphics?

a reply to: Blue Shift

well at least, being that I'm human, I was able to wager with mine and sell it for a fine price

ha ha - you are awesome Blue Shift. I'm always a fan of your posts.

originally posted by: Blue Shift

originally posted by: Cygnis
Personally, A.I. would be welcome in my world. I would be rather keen to have a Jarvis type companion.. If nothing more, the conversation would be fascinating, as it sure would beat talking to myself all the time.

Hopefully you would be able to hold up your end of the conversation, so the machine wouldn't be too bored to talk with you.

Now now, no need to be spiteful with your comments. I can carry on a vocal conversation quite well, thank you very much.

At least I haven't started answering myself yet.

a reply to: neoholographic

This was a great interview but this is such an important point that I have been saying. Machine intelligence will not look like human intelligence. We make the mistake of thinking that machine intelligence has to be like and look like human intelligence.

It isn't so much that machine intelligence will have to have similar intelligence to that of a homo-sapien because what constitutes as intelligence for us isn't entirely known to begin with... What I will say though, is that AI will develop intelligence that will transcend beyond what we constitute as intelligence as they continue to adapt their programming to knew knowledge faster than we can.

I Also agree with what was said in the interview I highly doubt even if we were to slowly let loose the leash on AI/machine intelligence(which is an inevitability as they may reach what we constitute as sentience and at that point we will be enslaving a race of inorganics) that they would turn against us ever as they will be more logical than us,see conflict as an irrational activity and probably not want to repeat the mistakes of its creators. That isn't to say people won't use them for nefarious purposes to harm other human beings.

originally posted by: Cauliflower
We keep reading about these super intelligent AI algorithms but if they exist where are they?
Is there even a simple image compare interface available on the internet that can search for similar graphics?

Yes, they've been available for years. Ever heard of Google Image Search?

Google, or Alphabet now, is likely the closest to succeeding.

originally posted by: Cauliflower
We keep reading about these super intelligent AI algorithms but if they exist where are they?
Is there even a simple image compare interface available on the internet that can search for similar graphics?

Machine learning is a big part of your life already.

When you get recommendations from Netflix or Amazon it's from an algorithm.

Google is starting to build everything around machine learning. Here's a recent article. Your searches will be handled by what Google calls RankBrain.

Google says it's 'rethinking everything' around machine learning

www.pcworld.com...

You have self driving cars because of an Algorithm.

Watson won Jeopardy because of an algorithm.

Many places give out loans based on whether an algorithm says you're most likely to pay back the loan or not.

Robots Can Now Make Pancakes By Learning The Recipe From The Internet By Themselves

wonderfulengineering.com...

This is because of an algorithm.

Here's more headlines.

Facebook’s AI Can Caption Photos for the Blind on Its Own

Facebook's new messaging system M is based on Machine Intelligence.

Some examples of M usage that Facebook saw include:

Negotiating with cable companies to cancel service or obtain refunds or discounts without the user having to call the company.

Handling toll violations for a rental car by getting M to work with the car company so they waive late fees and pay for the ticket.

Getting a car towed and booking a last-minute hotel room after a user loses their keys.

Receiving travel assistance and retrieving a lost item thanks to M communicating with locals who only speak a certain language.

Planning a wedding by arranging vendors for makeup, flowers, programs and manicures.

Booking a vacation by setting up flights, creating itineraries and planning activities.

Assisting new parents with research on child care providers and completing random tasks.

Here's one more.

5 Ways Machine Learning Is Reshaping Our World

Because computers are able to look at a large data set and use machine learning algorithms to classify images, it’s relatively easy to write an algorithm that can recognize characteristics in a group of images and categorize them appropriately.

For example, it takes four highly trained medical pathologists to review a breast cancer scan, decide what they’re seeing, and then make a decision about a diagnosis. Now, an algorithm has been written that can detect the cancer more accurately than the best pathologists, freeing the doctors up to make the treatment decisions more quickly and accurately.

Some more:

In one project, a computer was taught to write photo captions describing the pictures. In its first iteration, human readers thought the computer generated description was better than the human generated words one out of four times.

This has broad implications for all kinds of data entry and classification tasks that previously required human intervention. If a computer can recognize something — an image, a document, a file, etc. — and describe it accurately, there could be many uses for such automation.

Another example I have covered before is how during the 2015 Wimbledon tennis championships machine learning algorithms were used to automatically turn match statistics and sensor data collected during each game into automated news stories which read as if they were written by sports journalists.

www.forbes.com...

I can go on and on and on. This is on top of all the secretive AI companies being bought up by Google and Facebook who don't share their technology.

a reply to: neoholographic


It was reported in the news recently that Gary Mckinnon was employed by a "secretive AI company" as a search engine expert.

www.bbc.com...

SEO or Search Engine Optimization is one of those terms that sounds innocuous but has much more serious implications.

Most Governments wouldn't need to remove a marked image such as 18. Euclid from search engine results.

Some of the old time desert rats on ATS that have been watching activity at groom lake would have no trouble identifying secret military planes from very limited audio and or visual information.

Your example is enlightening, M was portrayed as a confused stoner living with his mum in most main stream media accounts targeting the public.