ricklentz

I think most companies would be suspect of an external performing this role. 

Why would someone with the domain knowledge as well as the political and financial independence necessary to  '... certify the soundness of big-data applications like anti-fraud techniques or stock-trading systems' not want to use that knowledge to build new and better services?

A key external today is tomorrow's competitor. One could argue the threat of lawsuits works - but these are only deterrents and temporary barriers, not absolute prevention techniques.  Remember, if an External Algorithmists violates these agreements you have to prove they did and you have to pay for the defense of them and, in some states, they are often difficult to defend.  Oh, and what if they are now residing in China, say working on their government-sponsored version of your core product or service?  Or they are now in a country like Finland or Iceland, who have different IP laws and your claims are just not valid in their jurisdiction.  The idea just

ricklentz

Subject/Behavior-Based Search - Maginot Line - Mosaic Theory 

1. The Supreme Court has ruled that what information falls outside of the protection of the 4th Amendment?  Explain why.

The  Court  stated  in  Smith  v.  Maryland that  an  individual  has  no  “legitimate  expectation  of  privacy  in  information  he  voluntarily  turns  over  to  third  parties.”  The cumulative effect of  the public  exposure and third-party doctrines renders data mining largely  “outside the  scope of the Fourth Amendment.
 
2. What are the three criteria that make dog-sniffing acceptable, (i.e., not a violation of privacy)?   The Harvard Law Review article finds dog-sniffing, and the three criteria, a good model for regulating governmental searches of data.  Explain.  (See Part IV of the article.)

The sniff must only analyze information that is legally obtained;  the sniff must only detect illegal activity; humans must not participate in any search until probable cause has been established by the sniff, and the sniff mus

ricklentz

How does the concept of “actionable analytics” fit with what has been presented so far in this course? 

According to Mike Wu, actionable analytics requires knowing the predictive power of the model, the input data, and the accuracy requirement of the prediction. Once these are model attributes are identified, data scientists can form and present meaningful comparisons between courses of action.  Using the model's prediction capability, the analyst generates and offers multiple forward-looking action paths for leadership.  

This process fits perfectly with what we have presented in this course so far.  We have used linear programming and constraint programming to translate business problems into models.  We have dealt with uncertainty through the use of probabilistic models with input distributions.  During the next part of our course, we will deal with increasingly real-world issues (dynamic programming, game theory, metaheuristics, and simulation).

ricklentz

Computer Ethics, Chapter 5: Questions 12-15

12) The four freedoms are:
to run the program, for any purpose
to study how the program works and adapt for your needs
redistribute copies
improve and release 

13) Using Locke's account, developers can argue that the software they create is rightfully theirs because they produced it with their intellectual and physical labor.  Appropriating the software without resource attribution, as stated in the text, effectively making Bingo their slave.

ricklentz

I agree with the DeepMind Ethics & Society mission and approach.  Unfortunately, some autonomous systems cannot be networked and need to function without a tether to the cloud.  

DeepMind doesn't mention that having this strategic oversight role allows them access to a broad audience including directly to many new academic claims.  This is a highly valuable position to be in during this rapidly maturing phase.  

I had to read this twice before it clicked:
"DeepMind Ethics & Society is a new research unit dedicated to this task, carrying out interdisciplinary research that brings together the technical insights of our team at DeepMind and the diverse range of people who could be affected by AI. 

We start from the premise that all AI applications should remain under meaningful human control, and be used for socially beneficial purposes. Understanding what this means in practice requires rigorous scientific inquiry into the most sensitive challenges we face and the inclusion of many voices throughout. We deep

ricklentz

sudo umount /dev/sdj1
sudo dd if=bone-debian-9.2-iot-armhf-2017-10-10-4gb.img.xz of=/dev/sdj1 bs=1M

ricklentz

Computer Ethics, Chapter 4: Questions 9-12.

9.) What would be the benefits and drawbacks of limiting organizations to using only crude categories in their information-gathering practices?

The businesses would become ineffective and would be displaced in the market by competitors without those limits. 

10.) The five fair information practice codes are:
a) Notice/Awareness
b) Choice/Consent
c) Access/Participation

ricklentz

Identify the current and potential competition. This can be done from the customer’s point of view (e.g. those companies vying for the customer’s business) and from the business’ perspective (e.g. those companies using similar strategies). 

Determine five strategic areas in which you are competitive. Zenos chose sustainability, product quality, media, customer service, and price.

Create a competition matrix. Down the left side column write the names of your five major competitors. Across the horizontal axis, list the five strategic areas in which you are competitive (e.g. sustainability, product quality, media, customer service, and price), filling in each box (see table 1 below). Evaluate the strengths and weaknesses of each competitor in relation to the associated strategic area.

Review the matrices to determine which areas your company is most vulnerable. Zenos sees pizza as a price competitive product; its competitive position is threatened if its quality and community appeal do not attract sophistica

ricklentz

Dolphin whistles example -> Move to the frequency domain and use Sakoe Chiba bounds to perform dynamic time warping.

Viterbi Path - most probable path through the trellis

Hidden Markov Models training notes:
  try many different topologies
  use cross-validation
  split independent signals that have the same logical meaning vs. trying to forcing the model (analogous coarticulation of phenoms):
    s1_telem_pos_updt_3M22_Zircon
    s2_telem_pos_updt_3M22_Zircon

ricklentz

Bayes networks define probability distributions over graphs of random variables.

To find how many probability values are required to specify a Bayes Network, identify the number of input variables K for each node(i) and sum 2^(K(i)).

Complex Bayes network representations offer considerable advantage over naive unstructured join representations. 

D- Separation: used for defining independence based on the graph structure and known variables.

Inference in Bayes Nets: Nodes can be Query (want to find), Evidence (know these) or Hidden (not Query or Evidence).