Homenaje a Juan del Gastor

The first time I met Juan del Gastor was at Carl’s home in Berkeley. This must have been in 2003 or 2004. I had booked a private lesson with him to learn Flamenco guitar. My flamenco mentor, Kenny, had really played him up, and told me that this was the guy to learn from, so I had high expectations.

I did not speak any Spanish, and Juan did not speak any English, so we got off to a great start. Each of us mumbled something in our own respective languages, but he said something that made a big impression on me. He said something that included the words sangre (blood) and mismo (the same). And then he put his arm next to mine. And I understood right away what he meant.

You see, Juan is authentic gypsy. Gypsies defined flamenco over the years of nomadic life: the gitano in canté gitano – flamenco –  means ‘gypsy’. And the gypsies’ origin is in India, where I am from as well. When he put his arm next to mine, our arms looked the same. And that is what he was saying – we are of the same blood. We are the same.

Juan’s real name is Juan Gomez Amaya. He is known as Juan del Gastor because of his uncle, the flamenco legend Diego del Gastor (1908-1973). After Diego died on July 7, 1973, his guitar legacy was carried on by the four sobrinos (nephews) – Diego de Morón, Austin Ríos Amaya, and the brothers Paco and Juan del Gastor.

Over the years, I learnt a lot from Juan. He taught me many of Diego’s falsetas. Every time he came to the US, there would be some sort of event, a fiesta, or a music program, and I would take lessons from him. But every time, it was a remembrance of Diego, and Juan was the man who would deliver Diego’s legacy. I recall thinking that living always in Diego’s long shadow must have been tough for Juan. On the one hand, being associated with the legend had brought him a lot of good things he would not have got otherwise – a following in the US, some fame, and probably a lot more money than he would have made selling fish in Sevilla. But any time anyone said anything about Juan, it was always as Diego’s this, Diego’s that ….

At the end of 2006, Kenny stopped coming down to the South Bay for his lessons and story sessions (more about that in a future blog). And my flamenco guitar playing started unravelling. I did travel to Sevilla in 2007, for the Fería de Abril, and took many lessons from Juan during that trip. I also met his wife Luci, who took me to various peñas unknown to the tourists, for some authentic gypsy flamenco. But my involvement kept reducing, and in 2010, I stopped playing the guitar completely, focusing on Argentine Tango instead. I could do tango at night, after my family activities were done.

And so, I did not meet Juan any more for many years.

Recently, as you may know from another blog post of mine, I turned 50. It so happened that the very next day, Juan, along with a bunch of others, including Kenny, were doing a show at La Peña in Berkeley. So of course I had to go. I felt very lucky that the timing was so good. At the concert, I met Luci and we chatted a bit, remembering old times. But Juan and Kenny were in the room at the back where the artists were, and I could not talk to them.

Of course there was the photo of Diego in the corner, as usual, since this was all in his memory. But I felt this time, something was different. Juan’s singing was different, and at some point in the middle, I noticed something that stood out: he was playing a falseta that was clearly all his own. This was definitely not Diego – not a single note of his in this falseta. I spoke to Luci about it afterwards, I told her the falseta was really pretty, and so different – she told me that Juan gets nervous on stage, and sometimes he just plays anything. She couldn’t really recall anything different or special. I will likely never know what it was, but I felt a strange elation nevertheless. Diego’s shadow was long, but not infinite, and Juan had stepped outside it. And it was beautiful.

Here are some memories of Juan over the years.

Diego’s Homenaje at the Thirsty Bear in 2005 – Juan playing guitar:

Juan dancing with Luci at the homenaje:

And some canté:

With Juan in 2006. I’m looking sheepish because I had a tough lesson.

The famous Spanish band Son de la Frontera performed at the Yerba Buena Center for the Arts in early 2007. After the concert, we all went to Carl’s house in Berkeley and partied through the night until it was time for them to go to the airport in the morning and take their flight back to Madrid. Here’s Juan playing the guitar.

Another shot from that same fiesta:

Juan with my youngest son – for some reason the latter was quite apprehensive of Juan. Probably from the energetic way he played the guitar.

Juan in his home in Sevilla, when I visited there in 2007.

 

And finally, Juan at the concert at La Peña on July 8, 2018. You can see Kenny too, under the ‘L’. It was a fabulous concert – I hope there are many many more!

Turn Fifty in a ’69 Kombi

As luck would have it, my fiftieth birthday fell on a Saturday. My wife planned a trip to Sonoma, where we would tour some wineries in a VW Kombi, and do wine tasting. Back in the ’70s and ’80s, when my parents lived in Kenya and Nigeria, we would go everywhere in these Kombis, better known in the US as the ‘VW Bus’. So, it had sentimental value for me. The idea was to leave the kids at home, and go for a day trip, just the two of us.

I admit, at first I was not sure it was a great idea. Turning fifty is not something I’ve been looking forward to with great excitement. But it slowly grew on me. I love wine, Volkswagens and vinyards. What was there to not like?

We left early in the morning and got to Sonoma exactly at 10:30 am despite a few twists and turns. Luckily, the parking lot we chose to park in turned out to be the place where we were supposed to meet. There was the green Kombi just a year younger than myself, already waiting for us with the driver, Tommy, and one couple (Dani and Logan). There was to be a total of 3 couples including us, so we waited for the last one.

They came within minutes and we set off for the first winery – Liana Estates. It turns out, they are taking entries for a new name, because this one is being disputed by a Liano from Italy, who says the name is too similar to his, which he uses for his own wines.

In the meantime, we started chatting among ourselves. It was instantly obvious that the company we had was just fantastic. Dani and Logan were a young couple from the University of Nevada in Reno. Dani had just graduated with a degree in journalism, and Logan would graduate in the winter. They had met in Italy where they had gone to study as part of their program, and were together for just 2 months. Logan was to turn 23 on the Monday.

Tommy, our host was a young guy just out of high school. But he knows a lot about wine-making and has a friendly and vivacious personality, just the right type for a trip like this.

Rebekah and Nate lived just a little south of Chicago, Champaign, Illinois, where Nate is a teacher, sports coach, and entrepreneur. Rebekah is a wedding photographer, which interested me immensely because it is what I have been thinking of doing for a while recently. We were to have several interesting chats about this along the way. Nate also had his birthday the same day, he had turned forty. Like us, Rebekah and he also had three children, that they had left behind with family, so we had that topic to talk about as well. Strangely enough, on a previous wine tour in Stellenbosch and Franschhoek near Cape Town, we had also met a family with three children and we had many family conversations with them too!

The wines at Liana were the best, but I think after a while we lost track of what we were having. Partly because we were drunk, partly because the company was too interesting, and we were really just very interested in getting to know more about each other and chatting. Here’s a photo of all of us: from left, Rebekah, Nate, yours truly, Gayatri, Dani and Logan.

At Liana, Tommy showed us the fermentation vats and barrels, and explained the wine-making process as well.

We had lunch at the Sixteen 600 Winery tasting house. The lady there explained a bunch of things but I was too drunk at that point to care – I just heard a few words about organic wine, vinyl records and something else. Lunch was a fabulous salad, with cheese bread. That helped a bit with the alcohol.

After that, we set off for the last winery, Buena Vista, which also happens to be California’s oldest premium winery. The tasting room is beautiful, with two floors. The upper floor is a gallery with a straight staircase leading up, next to a lovely chandelier.

We had plenty of time at Buena Vista to sit on some comfortable sofas and carry on our conversations. And watch the stuffed peacocks in the corners. For a while we were also just standing around the wine tasting bar and looking at other people below doing the same thing.

Before the tour ended, we stopped at some vineyards at the side of the road, and took many photos. Here’s one with our green ’69 VW Kombi from West Wine Tours.

And here’s one of Rebekah taking a photo of Dani and Logan with her Canon EOS 5D MkIV clad in red, that she uses for her wedding photography business.

Every good thing has to end, and so it was with our wine tasting tour. Tommy deposited us all in the parking lot, three unbalanced couples tottering their way back to their cars with their purchases.

Gayatri and I decided to stay in downtown Sonoma for a bit till we got back to our senses for the almost 2-hr drive back home. Walking around, we found Figone’s a family-owned business that specializes in olive oil and balsamic. It was the smell that drew us in from the sidewalk, it was so mesmerizing. Frank, the owner told us that he had started a line of body products too, that by itself would sustain his business, but we told him that he needed to keep the oils just because of how the smell would get him more customers. Here he is, with his wife and daughter.

The oil tasting experience was new for us. Frank gave us a lot of different types of oils and balsamic in these little cups. Frankly, I would have liked it better with some bread. But the concoctions are quite tasty, we had a lot of fun trying so many.

The drive home was anti-climactic but uneventful. We got home to this lovely cake selected by my kids. I cut it the next morning with my proud and happy parents looking on and cheering via Facetime video.

I don’t know whether I should be happy or not having got to 50. I reflect a lot these days on how I have lived my life so far, and what I should do in the years I have left. But I’m eternally grateful for the wonderful family and friends I have, the irreplaceable experiences, and the chances I get to meet so many new and fabulous people all the time, from all over the world. I couldn’t ask for more.

En una Milonga, hay que Bailar, no hay que Practicar and other El Flaco Dany Memories

In February 2010, Almirante and I visited Buenos Aires, Argentina, and Montevideo and Colonia in Uruguay. He had been there before and I was fascinated by the tango stories. I had just started dancing tango and wanted to see what it was really like.

In Buenos Aires, we stayed very close to Salon Canning, which is well known for its tango milongas, so we went there a few times. On one of those occasions, I met an Argentinian guy who spent some time talking to me. He pointed out to me an old man, sitting at a nearby table. He told me, this is El Flaco Dany – he is a famous milonguero. I was sufficiently awed by the man that I took a few photos to take back with me as memories.

Fast forward 7 years, and to my surprise one day, I saw a Facebook post that said El Flaco Dany was coming to teach workshops at Gustavo and Jesica’s in San Mateo! Needless to say, I signed up for and did these workshops, and got to know this wonderful man. The world is indeed a small place!

 

I’m eternally grateful to the Argentinian man I met at Salon Canning. He was a serious tanguero. While he was chatting with me, Almirante was having a conversation with a Brazilian couple who was also sitting at our table. The lady told Almirante how she had been trying some steps she learnt in class in the milonga. She encouraged him to try it too. Immediately the Argentinian guy stiffened significantly, leaned forward and told her: “En una milonga, hay que bailar, ¡no hay que practicar!” – in a milonga, you must dance, not practice!

Machine Learning and Job Loss

Machine Learning (and related Deep Learning, Artificial Intelligence) are all in the news these days. Every enterprise and every person using any form of technology will very soon encounter some form of machine learning, even if they don’t realize it. Examples of this are customer service bots, automation bots, automated phone services, self-driving shuttles or seemingly cute information-gathering robots at malls.

This is because we have become rather good at accumulating vast amounts of data, but are not that good at processing it. Even something simple as figuring out what data is useful and what is not, is in reality very difficult. Well, this is exactly what machine learning techniques are good at, so it makes sense to employ it where humans would not be very efficient.

That inevitably brings about the conversation around job loss. Many people fear that ML/AI will destroy large numbers of jobs, and create employment problems. Others argue against this, saying that in fact this technology will create more jobs. Here are some job loss possibilities:

  • Taxi/train/shuttle/truck drivers, pilots
  • Store check-out personnel
  • Mail/parcel delivery personnel
  • Furniture carpenters

Proponents of ML/AI say (with a big grain of truth) that this is nothing new. Throughout the history of mankind, there have been disruptive innovations that brought in increased levels of automation, and jobs were lost. Telephone operators, farmers, horse-cart drivers and many others have seen their job numbers decimated by technology. This is normal – what we do is move on, learn new skills required by the new technologies, and are better off for that.

This has worked well for thousands of years. But it seems to me that now, we are approaching a point rather quickly, where this cycle is showing signs of strain. Gaining new skills is a good thing – it typically results in better paying jobs, an improved lifestyle, and improvements to the economy. But what if we come to a point where the technology changes fast enough that there are vast sections of the human population that cannot gain skills fast enough to stay relevant? What if there are enough people that do not want to put in the effort to gain new skills, and are happy with the work they are doing that is now being made obsolete by new technologies such as machine learning?

What do we do in this case? In fact, I think we are already here – that the pace of technology is already at a point that the majority of people are unable to keep up with it. This is why we see political uncertainty in many countries around the world, including the US.

What is the solution for this problem? Some would say that we should stop creating technology for technology’s sake. We should stop the adoption of machine learning techniques, to be specific to the topic on hand.

Is this realistic? Can we as a population actually stop the march of what we consider progress? I think not. I may be wrong here, but I feel the momentum of technology is too strong, and not enough people feel the reason to stop it, including myself.

The correct approach should be to recognize that there will be many different types of people, including those that either cannot, or do not want, to significantly improve their skills to stay relevant in the new world after a technology disruption. And, part of the solution is to create ways for them to stay relevant in this new world, and still contribute.

So, can we use machine learning to create new jobs, ones that will keep people relevant without having to invest significant effort in obtaining new skills? This is the question I would like to put out to the community at large. I can think of a few to kick it off:

  1. Furniture / decorative item design: ML can be used to improve on hand-made craftsmanship for materials like woods, metals, etc. This could enable artists to create real-world utilitarian items that are also beautiful and cheap.
  2. Data labeling / preparation: For supervised learning, the data used to train models must be labeled. This is often a tedious and manual operation that highly-trained data scientists should not be doing.

The big problem is, who would be interested in doing this? Is there enough incentive for business to invest in people? Probably not. Should government play a role? Also probably not. Then what? This is the reason I wrote this blog post, to hopefully kick off a discussion. I came across an interesting point about how things work in Germany. If a company wants to lay off anyone, they have to give a 1-yr notice. Meaning, layoffs are difficult to do. Maybe this forces businesses to forgo this country – however I don’t see that happening in Germany. So maybe what it does is to force businesses to think about re-deploying personnel when they are not needed in what they are doing. Is this the way forward perhaps? Or something like it?

Website : WordPress = Machine Learning : KuberLab

HTML site

Over 15 years ago, I decided I needed to do a website for my photography hobby. I also wanted to do one for travel. Being a software engineer and a geek, I wrote it in HTML. Websites then were very simple, so what I had wasn’t fancy either,  just a couple of pages with a blurb about me, and a bunch of photos.

CGI-Perl

The next version was fancier – I wanted it to pick up photos automatically from a directory structure, so I wrote a bunch of Perl scripts and had an automated website. If I uploaded photos into the structure in a particular way, the scripts would automatically pick them up and display them.

Website Template

The third version (and partly my current fourth) was much fancier. I purchased a nice template, and customized it by hand. I was still fluent with Emacs and HTML, so it was easy.

WordPress

My current version is still a template, but now I’m not so good with Emacs and HTML any more, so for the parts I cannot manage, such as my blog, I use WordPress. WordPress (or Wix, Hugo, etc.) is what anyone would use these days if you wanted to build and run a website. This has become the norm – people expect slick websites, and nobody wants to actually code them. And WordPress is not just for building, it’s used to manage the website for ever and monitor, update – all those things required for production.

The Relationship to Machine Learning

So, what WordPress did to websites is mostly what KuberLab is doing to Machine Learning (KuberLab does more, but this is a good analogy). In the early days the companies like Google and Facebook, that have already crossed the Machine Learning Application deployment chasm, probably did the equivalent of what I did with my first HTML website. Then of course, they developed tools and frameworks, and now it became like my CGI-Perl website. With the current movement towards democratization of machine learning, there is now a lot of code available via the Kaggle competitions, Github repositories and so on. So, this is like the template website, one can pick up somebody’s code, tinker with it, and use it for oneself. This is also what is happening on the cloud, with AWS, GCP, Azure and others providing many tools, frameworks and templates for building machine learning applications.

Enter KuberLab

KuberLab is like the WordPress of machine learning. If you are an enterprise wanting to deploy a machine learning application, this is what you want. Sure, you could put time and effort, spend lots of money, and build it yourself and perhaps even feel good about it. But that effort doesn’t scale, deviates from your core business, and will not be production grade for everyone. For this, you need KuberLab. KuberLab is the platform that helps enterprises accelerate their adoption of AI applications. See http://www.kuberlab.com, or try it out at https://go.kuberlab.io

What Makes Reggae, well, *Reggae*?

I love Reggae music. It’s my absolute favourite type of music. There’s something about it – it’s clearly the music of ordinary people, which gives it depth, and the rhythm is just addictive. You just cannot stand still when you hear it.

What’s so special about reggae rhythm? At first look, it’s just the regular 4/4 beat. But it’s somehow different at the same time. And even within reggae, the pop style of UB40 for example, is very different from the roots reggae of Bob Marley. This got me wondering.

Then, a few months ago, I saw the 2012 documentary ‘Marley’. It is an amazing mosaic of the life of an amazing man. And I understate it severely when I say ‘amazing’. It was available on Netflix for a while, but I see that it has now gone away, at least here in the USA. You can still buy it from Amazon, or watch it in low-resolution here. I highly recommend purchasing and keeping it, it’s that good.

From this documentary, I learnt about how reggae came to be, and how it is different. A typical 4/4 beat emphasizes the first beat the most, and the 3rd as well. Here’s a good example, Bohemian Rhapsody by the popular rock band Queen. You can clearly hear the 1-2-3-4 if you look for places where the drums start, as they do at 02:00 in the video.

In this case, the numeral in bold and italics is the strongest beat, and the numeral in bold alone is strong too. We could represent it like this pictorially:

The thicker and longer the line, the stronger the beat. You can imagine beating a drum with a severity equal to the weight and length of the arrow.

In reggae, it is the off beat (#2) that is emphasized, as well as two next beats. The ‘2’ and ‘4’ are typically guitar, and the ‘3’ is the base drum. Now the beat is 1-234. Bunny Livingston, one of the original Wailers, explains it really well in this short clip from ‘Marley’:

So in our pictorial representation, this is what the reggae beat would look like:

This reggae beat was discovered most likely by accident. Bob Andy, the recording artist of Studio 1 where Bob Marley and the Wailers did many of their early recordings, said that there was some equipment they had brought from the USA, which they tried one day. It was a tape delay, and it caused an echo of the guitar strum, which led to the ‘chekke’ sound typical of the reggae guitar. Carlton Davis, session drummer of the Wailers talks about this sound in ‘Marley’. And  you hear it clearly in the Bob Marley song that follows as well:

The drum beat comes on the ‘3’ and is usually quite strong, which is why is is 1-234. The ‘chekke’ comes on the ‘2’ and often on the ‘4’ as well. And the bass guitar also comes in on the ‘2’ and ‘4’. Aston Barrett, bass guitarist of the Wailers explains it here:

This is what makes reggae unique. This idea of guitar and drums is really important. If you switch the two, drums on the ‘2’ and ‘4’ and guitar on the ‘3’, it sounds totally different, as you can hear in this song from the 70s by The Police:

So, now that you understand the reggae beat, let’s leave you with a very nice reggae song from Steel Pulse. Enjoy!

Why Machine or Deep Learning, and How to Choose

Two incidents happened recently, which prompted me to write this article. First, I was writing a proposal for a potential partner for my startup, on how we could partner with them to bring AI into their products and applications. They then turned around and asked us why they should want to develop AI applications in the first place.

Second, I was chatting about my startup with an investor friend. After he understood that we did a platform to accelerate the adoption of AI techniques by enterprises, he told me: “Shekhar, you are two steps removed from the problem. Enterprises today don’t even understand what AI is, how it differs from traditional techniques. They need to get this first.”

Of course I knew this, and had been talking about it, but I realized I needed to put it down. Now, a real comprehensive reasoning will be customized to the specific customer and the problem. I do have some general points however and here they are.

Analytics has been and can be done without AI of course. In many cases, the non-AI approach can be severely limiting, for the following reasons. This is not intended to be an exhaustive list, it only contains some of the main reasons why many enterprises are moving to AI models for data analysis.

  • The traditional data analytics approach is reactive. Data is collected from various sources, and analyzed with tools that display it in various ways, with dashboards, graphs, logs, etc. The patterns that these tools look for are based on existing knowledge. This is akin to monitoring, and the response to triggers requires human intervention anytime something new is seen. With AI, the responses are proactive. AI is used to predict the triggers, and automatically apply the remediation, hence it is a further level of automation with a significantly reduced need for human intervention.
  • AI makes prediction decisions based on past and current input of data into the model. The model learns from the data input, and the more realistic data is used, the better it learns. This then improves the predictions and decisions the model makes – so, a prediction made in the past would generally be different than a prediction made later, even with the same data input, because in the time in between, the model has learnt more and has become more accurate in its predictions. In the case of non-AI predictive models, the same data input always results in the same predictions irrespective of when this happens, since the model is not learning.
  • Traditional models were designed to work with less data. Nowadays, every organization collects large amounts of data that were not possible before. The newest deep learning algorithms perform better with an abundance of data which was not the case before. The amount of data used for these techniques would often be overwhelming for traditional techniques.
  • With unsupervised learning, the newer AI techniques can detect patterns in the data that would not be obvious to humans. For example, recently it was determined that the plague or Black Death in 14th century Europe was more likely spread by humans via lice, rather than rats as was earlier thought. This was determined by simulating outbreaks in various cities with different models (rats, airborne, fleas/lice) and finding out which fit best. If the data had been fed to a deep learning model with unsupervised learning, it would have discovered patterns that led to the conclusion faster, without the need for simulations.

Within the broad category of AI, there are multiple categories of learning. The latest, and the one most cited these days is Deep Learning. The main difference between deep learning and traditional machine learning is in the classification of features required to solve the problem. With traditional machine learning, the feature classification is done manually. With deep learning, the system figures out which features are important for making the prediction and automatically evaluates them for making decisions.

For example, if our problem is weather prediction, and we have collected data on various weather phenomena from the past. To be more specific, let’s say we want to predict the likelihood of a natural fire in wooded areas. We may define the features required for this prediction are: humidity, temperature, the density of trees in a given area, among others. For traditional machine learning, we would need to create algorithms that predict the possibility of a fire as a function of these features. A deep learning algorithm would automatically figure out which features are needed to make a prediction. Suppose we added a feature such as the population of rabbits in the given area. Then, the traditional machine learning model would use this data for its prediction if it were programmed as such, however, the deep learning model would recognize from its training that the rabbit data is not relevant to the prediction and would not evaluate it.

Deep learning also works better with the amount of data fed to it for training. Traditional models taper off after a while, as seen from the graph below. Hence this is a good technique to use in cases where a lot of training data is available.

Some good examples of where to use deep learning is in threat detection where there are many samples of attacks (such as network intrusions) or predictive maintenance for IoT devices, where a lot of failure data is available (Hitachi uses this on their remote earthwork machines).

This also means that in the case where very little training data is available, deep learning is not a very good technique. An example is spearphishing attacks, where the number of successful attacks is very small, hence there is not much training data available for deep learning models, and the number of false positives is so high that this is not a good technique for predicting these specific types of attacks (see this paper on spearphishing).

Here is a cheat sheet for deciding what type of machine learning technique to use for a given problem. The original has links for each individual technique, to explore more details on that technique, and can be found here.

My Podcast on Argentine Tango

Some time ago, I did a podcast on my experiences with Argentine Tango dancing. Tina Baumgartner interviewed me and put it up. The link is here: http://avariedlife.com/passion-for-tango/

Lunar Rainbow at Victoria Falls

Victoria Falls is by itself an amazing wonder. But about three or four times a  year, an amazing phenomenon occurs. During the nights of the full moon, there is enough light to create a lunar rainbow over the falls. This happens in the months of June and July, when it is winter in the southern hemisphere.

At the same time, the water over the falls is at its maximum. In the summer months, there is less water, with the Zimbabwe side seeing the most, and the Zambia side mostly dry. But in the winter, there is a single sheet of water all across the falls.

The Avani Livingstone Resort on the Zambia side is probably the best place to see the falls from. It is right next to the falls, and it takes about 5 minutes to get to the view sites by foot on the falls trails.