I’d like to introduce you to your chauffer for the day. His name is Mobli, and he comes from Israel. Unlike most Israeli drivers you may have had the pleasure of experiencing, he is very quiet, and he drives quite smoothly. You will not be engaged in lengthy debates about politics or religion, you will not be asked how old you are, why you are not married yet, or how much money you make. You will not find yourself careening around traffic circles, speeding up into oncoming traffic, or lurching to a stop so sharp that your rugelach fly out of their plastic bag and stick themselves to the windshield. Mobli is a whole new breed of Israeli driver….
The reason Mobli is so well behaved is because he doesn’t even sit in the cabin with you. Mobli sits quietly under the hood of your car, and he fits in a plastic box about the size of a school lunch box. Mobli is a computer, his arms are wires snaking all around the hidden parts of your car, his fingers, sensors of all kinds, and his eyes are an array of radars and lasers mounted around your car. He thinks faster than any human driver, calculating his driving patters over one hundred times a second. He is made by a company called MobilEye (Mobl-i, you get it?), and he is already installed in a fleet of Audis tooling around the US and Israel.
Mobli represents the future driver of not only your car, but the driver of trucks, buses, and trains as well. I welcome the day that Mobli and his computer friends start driving our cars, they will be far more safe and efficient then our current driving population; sleep deprived, coffee swilling, texting, shaving, makeup applying human beings.
But what really makes Mobli a great driver is not what he does while he’s driving you around, it’s what he does in his spare time, while you’re in the office, at home, or out for dinner. Mobli is not just one computer sitting in your car, Mobli is part of a network of all the computer drivers made by MobileEye, as well as the massive supercomputers in MobilEye’s headquarters, and every day, all day, they are running simulations of millions of scenarios to see how to best drive through them.
If human engineers had to tell the computers what to do in every possible driving scenario, it would take them centuries. There are so many billions of different possible scenarios; driver speeding up in the left lane with pedestrian crossing on the right, skidding left on icy road toward semi-trailer, human driver running a stop sign into intersection in front, two lanes merging onto highway in heavy traffic during thunderstorm, the list is endless.
Instead of humans trying to tell the computer what to do in each scenario, the humans simply tell the computer what outcomes they are looking for and they let the computer run through billions of simulated scenarios and learn from what happens. The Mobli of today is exponentially smarter than the Mobli of five years ago, and five years from now, when I believe we may actually get to use him as our driver, he will be exponentially smarter than today.
This is the beauty of machine learning. It never forgets a lesson, it incorporates all that it learns into its future driving immediately, and it tests scenario after scenario tirelessly. We humans learn from experience; a forty eight year old driver with thirty years and a million miles under his belt is far better than a sixteen year old with a newly minted license, and their auto insurance rates reflect that. But a Mobli driver has the experience of billions if not trillions of miles driven, both in the real world from all the Mobli cars on the road, and in the simulations constantly being run by the supercomputers.
The idea of building knowledge on previous experience, called reinforcement learning, is nothing new. Edward Thorndike, a psychologist, demonstrated it years 100 ago in a series of experiments with cats. The cats were placed in a box, and the only way for them to get out was by stepping on a lever. The first time they were placed in the box, they paced, they meowed, they snarled and eventually by mistake they stepped on the lever. The second time they were placed in a similar box, they were much quicker to step on the lever, and a few times later, they would simply walk over the lever and press it immediately upon entering the room. The more they reinforced their learning of the trick to the room, the faster they got at activating it.
For decades, people have been trying to do this with computers. In 1992, a researcher at IBM invented a computer program that taught itself how to play backgammon, becoming better and better until it eventually beat the best human players. But chess, with far more possible moves and outcomes was too great of a challenge for computers. That challenge ended in 1997 when the Deep Blue computer beat chess master Garry Kasparov. The world then turned to Go, an ancient Chinese game with simple rules but far more possible moves than chess. For years, humans had the edge, until last year when Google Deep Mind’s Alpha Go computer beat the world champ, Lee Sedol. The Alpha Go computer used reinforcement learning techniques, playing billions of simulated games, and constantly learning from its errors.
We humans have been using reinforced learning for as long as we have been around. The hockey player who is skating down the ice at incredible speeds, expertly shifting the puck along with his stick, and then with the flick of a wrist shooting it into a tiny opening between the goalie and the posts does it almost without thinking, his brain relying on the thousands of times he’s done it in practice. The court stenographer, typing furiously at 250 words per minute, is not really thinking about what she is doing; her hands fly across the keyboard with decades of reinforced learning pushing them. My five year old may struggle to read simple words like cat, hat, bat, and mat, while I can read a college textbook with ease.
In the 1970s, Noel Burch, an employee at Gordon Training International described human learning as The Four Stages of Competence. Let’s take reading and use it as an example. The first step is Unconscious Incompetence; I don’t understand how reading works, and I can’t do it. Then I sit down in a classroom and learn all about how to read; what sounds the various letters make, what to do when I get to a comma or a period, the difference between hard vowels and soft vowels, etc. Now I’m in the stage called Conscious Incompetence; I understand reading, but I can’t really do it. Then I get some books and start working. I keep thinking of everything I learned, and I’m applying it, progress is slow, but I can crawl through The Cat in the Hat. Eventually, I get to the highest level, Unconscious Competence. I can read very well, and I never really think about it, it’s as natural to me as breathing.
Humans get to Unconscious Competence because we have neural pathways that get strengthened by positive experiences and fade away when not used for a long time, hence we have reinforced learning. For years, computers didn’t have that ability, they simply followed the instructions being programmed into them. Only in the last few decades have computers begun using reinforced learning, and the progress they have shown is remarkable. Self-driving cars are only the tip of the iceberg. Computers will be able to run billions of simulated tests to discover what chemical formulas can cure various illnesses, computers will be able to perform surgeries with far more precision than humans, and most wondrous and scary of all, computers will be able to create far more efficient computers that can do more things better than us!
While reinforced learning has a lot of positive sides to it, there are areas where reinforced learning takes away from the process. I came into contact with one of those areas a few weeks ago, and now make a concerted effort to roll back my reinforced learning to better my life. Thank G-d, I can read Hebrew fluently. With or without vowels, give me a page of Hebrew text and I can fly through it. My Hebrew reading is at the level of Unconscious Competence. This may be good if I’m trying to read a Hebrew newspaper. It is not good when I’m trying to pray. The goal of prayer is not to be efficient, it’s not to move unconsciously through the text; the goal of prayer is to be Consciously Connecting.
As a matter of fact, when prayer becomes something that we do with rote, it starts to lose its meaning, and starts becoming boring or burdensome. There is a prayer called Tachnun, which is a deep supplication to G-d to help us through our moral struggles. On most weekdays it’s only a few paragraphs long, but on Monday and Thursday it is about three pages long. I found myself disliking Monday and Thursday services because the Tachnun was soooo long, you just sit there and read for seven minutes straight paragraph after paragraph.
I was bothered by the fact that I really didn’t like those prayers, so I decided it was time to roll back my reinforced learning and reading competency, and instead of being quick and efficient, I would be slow and deliberate. I may not be able to finish all of the prayers in time with the rest of the congregation, but I would really focus on them. A wondrous thing happened. I slowed down my reading to a crawl, but suddenly started perceiving much more, concentrating on the words, thinking about their beautiful meaning, and feeling those meanings permeating my soul. It has been a few weeks since I started doing this, and now I actually look forward to Monday and Thursday prayers, when I step back and let all that feeling wash over me.
The same goes for connecting with your children, your spouse or a friend. If you are looking to efficiently carry out your duties as a friend, spouse, or parent, you are bound to be frustrated. The kids will never be ready in time, no matter how many systems you put in place. You can’t share your day with your wife, and properly hear about her day in five minutes, and you friend who’s going through a tough time can’t tell you all that’s going on in his life in ten minutes. You need to slow it all down, if you really want to really be a parent, spouse or friend.
Reinforced learning is great for many things, especially things that need efficiency, but efficiency and emotion don’t work well together. Emotion is the feelings that you feel at a particular time, efficiency is trying to do things in less time. More efficiency often means less feeling. That’s great for surgery or driving my cars, feelings should have nothing to do with navigating a road safely, or avoiding cutting the wrong areas in a brain surgery. It’s not great for human relationships, whether they are with other humans, or with the Divine. Human relationships are built on emotions, and emotions take time to come out.
The world will be a safer place when Mobli drives us to work every day. But the world will be a better place if we start taking the extra time that efficient technologies free up for us, to slow down. To say a prayer with consciousness and feeling, to call a friend and really listen to them, to take a walk with our teenager, to live life with Conscious Competence.
Parsha Dvar Torah
In this week’s Torah portion, Terumah, we read about the construction of the Tabernacle and most of the vessels that were in it. We find an interesting phrase by the instructions for the menorah. When describing the making of the menorah the Torah tells us, “You shall make a Menorah of pure gold, hammered out shall the menorah be made” (Exodus 25:31). The commentators ask why it says “shall the Menorah be made” instead of saying, “shall you make the Menorah” which indicates that it was not made by Moshe?
The Medrash tells us that Moshe had a difficult time understanding how to make the Menorah which was extremely complex and had all sorts of decorative cups, knobs, and flowers. So G-d showed him an image of it in fire, and he still couldn’t figure out how to copy it. Finally g-d told Moshe to simply throw the gold ingot into the fire, and miraculously the Menorah emerged fully made. Hence, the menorah was made, although not by Moshe. But this leaves us with a question. Once G-d saw that Moshe could not make the Menorah unaided, why did he make him work on it again with a fiery image without success, and only after that allow him to throw it into the fire? Why didn’t G-d save Moshe the bother and the failure, and simply make it miraculously first?
The Sfas Emes, the second Rebbe in the Gerrer dynasty (1847-1905, Poland) answers this question with a fundamental lesson. Many times we are faced with a challenge that is simply too difficult. We struggle and struggle and then suddenly we have a Eureka! moment and it all works out. That is G-d giving us a hand from above, pushing us across the finish line. However that only happens if we force ourselves to the limit of our capacity. If we quit early, we will miss that final push from above, and we may never make it through the finish line.
That is what happened with Moshe and the Menorah. First he tried, and it didn’t work. G-d gave him another angle, and he had to work at it. Finally after Moshe tried his hardest every which way, he was given the big push that got the job done with ease.
It is empowering to know that this is the way G-d operates with us. Sometimes we work really hard on a project, on fixing a character flaw, on bettering a relationship etc. and we hit a point that we’re ready to give up. But now we know that if we just push a bit harder, and really max out our effort in this area, we may just get that big push from above that will push us across the finish line!
In this week’s portion G-d asks the Jewish people to build a physical dwelling place for the Divine Presence. The Sages tell us that the real goal is that we each build a Tabernacle inside ourselves, but that the building is the physical expression of that idea, and one we can relate to much more easily. The Jews were asked to donate the many different materials with which the Mishkan (the Tabernacle), its vessels, and the holy vestments for the Kohanim would be made.
The items the Jews were asked to bring were: gold, silver and copper, turquoise, purple, and crimson wool, fine linen, goat’s hair, red-dyed ram’s skins, tachash skins, acacia wood, oil, spices, and precious stones. G-d tells Moshe that He will show him a model of the Tabernacle and that the real one should be built exactly like the prototype.
After that, the Torah begins to detail the design of many of the vessels. The ark was made of three boxes, the outside and inside ones of gold, and the middle one of wood. On top of the box was a special lid that had two childlike forms with wings engraved onto it. There were four rings in which poles to carry the aron were placed and, specifically regarding the ark, the Torah stipulates that the poles were never to be removed.
The Table was a vessel used to hold twelve loaves of showbread that were placed there for a week at a time, from Shabbos to Shabbos. The table was made of gold-plated wood and had a small crown-like ornament rimming it. It had a special system of poles and supports so that the showbreads could be held up properly.
The Menorah had to be carved out of one block of gold. It was about 70 inches tall and had one central mast with three branches leading off to each side. It was heavily adorned with sculpted flowers, knobs, and decorative cups.
The building itself was made of dozens of wood planks covered in gold and held in place by silver sockets. There were also gold plated wooden bars that held them together. There were two heavy tapestries covering these planks. The inner one was made of twisted linen woven with turquoise, purple, and scarlet wool and was held together with golden hooks. The outer one was made of a more simple material, woven goat’s hair, and was held together with copper hooks. The Sages tell us that this teaches us that a person’s home should always be more beautiful on the inside than on the outside. (Please note: There are so many lessons taught from everything in the Tabernacle, but space doesn’t permit me to list all of them. However, please discover these gems for yourselves!)
The altar was a hollow rectangular cuboid (the width and length were the same, the height was not) made of wood and covered with copper. It was filled with dirt. It had protrusions at each of the top corners that were exact cubes, netting surrounding it like a belt, and a protrusion in the middle that was large enough to walk on. Leading up to it was a long ramp, as no steps were allowed on the altar (see the end of Parshas Yisro).
Finally, the courtyard was swathed in a white linen sheet which was held in place by wooden pillars with copper sockets. The pillars had bands of silver going around them, and they held up the material with silver hooks. If it sounds like a beautiful place, that’s because it was one. May we all merit to see the rebuilding of the Temple, and may we once again have a place on earth where G-d’s Presence can reside in all of its Glory!!!
Quote of the Week: Action will remove the doubts that theory cannot solve. ~Tehyi Hsieh
Random Fact of the Week: The S.S. in a ship’s name stands for “steamship.”
Funny Line of the Week: His mother should have thrown him away and kept the stork!
Have a Swell Shabbos,
R’ Leiby Burnham