Ok to set the table, you’ve heard me speak about robotics, machine learning, and AI and within this course, we see how Facebook posts and Google searches may be manipulating us by showing us what they think we want to see or what the programmers want us to see but – did you know that some of the newspaper articles you read are being generated by computers (Open AI project => GPT)?

Robots write thousands of news stories a year

News-Writing Bot Is Now Free for Everyone

Now in the first article, consistent with AI & Robotics advocates, the Associated Press states automated news will provide better coverage but what happens when the computer can assess and correlate every image, every conversation and communication and available data… that is coming with Machine Learning and Big Data.

Setting the stage – here is GPT-3

And lilmiqueia with 1.8M+ followers

https://www.youtube.com/watch?v=6bn3tUUtj2M

AI Intro

First, it is important to obtain a grounding to understand the complexity of AI. Traditional AI problems required programs that were largely classifed as NP-Complete which means the complexity is beyond polynomial time (i.e. Not Polynomial => NP).   As a basis please read about Analysis of Algorithms and how their complexity is determined. Moving on, please read about NP-Complete

Now to complete our AI classification of complexity/programming,  AI-Complete problems represent the most difficult problems that are equivalent to making computers as intelligent as people and are hypothesised to include computer vision, natural language understanding, and dealing with unexpected circumstances.  while solving any real world problem.  A simple example of these are CAPTCHAs

Traditional approaches to solving AI problems includes Fuzzy Logic and Probablistic Logic.  These use continuous or varying truth values between 0 and 1 in contrast to digital logic with only true and false.  Another approach is to capture and apply heuristics as is done with Expert Systems.  I have a demo of Inference Engines in the Prolog Menu System.

Machine Learning

Now we were progressing at a reasonable pace with the traditional AI approaches above however Machine Learning has quickly emerged and evolved and when coupled with Quantum Computing represents an issue that we need to assess very carefully.  Stephen Hawking has recently stated that “AI could spell the end of the human race“.

Here is Elon Musk on the dangers of coupling Quantum Computing and Machine Learning.

Ok, recall the Jurrasic Park quote that essentially stated that the researches proceeded to see if they could do something without stopping to think about whether the should do it. Researchers research and that’s what they/we do but we have to reflect on the repercussions.

One of the presenters spoke about machines learning from their creators and that we may not have anything to worry about but please keep in mind Organized Crime and even Governments are also working on this for hacking and other functions that may have untoward negative effects.

Deep Learning introduction, examples and extrapolated extensions

Now in this past presentation we saw that computer’s/Deep Learning can listen and watch of course they can read,  Are many of our communications digital.  Is our environment increasingly populated by microphone (consider the IoT and even your cell phone in you pocket that could be turned on without your knowledge).  Also consider LiFi.  Now let’s consider Deep Learning applied to human behavior.

AI, Ethics and Manipulation

Here is one application of machine learning and we will discuss this further in privacy as facial expressions and body language can be used to infer intention but equally concerning consider Google’s recent patent that will allow robots to assess our mood and then alter their interactions with us dependent on our mood.

http://spectrum.ieee.org/automaton/robotics/robotics-software/why-googles-robot-personality-patent-is-not-good-for-robotics

So should we worry about ML/AI?

Cal Berkley’s Stuart Russell

Image Understanding

Deep Learning application – facial expression understanding.  

Well, Deep Learning can already be used in facial recognition and now it can be used to learn and understand expressions.  The following research was done to add emotion to digital communications but as we must, we must assess the possible negative ramifications.  We have also seen that facial expressions and body language can be used to infer or predict action intentions so what happens when a computer can detect emotions.

Now consider Google’s patent allowing robots to detect emotion, upload it to the cloud and distribute this to other robots so that they can interact with us.  Note I stated interact with us but I could have stated “manipulate us”.

Putting it all together

Lastly, what happens when our computers get smarter than we are, have access to comprehensive historical data and can listen to, observe and even statistically predict everything we do?  Again, we are the dominant species on the planet because of our intelligence but what happens when we are no longer the highest intelligence on the planet?  Also consider Google has created a robot that has gained emotional intelligence.

Resources

[listyofiles folder=”wp-content/uploads/knowledge-management” options=”table,filesize,icon”]