Artificial Intelligence – Future Ahead

Among all the debate about AI involving visionaries and business legends, here is a wonderful discussion with one of my favourite AI experts (and teacher) – Andrew Ng and some of his more humane concerns, considerations. I am glad he acknowledges reluctance of re-learning (reskilling) as one of the bigger challenges. I believe this is going to be a large part of the counselors’ or therapists’ work in the next few years as more people become redundant in their organizations.

A must read for anyone curious about AI and its future –

“As an AI insider, having built and shipped a lot of AI products, I don’t see a clear path for AI to surpass human-level intelligence,” he said. “I think that job displacement is a huge problem, and the one that I wish we could focus on, rather than be distracted by these science fiction-ish, dystopian elements.”
“I’ve been in a lot of private conversations with AI leaders, or business leaders who are working on new AI products that will wipe out tens of thousands of jobs in a single company, maybe more across multiple companies,” Ng said. “And the interesting thing is that a lot of people whose jobs are squarely in the crosshairs of the technologies, a lot of people doing the jobs that are about to go away, they don’t understand AI, they don’t have the training to understand AI. And so a lot of people whose jobs are going to go away don’t know that they’re in the crosshairs.”
“I think one challenge that’s harder to get around is that if I am a master welder, and a lot of my identity is bound up in the respect I command as a master welder, needing to take on a new role where I’m now a novice, I think that’s emotionally challenging,” he told VentureBeat. “That’s actually a significant challenge we need to get through. For what it’s worth, once I was a master Basic programmer, and then I had to learn Python.”

Read the complete article here: AI expert: Worry more about jobs than killer robots

And here is the reference to the debate between Elon Musk and Mark Zuckerberg

Featured image: Kismet robot with rudimentary social skills at MIT, PC: Wikipedia


Essential Machine Learning Algorithms in a nutshell

I am sharing some brief but insightful videos that explain the essential Machine Learning (ML) algorithms quite well. All these videos are part of Data Science and Machine Learning Essentials course by Microsoft on edX platform.

If you’re interested in learning Machine Learning thoroughly, I would highly recommend longer Machine Learning course by Stanford University professor Andrew Ng on Coursera platform. It is one of the best CS courses I have ever taken!

Watch these wonderful videos –

  • Classification – In classification we try to predict if the given test entity belongs to a specific class or not based on the training set we use to train the algorithm. Thus, classification is predicting a true/false value for an entity with a given set of features. For example, we use classification to determine if the given email is a SPAM or not. The mail is checked for various features such as presence of certain words in its contents, the sender etc. to determine f it can be classified as a SPAM or not. It can also be used to detect credit card frauds, detecting if tumour is malignant or not and many such classification problems.

  • Regression – Regression is used to predict a real numeric value outcomes. It can be used to predict sales figures, number of customers for the business based on the training set we use to train the algorithm. The training set examples contain features that denote factors that are most likely to have effect on the outcome. For example, to predict selling price of the house, its total built-up area would be one of the most important features.

  • Clustering – Unlike Classification and Regression, clustering is an unsupervised ML algorithm. In clustering, we try to group entities with similar features. For example,clustering can be used to determine the locations of telephone towers so that all users receive optimum signals. We may also use clustering to group products or customers where we may not have established categorization.

  • Recommendation – Recommendation is used to recommend an item to a user based on his previous usage/purchases or preferences of similar users. For example, it can be used on online shopping sites such as Amazon to recommend new books or items to a user. Netflix uses it to recommend movies to their customers.


Exploring Images using Principal Component Analysis (PCA)

I contemplated using Principal Component Analysis (PCA) for one of my recent projects in Machine Learning (ML) with Python as we were trying to figure out and eliminate some redundant features in our data.

As it turned out, PCA wasn’t useful for doing what we were trying to do and we had to use another algorithm for feature elimination. Nevertheless, it allowed me to dig deeper in PCA and figure out how it exactly works.

My key leanings from the exercise about PCA – PCA can reduce dimensions, but not eliminate them. PCA doesn’t eliminate dimensions and keeps others from the original data. It transforms your data in a number of dimensions whose data are completely different from the original ones. And that made us chose another algorithm for dealing with our problem which needed us to eliminate few few features and run Logistic Regression (LR) on that.

Remembered all that while reading this post showing beautiful visualization of PCA extracting features from photos. It is wonderful to read about recreating those images by reducing components. The image breaking down is done with faces and he has chosen fashion to illustrate PCA. Brilliant!!! It even contains link to code on GitHub. 🙂

Definitely worth reading if you are interested in Principal Component Analysis (PCA), Eigenvalues or Machine Learning (ML)!

Principal Component Analysis and Fashion

Principal Component Analysis and Fashion