Show of 05-07-2022

Tech Talk

May 7, 2022

Email and Forum Questions

  • Email from A in Smithsburg: Dear Tech Gurus. This popped up in one of my security feeds. While I just can’t wrap my head around the concept of “owning” imaginary “property”, the idea that it can be stolen really gives me a headache. However, this raises the question, what’s the difference between one Bitcoin and 12,000 acres in Metaslovakia? (Other than the Bitcoin is relatively easy to spend… at least now…) On a side note, going back a couple of weeks, I loved Andrew’s Dean Dean observation… back when I was a student at UT, we had Dean Martin (title) … and thought you would enjoy hearing that Dougherty Hall, where Dean Dean would have had his office is flanked by Ferris and Buehler Halls. (Hard to believe that got past the planning stage!) Your faithful listener, Al Metzel from Smithsburg, MD (very small town, but our guys answer Camp David’s 911 calls)
  • Tech Talk Responds: The key is the block chain, not the gas that pays for the validation. A blockchain could record the deed for 12, 000 acres in Metaslovaikia and make it impossible for a corrupt official to steal it because the ledger where it is recorded is public and immutable. The magic is in the blockchain. The trick is giving a string of 0s and 1s to entice someone to validate the blockchain. Brilliant.
  • Email from James Messick: Andrew, I just wanted to tell you that I think you have been doing a great job as the new co-host of Tech Talk Radio. I know you had big shoes to fill, but I think you are doing an excellent job. But I am glad that Mr Big Voice stayed on staff. James Messick, Kernersville, NC
  • Tech Talk Responds: Thanks for the feedback. I agree with you.
  • Email from Jim in Bowie: Dear Tech Talk.  I just bought a Roku box I have the choice of connecting it to the Internet over Wi-Fi or with an Ethernet cable. The Roku box is close enough to the router to connect them with a cable and the Wi-Fi signal is strong there. My question: Is there an advantage to using one method over the other? Jim in Bowie, MD
  • Tech Talk Responds: My rule of thumb for connecting devices to a router is to ALWAYS connect the device via a wired Ethernet connection if that is an practical optio. In most cases a wired connection is going to be faster than Wi-Fi, and it’s virtually guaranteed to be more stable. A wired Ethernet connection should help prevent buffering and temporary streaming interruptions due to Wi-Fi signal loss. There’s also another advantage to keeping streaming devices off your Wi-Fi network whenever possible. Streaming video requires lots of bandwidth, so connecting your Roku via a wired Ethernet connection will free up all of your Wi-Fi bandwidth for your smartphones, laptops, tablets and other devices that really need to connect via Wi-Fi.
  • Email from Ashley in Virginia Beach: Dear Tech Talk. I use Chrome as my main browser and it tends to load pages that have many photos at a snail’s pace. I know this is a Chrome issue because the same pages load a lot faster in Edge and Firefox. How can I make Chrome load photo-heavy pages faster than it does now? It is my favorite browser by far, but this is driving me nuts! Ashley in Virginia Beach
  • Tech Talk Responds: This is a known issue with Chrome and there’s a simple solution that seems to work for most people.
    • With the Chrome browser open, type (or copy and paste) the following into the address bar: chrome://flags/#enable-gpu-rasterization
    • If the GPU Rasterization feature is available on your platform, toggle its state to Enabled.
    • Exit Chrome and then relaunch it.
    • Now visit one of the pages that was loading slowly before and see if it loads faster now. My guess is it will.
  • Chrome 37 introduced a GPU rasterizer. When enabled, some paint workloads can go from 100ms/frame to 4-5ms/frame. With GPU rasterization, part of the workload is moved from the CPU to the GPU.
  • Email from Tom in White Stone: Dear Tech Talk. I am renewing my Internet package this month and have my options from budget options to the Gigabit for a premium. How much download speed do I actually need? Tom in Whitestone, VA
  • Tech Talk Responds: To answer that question, let’s look at some common activities, ranked from the least demanding to the most demanding.
    • Email                              1 Mbps
    • Music Streaming            2 Mbps
    • General Web Browsing  3 Mbps
    • Social Media                  5 Mbps
    • Online Gaming               5 Mbps
    • Video Conferencing       5 Mbps
    • HD Video Streaming      5 Mbps
    • 4K Video Streaming       15 Mbps
  • It might come as a surprise to many people, but when you look at individual internet activities they simply aren’t that demanding. Low bandwidth activities like using email (or any other text-based communication like chatting), streaming music, or just browsing around the web searching for things or reading posts on your favorite forum, just don’t use that much bandwidth.
  • With a gigabit fiber connection, you could likely stream 4K video to TVs in every single room in your house, plus all the handheld devices, and still have some bandwidth to spare. It is important to emphasize that more bandwidth doesn’t make a less-demanding bandwidth activity better.
  • For the vast majority of people, purchasing the highest tier of internet service available to them, especially if that’s gigabit speeds, is overkill.
  • Email from Donna in Pittsburg: Dear Tech Talk. I just got a new Internet connection and my download speed is much slower than advertised. What might be causing this discrepancy and can I fix it. Donna in Pittsburg, Kansas
  • Tech Talk Responds: The router’s advertised speed is theoretical. The speeds advertised on the box and in the documentation for a particular router are the theoretical maximum speed the router can sustain under perfect conditions and when paired with an equal or better test device in a lab.
  • When you’re using your iPhone, Xbox One, or whatever device on your Wi-Fi network, you are limited to the connection that device has negotiated with the Wi-Fi router. On that AC1900 router, for instance, the bandwidth is divided between a single 2.4Ghz band that maxes out at a theoretical 600 Mbps and a 5 GHz band that maxes out at 1300 Mbps. Your device will be either on one band or the other, and can’t take advantage of the full capacity of the router.
  • Because of overhead in the Wi-Fi protocol, you can expect anywhere from 50-80% of the expected “advertised” speed based on your gear. Newer routers paired with newer devices are more efficient, older devices and older routers, are less so.
  •  

Profiles in IT: Sebastian Thrun

  • Sebastian Thrun  is a German-American entrepreneur, educator, and computer scientist, best known for autonomous vehicles and founder of Udacity and Google X.
  • Thrun was born May 14, 1967 in Solingen, Germany.
  • Thrun was a geeky kid, spending much of his free time in libraries or in front of a NorthStar Horizon home computer, on which he tried to write software programs to solve puzzles and play solitaire.
  • He completed his Vordiplom (intermediate examination) in computer science, economics, and medicine at the University of Hildesheim in 1988.
  • As a lonely undergraduate, Thrun thrust himself into trying to understand people better, dabbling in psychology, economics, and medicine.
  • At the University of Bonn, he completed a Diplom (first degree) in 1993 and a Ph.D. (summa cum laude) in 1995 in computer science and statistics.
  • In 1995 he joined the Computer Science Department at Carnegie Mellon University (CMU) as a research computer scientist.
  • In 1998 he became an assistant professor and co-director of the Robot Learning Laboratory at CMU.
  • As a faculty member at CMU, he co-founded the Master’s Program in Automated Learning and Discovery, which later would become a Ph.D. program.
  • Thrun left CMU in July 2003 to become an associate professor at Stanford University and was appointed as the director of SAIL in January 2004.
  • Thrun led development of the robotic vehicle Stanley which won the 2005 DARPA Grand Challenge, and which has since been placed on exhibit in the Smithsonian Institution’s National Museum of American History.
  • His team also developed a vehicle called Junior, which placed second at the DARPA Grand Challenge in 2007.
  • From 2007–2011, Thrun was a full professor of computer science and electrical engineering at Stanford.
  • The fateful lecture. He heard Sal Khan talk about the Khan Academy at TED.
  • In 2011, he sits down in his living room with an inexpensive digital camera and starts teaching Introduction to Artificial Intelligence.
  • Some 160,000 people sign up from more than 190 countries. None of the top 400 students goes to Stanford. The experiment starts to look like something more.
  • In 2012, he co-founded an online private educational organization, Udacity.
  • On April 1, 2011, Thrun relinquished his tenure at Stanford to join Google.
  • He was a Google VP and Fellow, and worked on development of the Google driverless car system. He focused on this because of the many traffic deaths.
  • Thrun developed a number of autonomous robotic systems that earned him international recognition. He always solved real problems and made them personal.
  • In 1994, he started the University of Bonn’s Rhino project. RHINO is a mobile robot designed for indoor navigation and manipulation tasks.
  • In 1997 Thrun and his colleagues Wolfram Burgard and Dieter Fox developed the world’s first robotic tour guide in the Deutsches Museum Bonn.
  • In 1998, the follow-up robot “Minerva” was installed in the Smithsonian’s National Museum of American History in Washington, DC.
  • Thrun went on to found the CMU/Pitt Nursebot project, which fielded an interactive humanoid robot in a nursing home near Pittsburgh, Pennsylvania.
  • In 2002, Thrun helped develop mine mapping robots in a project with his colleagues.
  • After his move to Stanford University in 2003, he engaged in the development of the robot Stanley, which in 2005 won the DARPA Grand Challenge.
  • In 2007, Thrun’s robot “Junior” won second place in the 2007 DARPA Urban Challenge.
  • Thrun joined Google as part of a sabbatical, together with several Stanford students. At Google, he co-developed Google Street View.
  • He contributed to the area of probabilistic robotics, a field that marries statistics and robotics. Probabilistic techniques have since become mainstream in robotics, and are used in numerous commercial applications.
  • In the fall of 2005, Thrun published a textbook entitled Probabilistic Robotics together with his long-term co-workers Dieter Fox and Wolfram Burgard.
  • He is currently CEO of Kitty Hawk, one of the pioneers in personal air taxis launched by Google co-founder Larry Page. This is a better option for transport.
  • He has run half a dozen marathons; he snowboards; he kite-surfs; and he is an avid road cyclist.

Observations from the Faculty Lounge

  • Innovation according to Sebastian Thrun
  • Innovation by itself is an experimentation and learning process. Yu have to embrace the future with uncertainty. Failure is your friend, because you are forced to learn something new, and it strengthens you in the ability to make good decisions the first time around. It is the only way forward. When you fail, it forces you to do something no one else has done before.
  • Sebastian Thrun says there are two types of people: type A and type B.
  • Type A is the expert, the one who knows everything. They come to you with a plan with all the ideas and details worked out. They want to make sure that everybody participates in a consensus of what they are doing. He avoids these people.
  • Type B however, you rarely see often. They have a vision, but are honest about the predicament and usually admit they have no clue how they will reach their end goal. This person has the strongest vision, and knows they will fail, but does not mind failure.
  • A team of 12 engineers, for instance, built Google’s driverless car. The team that was not afraid to break all the rules. The more people you have on a team that compromises on a situation, you will have a hard time breaking any rules.

A New Approach to AI: Thinking with Analogies?

  • Suggested by Bob from Maryland
  • Melanie Mitchell, an AI researcher, says AI never truly be like ours until it can make analogies.
  • She thinks machines need to be able to make good analogies before they can approach humanlike artificial intelligence.
  • She spent six years collaborating closely on Copycat, a computer program which, in the words of its co-creators, was designed to “discover insightful analogies, and to do so in a psychologically realistic way.”
  • The analogies Copycat came up with were between simple patterns of letters, akin to the analogies on standardized tests.
  • Mitchell says making analogies, allowing AI systems to apply existing knowledge to new problems, will help them truly understand the data they’re manipulating.
  • For example, we want self-driving cars, but one of the problems is that if they face some situation that is just slightly distant from what they have been trained on they do not know what to do.
  • How do we humans know what to do in situations we have not encountered before?
  • Well, we use analogies to previous experience.
  • There are new approaches, like meta-learning, where the machines “learn to learn” better. Or self-supervised learning, where systems like GPT-3 learn to fill in a sentence with one of the words missing, which lets it generate language very, very convincingly.
  • Deep neural networks are great at certain tasks, but that’s not enough for Mitchell.
  • One of the theories of why humans have this particular kind of intelligence is that it’s because we’re so social. One of the most important things for you to do is to model what other people are thinking, understand their goals and predict what they’re going to do.
  • And that’s something you do by analogy to yourself. You can put yourself in the other person’s position and kind of map your own mind onto theirs. This “theory of mind” is something that people in AI talk about all the time. It’s essentially a way of making an analogy.
  • Does that mean that for an AI to make these kinds of analogies, it also needs a body like we have? That is the million-dollar question.
  • Mitchell’s intuition is that yes, we will not be able to get to humanlike analogy [in AI] without some kind of embodiment.