College Cool LOL

Obama Kills Kim Jung Un

Did this for my Photoshop assignment. Lol.



An Expanding and Expansive View of Computing

I attended Jim Kurose’s talk on the advances in computer and information science and engineering primarily because I wanted to gain some insights on the state of technology today, as well as get his novel interpretations of the future of the field, being at the crossroads of both industry and academic in his role as Assistant Director at the Computer and Information Science and Engineering (CISE) Directorate at National Science Foundation (NSF).

During the first half of the talk I learnt that the National Science Foundation mission was more than just advancing science without a specific aim. I had always thought that grants were given out as long as the idea seems plausible and if it succeeds, would boost the development of science and technology to a certain degree. It turns out that NSF’s aims to advance science so as to boost national health, economic wealth and social welfare as well as defense. There are 5 main arms that CISE is focusing on – these 5 are part of the White House Initiatives. These include data science, smart systems, understanding the brain, smart cities, and a nationwide strategic computing initiative. Of those facets, I am most familiar with the one on brain science with the creation of the BrainHub to look into brain science at CMU. I believe this was done after President Obama announced an initiative to research into brain science so as to find new ways to treat disorders such as Alzheimer’s or autism. I think these are great domains to research into, though we should not neglect the security aspect of it. Sometimes, trying to network anything leaves us more susceptible to attacks via side channels. I believe that an engineer can add value to brain research. For example, in 18-220, we learnt about Telegrapher’s Equations in a coaxial transmission line which incidentally can be used to model the transmission of an action potential, an electrical impulse through the axon of a neuron in our nervous system. After all, both science and engineering are cross and multi-disciplinary – both fields are not mutually exclusive.

He later talks about the details of CISE such as its organization chart and finances, which aren’t that all interesting. However, the partnerships that NSF has forged are essential to their success at each cycle of development – from discovery to innovation. This can be seen at present as many companies license patents from universities to be used in consumer-facing products. This way, researchers can leverage their resources and increase the speed at which R&D is being conducted.

Kurose then talks about each emerging field of CISE. He recognizes that a national-scale experimental infrastructure would benefit research so that parallel computation data can be exchanged across institutions from both coasts quickly. He proposes the use of a public cloud infrastructure to achieve this. He also goes into the central theme of his talk, which is that computing should focus more on societal applications, while not neglecting the human aspect of it. It is also beneficial to let computing be pervasive around us, to the point that it does not get in the way of everyday life (ie. it just ‘knows’). Finally, he notes that investment should not solely be poured into upstream research, but also into its most basic tenets – quality computer science courses for students. With that, they have developed pedagogies and curricula for a PreAP as well as Advanced Placement course in Computer Science Principles. I think this is a far-sighted move on their part since there is likely to be more positive externalities in quality computer science education. One proposal to think about for engineering is whether our pedagogy can transition to a more application-based, hands-on education, rather than the systematic, theory-based one we have now.

In sum, I feel that I gained some insights into the role of NSF and what they have done in improving the quality of research and education. However, he did not cover some of the points he promised to make in the questions the audience posed in the middle of the session, such as the one on Power and Water sustainability – how much research has been done in ensuring such natural resources remain available into the future? Also, I would actually like to learn more about the NSF funding process – how grants are given and what the process is like. Last year, I went for a talk on patent registration and I found the whole process very enlightening – I was hoping this talk would shed some light on this process.


Pusheen’s Facebook News Feed

Pusheen and Facebook

College Commentary

Thoughts on Yelp Tech Talk

I attended the Yelp Tech Talk, organized in partnership with Society of Women Engineers. It was held at Margaret Morrison Room A14 on the evening of 29th September. What appealed to me about the event was the topic on Yelp online advertising and how Yelp optimizes it so that everyone benefits – the advertiser gets a steady stream of customers, customers get to discover great new food, and Yelp earns a cut from the advertising fees. Personally, I have been keeping abreast of the pay per click advertising industry because it is a fascinating field. From advertising optimization to fraud detection to machine learning analytics, there are many domains in advertising which interest me. Hence, I wanted to explore how Yelp does things differently from Google, if they are.

The first half of the talk was more of a pitch to tell students how great Yelp is, and how each Yelp employee can look under the hood and work on a feature based on feedback from customers. They gave the case study of a person asking Yelp why they could not order take-out directly on the Yelp website, when they were already searching for what to eat. Subsequently, he discussed how his team made an API, or an interface, for other companies to integrate their services on the website in a tighter way. For example, OpenTable is able to use this special API link to display the number of free reservations on Yelp, while other on-demand services can choose to provide deliveries for a particular food place.
What I garnered from this segment of the talk is that it does not pay to go all-in to every segment in your market. Rather than creating a whole new subsidiary like Yelp On Tap to provide food delivery or reservations, Yelp chose instead to provide the necessary infrastructure for other companies to more tightly integrate their services on Yelp’s website. This is counterintuitive to many business people – after all, why lead people away from your site when you could be the one profiteering off this new segment? However, I believe Yelp might have reasons for doing this even though it was not discussed. Firstly, they might not have the necessary knowledge to succeed in that niche. OpenTable probably knows how to do it right; why reinvent the wheel and try to compete with them? Secondly, Yelp is a lean company and they might not want to waste resources on things that do not add any innovation to the industry. After all, integrating services is not innovation. Finally, they might be emphasizing Yelp as a platform, much like how Facebook is. When Facebook tightly integrates features like embedded YouTube videos on their site, they are not leading users away; rather, they actually gave more reasons for users to remain on the website since they can watch the video on Facebook. Similarly, Yelp users do not need to navigate to another app to make a reservation – they can do it all from within the app.

The latter part of the talk discusses the Yelp Auto Bid Genius, which is an internal algorithm which determines what ad to show to customers. She demonstrated that the highest paying ad does not necessarily get the top spot – a combination of click through rates as well as bidding rates do. She discussed about Yelp ad formats, including click-to-call where the advertiser pays each time a call is made to the business from Yelp. I was disappointed that she didn’t go in depth in exploring conversion tracking – how effective advertising on Yelp can be when measured quantitatively. Yelp advertising can cost up to $600 for every 1000 times an ad is shown, which makes it insanely expensive. There are many occurrences of competitors viewing a particular ad many times to drain the advertising budget of their opponents. How does the business know that the customer came via an ad on Yelp? Some of the methods off my head include using special coupon codes which can be assigned to the viewer of the ad. However, I was wondering if there is anything Yelp can do to further increase the tracking and engagement of its users. For example, if I view a listing on Yelp and I go to that place to have a meal, paying for it using my Yelp-branded credit card, then Yelp can perhaps give me a 5% loyalty rebate for using Yelp. Although Yelp technically loses money, it will gain a lot of insights into the users and activities of Yelp. Of course, the privacy concerns must be looked into, but as it is, Yelp engagement is practically non-existent. Users only fire up the app or website only when they need to, which can be a good thing too, since this means each visitor on the website is a ready paying customer.

It was an interesting evening to think about advertising on a ‘new’ platform like Yelp. I was previously more acquainted to text ads on Google and those intrusive display banners on the web, but I have gained some appreciation of what it takes to make advertising on Yelp more effective, and the technology to make it work effectively.