Tag Archives: Google

The Future of Everything Festival AI Track

 

 

The Future of Everything Festival

From the Wall Street Journal

May 8th, 2018

Artificial Intelligence Session

Today I attended the #wsjfuturefest at Spring Studios in Tribeca.  The session covered all things AI and included discussions from a number of NYU faculty members and AI experts. Screen Shot 2018-05-08 at 2.03.24 PM

The people who spoke in the track sessions I attended included:

  • Gary Marcus – Professor of Psychology and Neural Science, NYU
  • Garry Kasparov – Author, “Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins”; former World Chess Champion
  • Amy Webb – Founder, Future Today Institute; Professor, NYU Stern School of Business
  • Nicole Eegan – CEO Darktrace
  • Sean Gorley – CEO Primer
  • Kate Crawford – Co-Founder, AI Now Institute; Distinguished Research Professor, NYU

 

Here are some of the key takeaways from the sessions:

AI is not the scary big brother that some believe, rather right now it is a pretty limited tool.  It’s great when there is a tremendous amount of data to show a particular specific thing, but doesn’t work the minute it gets complicated.

We want a Rosie from the Jetsons and instead get Roomba, which doesn’t seem to know when it is spreading poop all over the house.

AI can do recognition tasks and classification.  Specifically, we already have speech image and natural language recognition.  AI can play board games and do advertising targeting.  It is not good with small data problems.  If we want better AI, we should start studying small people.  They are good problem solvers.

Right now particular firms are largely responsible for AI. They are Baidu, Tencent & Alibaba from China and the US entities of Amazon, Google, Microsoft, IBM, Facebook and Apple.  Whereas the US is great at algorithms and hardware, we are not as proficient as the Chinese in data – and they have much much more of it given their government and huge population.  They will eat our lunch in tech in the future.  China is fast building as the US is retreating.

Their more advanced systems are now being used to control people.  For example, China has a social ranking system of points that can be used to put people on a no fly list or no school entry designation based on their behavior that is continuously monitored.

What will the future look like?  It looks like the end of smartphones as we move to personal systems with interaction via voice.  We should think of ourselves as data or oil as data are the new oil.  We will have less autonomy and control of our information with an increasing number of devices.

Most panelists agreed that we have little security, but few supported the GDPR.  There was concern both about regulation and lack of regulation.

The CEO of Darktrace talked about cyber security and how certain bad actors seek to invade networks.  For example, there was an Internet attack on cappuccino makers in train stations connected to the train network.  Another entity used the fish tank at a particular casino to obtain network entry through the automated thermostat system.

AI uses generative adversarial networks, which are two neural nets competing against each other. There is a generator and discriminator from two networks that work together to solve the problem.  However, there are means of protection. Specifically, AI can also be used to detect attacks.  One method involves a data set of all prior attacks that is used to categorize and predict future attacks with supervised machine learning. Also, self learning of unsupervised data can find new threats modeled after human body immune response.

In the casino fish tank with internet connected thermostat, the system looked for the key term “high rollers.” The internal controls found the thermostat using too much data and shut it down.

AI can be used to video potential employees during hiring to record each movement and tick of a person. The goal would be to match the patterns to the people who are already in the organization and performing well to reduce turnover and poor performance.

 

 

Advertisements

Bloglette: Google Analytics for Strategy

LaTienda, a brand of Spanish specialty foods used Google Analytics to determine whether shipping costs affected  retailers’ likelihoods to abandon their shopping carts due to the shipping costs. After segmenting customers by region, La Tienda determined that stores from Region B were 48% less likely to complete the purchase than those in Region A, in which shipping costs were lower. As a result La Tienda implemented flat rate shipping for Region B and completed transactions increased 70%, while conversions did not change in Region A.

The full case is available from Think Google. I use these cases in my courses to illustrate the importance of making informed decisions.

/home/wpcom/public_html/wp-content/blogs.dir/108/51763905/files/2015/01/img_3031.jpg

A Sick Application

Image

Last night when I was suffering from the stomach flu, I searched Google using the desperate and relatively long tail key word phrase “how to stop throwing up.”  Clearly I was in pretty bad shape, but somehow still thinking of my blog.  I remembered an article  I had read a while back that explained how the CDC, The Center for Disease Control in the US tracks social media for illness to determine outbreaks.  My memory was correct in that the CDC not only tracks flu outbreaks via social media they have even launched a contest to encourage citizens to predict the 2013-2014 flu season.  Here is the beginning of the press release from the CDC:

CDC Competition Encourages Use of Social Media to Predict Flu

November 25, 2013 — CDC has launched the “Predict the Influenza Season Challenge,” a competition designed to foster innovation in flu activity modeling and prediction. The registrant who most successfully predicts the timing, peak and intensity of the 2013-2014 flu season using social media data (e.g., Twitter, internet search data, web surveys) will receive an award of $75,000 and CDC recognition. Full details of the contest requirements – including eligibility rules, how to enter the contest, and scoring – are available via the official contest announcement at https://federalregister.gov/a/2013-28198.

Aside from the CDC there is an app called Sickweather that uses social media to track outbreaks of a variety of illnesses scanning Twitter and Facebook.  The result is an interactive map showing the areas around the country for infections and other health issues.  I thought this was interesting, but not particularly useful for me.  I was quite aware that the stomach flu was going around.  At least three kids in my child’s class had the flu and on Wednesday, my child got it.  It was passed to me in spite of extra hand washing.

Unfortunately even knowing that the illness was present did not make it possible for me to avoid getting ill. The sickweather app might provide some interesting information, but can’t make up for the realities of parenthood.  One area where it may be helpful is in cases of food poisoning.  Sickweather could track names of restaurants that made people sick.  However, publicly posting this information without clear evidence may have its own issues.

The amount of knowledge and potentially predictive behavior is fascinating, though my illness would not have shown up on Twitter or Facebook.  Google likely has the stronger data set given that it includes search information and also social media.

Here is to avoiding illness and keeping up the fluids…

Click here for: The article on Sickweather from the Star Tribune

Image

Should Marketers Buy Popular Key Words?

The Marketing Science Institute publishes a very interesting marketing newsletter highlighting research in social media and sometimes mobile marketing.  In MSI Insights second issue in 2013 they summarize a study by Jerath and Ma titled “Consumer Click Behavior at a Search Engine: The Role of Keyword Popularity” which examines consumer online search behavior.  The researchers found that key words with low popularity were more likely to lead to clicks by viewers on both organic and paid results.  The data suggest that as consumers delve more deeply into examining a purchase the words become more specific to their needs and they are therefore more likely to click on those results.  The implication for marketers is that they should consider less popular key words and determine the likely scenarios for search as consumers get closer to the purchase.  More general, popular key words may play a bigger role earlier in the purchase funnel.

Image