Wednesday, November 14, 2018

Technological Singularity




Who is Ray Kurzweil?
Ray Kurzweil is a Computer Scientist who received his Bachelor's from MIT. Kurzweil is an inventor, author, entrepreneur, and futurist. He has been involved in many AI related fields, such as Text-to-Speech Synthesis, Optical Character Recognition, and Speech Recognition. He is currently the Director of Engineering at Google; a position he has held since 2012.


Why is Ray Kurzweil relevant?
Ray Kurzweil believes that we are rapidly approaching what is known as a Technological Singularity. A technological singularity is the idea that we will invent an artificial superintelligence that will trigger massive, uncontrollable, unfathomable technological growth. The belief is rooted in the idea that if machines can learn, then it may be possible for a machine to continuously learn until it becomes “smarter” than its human creators.




1990: A computer will defeat a world chess champion by 1998.
Outcome: In 1997, IBM’s Deep Blue defeated Garry Kasparov. 



1990: PCs will be capable of answering queries by accessing information wirelessly via the Internet by 2010.
Outcome: PCs were capable of this long before 2010.


1990: By the early 2000s, exoskeletal limbs would let the disabled walk.
Outcome: Companies like Ekso Bionics and others now have technology that does just this and much more. In fact, University of Pittsburgh developed artificial arms that could allow someone to “feel” through the prosthetic limb.




1999: People would be able to use their computer via voice commands by 2009.
Outcome: Computers were in the early days this feature by 2009, however today we have very sophisticated software such as Apple’s Siri, Amazon’s Alexa, and Google Now.

 


2005: By the 2010s, the software would be able to perform real-time language translation for words spoken in a foreign language and display the translated text to users via the glasses they are wearing.
Outcome: Skype Translate, Google Translate, and more have been developed.

 




Late 2010’s: Glasses will have images show up onto the retina.
"Google Glass"


2020’s: Diseases will be gone because of the newest technology of nanobots and they will become more advanced than current medical technologies. Then the normal human eating will be replaced by nanosystems. Next, the Turing test will be passable, and there will be self-driving cars which then people will not be able to drive on the highways.

"Self-driving car"
"Nanobot fighting off diseases"        


2030’s: Virtual Reality (VR) will be 100% real and make people feel like they can actually be there. Then we will be able to upload our minds/consciousness by the end of this decade.



2040’s: "Non-biological intelligence will be a billion times more capable than biological intelligence (a.k.a us)." Then nanotech foglets will be able to create food out of air and objects that are in this world.

  

2045: Humans will be able to multiply our intelligence a billion-fold by connecting wirelessly from our neocortex to an artificial neocortex in the cloud.

 


Our Thoughts



It is hard to argue with someone who is so established in the field. Ray Kurzweil has made many contributions and currently works at one of the leading innovators of our time. Plus, many of his past contributions have come true, or at the very least have comparable modern technological products. Despite this, it is still hard to believe some of the predictions. For example, the idea that by the 2020s we will have nanobots that cure MOST diseases when our current technology has yet to produce cures for illnesses such as cancer, HIV, heart disease, etc. Uploading our mind/consciousness to a computer is also particularly hard to believe. As a Computer Science major, I have experience in studying/developing systems with data storage/manipulation, algorithms, and AI. While I am far from an expert in my field, I understand many of the challenges we face today in knowledge representation in a computer system. The idea that we could encode human information in the brain, when we ourselves still barely understand the brain, makes me skeptical to say the least. Even more hard to believe, is the idea that once encoded, this information could somehow be decoded and transmitted to a new brain. Unless that “brain” was mechanical or artificial itself, I don’t see us having the technology to be able to implant that much data into something as complex and fragile as the human brain anytime soon.


While the AI we have developed in the modern age is impressive, it still lacks the true intelligence to surpass us to the point that it would be able to develop encoding for something as complex as the brain without us first understanding the brain itself. A good example, would be a neural network. A neural network is a popular approach to machine learning in the modern age. We use it for things such as stock price prediction, OCR, face detection in images, reading human hand-writing in images, email spam filtering, natural language processing, predicting movie user review ratings, etc. Neural Networks are used in a vast majority of Machine Learning related applications and have one “setback.” They are problem specific. A neural network is trained on a set of inputs that represent a specific task. A neural network that is built and trained to predict stock prices would not know what to do with an image of a face. To train a neural network on the amount of input that would represent the wealth of human decision making/knowledge wouldn’t be feasible. Both in structure and in the amount and complexity of the data that would be needed to do so. The idea that we are close to the type of virtual learning necessary to reach human intelligence is wishful thinking.


There are some predictions he has made that are believable though. Self-driving vehicles becoming more popular on the road by the 2020s is not a far-fetched concept at all. We already see a lot of support and success in self-driving vehicles today. The Turing Test is already passable in certain scenarios, provided the human has no physical or voice-related contact with who they’re speaking to. For example, someone chatting with a customer service bot may not realize they are speaking to generated text, if their conversation strays from becoming too broad.

No comments:

Post a Comment