Since time immmorial, humans have been interested in astronomy and space. This can be dated back to the work of Aryabhatta and the ideas of Greek astronomers like Aristotle. However, human journey to space, as we know today, happened much later. Today, we are in a race to conquer the planets and use our knowledge of space acquired over centuries to advance many technologies.

Space is an important frontier for many technological advances. Global Positioning Systems (GPS), International Space Station (ISS), communication satellites, and (hopefully) very soon low earth orbit satellites beaming wireless internet are some of the greatest engineering marvels. Beyond space tourism and inhabiting Mars, there are many more important reasons why space might be very important.

This fun post shows the number of trips made by American and Russian astronauts and the various expeditions to space over the last two decades.

We perceive our sense of self as an amalgamation of the different sensory inputs - visual, tactile, locomotive, etc., that our brain processes and interprets. Or in other words, our sense of self is an interpretation made for us by our brain (mind) after processing several inputs. In olden days philosophers who tried to understand one’s true self found it difficult to reason about the interpretation of sense of self - are we really who we are? or are we manifestations of our mind’s interpretation of several inputs? This is a fascinating topic and lately neuroscience research is trying to understand which parts of the brain are processing and how our brain creates such interpretations.

In his book, The Man Who Wasn’t There, journalist Anil Ananthaswamy takes us through the neuroscience research on this topic. The main message of this book is that our sense of self is an interpretation of several inputs which are beyond just sensory inputs but our own physical body and society play a vital role. And in this age where Computer Scientists are working on Deep Learning as well as trying to understand the neurological pathways in order to build systems that mimic the human decision making apparatus, while that apparatus itself is not completely understood.

What fascinates me about this study is how the mind’s computation of sense of self can be influenced (or cheated) by changing the mapping (think of it as lookup table entries) the mind preserves for different sensory and other inputs. In many cases, what we feel and believe our sense of self is, is just an interpretation (an entry in the table) that is fetched given the inputs. We can influence such inputs to pick up a different entry that would alter our sense of self. There is also a classic neuroscience experiment from the 90s that demonstrates how the mapping in our brain, representing our physical sense of self, can be altered within minutes. You can take a look at an example experiment below:

I am intrigued about the possibility of changing our emotional sense of self and if it could be done non-invasively using technology. Such a possibility would be a big boost to curing several mental disorders just by altering the mapping in the mind, when we know how the mapping is created, where it is stored and how it is interpreted. Also, one needs to be aware of ramifications of this - can such mappings be altered by say adversarial inputs - visual or societal - in the form of social media posts or news articles. We might be destroying our own sense of self by letting our mind interpret the inputs that we consume, without us realizing the end result - thinking we are who our mind interprets (influenced interpretation/self) rather than who we really are (actual self).

With online education becoming the norm due to the pandemic there is a growing debate about inequality that is brought about by access to reliable, affordable internet connectivity. At the same time, many countries in the developing world are seeing a huge increase in people connecting to the internet, many of them for the first time. With prices of internet becoming affordable in many parts of the developing world, more and more services are being offered online.

If the number of subscriptions in 2019 were any indicator of the growth, the pandemic has only accelerated the subscriptions. More and more of our daily activities like education, work, essential services like shopping groceries, etc. all have come to rely on internet connectivity. The mobile phone, being the most affordable computer in many parts of the world today, with an internet connectivity dictates whether you can participate in these activities or not.

While growing internet connectivity is certainly good news, let us see how affordable they are across the globe. The cost of internet in most parts of the developing world is much less than in the west. However, the quality of internet is not as good as in the west. The visualization below shows the cost of internet (in USD) in different parts of the world as well as the mean download speed. We can see that the mean download speed is relatively poor in many parts of the world. Crucially several activities like education, work, etc. require not just a connectivity but a good one.

Even when the pandemic ceases to exist, many important everyday activities, commerce and essential services would require internet connectivity. With AI based upscaling technologies, we can hope that many of the educational material could be delivered even with a poor connectivity without reducing the effectiveness of the medium of delivery. In parallel, technologies like low orbit constellation of satellites delivering internet, like the ones being planned by Starlink and OneWeb should see its deployment fast enough in order to rebalance the loss created by internet inequality.

One of the many issues the global pandemic has exposed is the (mostly) single point of failure in global supply chain systems. Over the last couple of decades global corporations optimized their supply chain to such an extent that it became far from flexible, while being extremely optimized. While the supply chain disruptions in today’s products - from face masks to medical equipments got exposed, this post is about the future supply chain challenges.

With many countries rallying towards fighting climate change, electric vehicles are at the forefront of that fight. With the cost per kilowatt-hour touching $100/kWhr or even less, electric vehicles would not only make economic sense but also reduce the total cost of ownership drastically. While American companies like Tesla are at the head of the pack in innovation and execution, the components for uninterrupted production as well as security lie elsewhere - the supply chain for electric car batteries.

As Benchmark’s Simon Mores says there are two ways of reducing cost of Li-ion batteries:

1. Controlling the supply of raw materials that go into batteries
2. Changing how batteries are made (scale)

While the prices of raw materials were at all time high, companies like Tesla maintain lead by controlling the manufacturing process - “the scale” as 25-40% of the cost of an electric vehicle can be attributed to the cost of the battery. That is where Gigafactories become interesting and so does controlling the supply chain. But so far, American auto companies, to a large extent, have only been assembling the components that constitute the battery. There is a lot of effort that goes into a battery before this final assembly.

Let us break this apart looking at the value chain of electric battery manufacturing:

If one were to superimpose the geographies that control different parts of the value chain, one could immediately see the premonitions of future supply chain challenges.

More than 50% of every one of the raw materials is dependent on one geography, which is a recipe for future disruptions. As Benchmark’s Vivas Kumar points out, the only part of the value chain that can’t be moved around are the mines themselves. Every other part of the value chain can be located in different parts of the globe. Most of the early stages of the value chain like Cathode, cell manufacturing, etc. are controlled predominantly by specific geographies that would be a supply chain and security risk when electric vehicles become like mobile phones. A wise thing for such an industry like Auto to do is to shore up the different components so that the critical component (battery) supply is not disrupted in the future.

One has to wait and see if the electric battery supply chain gets reshored and does not become another disruption waiting to happen or if short term optimization goals will lead to long-term disasters, just as it is happening in semiconductors and electronics.

In a previous post we saw the challenges in accessing quality education digitally. In this post, we will look at some of the disparate pieces of technology that could work together to make online education less expensive, more efficient and hopefully more accessible.

From a 45000 ft. view, online education is just about a video playing in a TV or a computer browser, much like any TV program or YouTube or Vimeo video would. But there is a lot more to it as you go down into the details. Even in its simplest form (i.e. no authentication, account validation, etc.,) most of them contain video and audio streams, which itself is the first bottleneck for places which have poor internet bandwidths, if at all internet exists.

So, the very first technical problem is to address bandwidth the way several streaming services address them - by encoding them in different qualities and formats, cache them in CDNs or edge networks efficiently, predict their access patterns and serve them efficiently. This alone may not be enough - for most of the world the only access to internet may be through a not-so-good mobile data network which might still be prohibitively expensive to stream high quality videos.

This brings us to the second problem: video quality. Since most educational videos, unlike action movie frames, are to a large extent relatively slow moving and not lots of dynamic pixel changes, AI techniques used for Super Resolution can be employed to create a higher resolution content on device even if the incoming videos from the nearest edge network are low resolution ones. But what about the device - does it need to be one with an expensive GPU attached to it? Not really.

That brings us to the third problem: device specifications. The most desired characteristic of the device would be longer battery life and quick recharge capability. Since most education content (the content itself) is generated before the class (the event) and the content is just delivered live, modest processing power should be enough to do the inference side of super resolution. Chromebooks and android devices are a first step in this direction, but a main area of improvement would also be how long the device lasts.

Most students are visual learners but most educational content is dry. One could use components of game development engines to develop tools for teachers so that they can create interesting animations and content for their class without having to learn and program using game engines. We have been living with Powerpoint for content creation for more than three decades and may be it is time to add more interesting content creation capabilities to Powerpoint or move to an application that provides such capabilities.

Most digital content is created in English, Mandarin and Spanish while there are many other languages and many research suggest that one’s first language is the best medium for early education.

While translating languages from one to another automatically is not effective as of today, future AI technologies would hopefully be able to bridge this gap as well.

Students with special needs would require more help both from technology and human beings in order to enjoy and benefit from online education. If technology can leave more time for teachers they could invest that gained time on students with special needs.

There are several other problems where technology could play a role. One could create authentication mechanisms for examinations and awarding certificate of completion of courses. Zoom like video conferencing technologies could be used both for delivering the content as well as break rooms for further guidance. Automated algorithms could test basics of the students’ knowledge of a topic and have progressively challenging exercises, rather than a one-size-fits-all teaching and evaluation model that is currently in place. A content created by one teacher or school can be securely shared with another school, optimizing the time and effort put in for content creation. The content itself could be stored in cloud storage and personalization algorithms could recommend what other YouTube or other content would be helpful in understanding the concepts in a specific topic, instead of the students straying all over the web looking for explanations to fortify their understanding.

There are lots of ways things could improve. But in reality, most such softwares are fragmented. One might use a chromebook while the slides sent out by the teacher would be in powerpoint format, one video would be on YouTube, one would be lying as an mpeg file in a shared folder. One would use zoom and the other Webex or Teams and there is no way these applications are play well together. Further, the current techniques for first party and third party integrations of content creation and marketing might bump up content rich in marketing dollars while relatively punishing content rich in content. Nevertheless, technology would play a huge role not only in school education but also enterprise reskilling (education technology would be an enterprise product as well) and the one who wins would be the one who builds the strongest platform combining the disparate technologies listed above, and some more.