Collision - 2019 Takeaways

These are my long overdue notes from Collision - 2019, which I attended at the end of May. The conference was very engaging and started with interviews with the Hollywood stars and prime-minister of Canada Justin Trudeau.

Quick stats:

• 25 000 attendees
• 125 countries
• 68% - senior management
• 45% - female attendees
• 3750 CEOs

Collision as its sister conference WebSummit in Lisbon I went last year has multiple tracks.

The one I was mostly interested in was a #FulSTK - the world’s leading developer conference; a gathering of innovators, data scientists, coders, and engineers programming the future. I devoted my full first day of the conference to this track.

Common themes through all the presentations for this track were cybersecurity and making the world a better place for everyone. 

The first talk I really liked was "The quantum application pioneers are here" by Alan Barat from D-Wave Systems. D-Wave Systems are pioneers in quantum applications development. They were able to build over 150 quantum applications and prove that quantum computing is ready to be used for enterprise. The most fascinating project they showcased was the application that can predict the spike of the terrorist attacks based on analyses of a graph of relationships between leaders of terroristic organizations. They discovered that the moment this graph is unbalanced there is a terror splash. Interesting fact - it all started with a graph for relationships of Romeo and Juliette families.

The most exciting thing is that everyone can start building their own quantum computing apps right now. You can register here and find tutorials, SDK and get free 1 minute of quantum computer per month. It sounds like very limited time but it can run 400 - 4000 apps in a minute so it is enough to play with.

Next talk “Petabytes are new terabytes” by Brian Carmody was about storage. He and his coworkers in Infinidat were able to build storage systems, that are bigger than any currently available. They do it for banking, core DNS (100% availability requirements) and for other systems that require scale, availability and the huge amount of data. Their average deployment 7 petabytes. maximum - 100 petabytes.

They not only a very technical company but also want to solve some problems that can impact all of us, such as the water crisis that is expected to be the next world crisis. One of the solutions – is precision agriculture, when plants are watered precisely where it’s needed when it’s needed and with the minimum amount of required water. Precision agriculture requires collecting a lot of data, that need to be stored and accessible with low latency. The storage also should be cheap so third world countries can afford it, as decreasing inequality – one of the goals of the Infinidat.

Building solution that can provide a lot of cheap, low latency storage based on the hardware is hard. Nowadays hardware is complicated and software is dumb. So, this company used a different approach – they use the cheapest dumb hardware and building smart software on top of it. They use Multiple-Value logic trees for it (don’t ask me what it though), Predictions, DRAM and NAND flash

Next talk was about blockchain (it is impossible to have a conference without this topic nowadays, right?). The name of the talk was “Blockchain: the latest cybersecurity hype?” by Zulfikar Ramzan from RSA Security. The main idea of this talk was although blockchain is good for solving some problems, for example, to use it in cryptocurrency, using it everywhere can be a bad idea. Blockchain got everyone's attention last couple of years, but it is really old and most of the blockchain ideas were developed in the early ’80s. Many tasks solved with it can be solved using old good algorithms and tools so you really need to ask you should I use blockchain. Some of the questions everybody needs to ask themselves before using blockchain are:

• Do I need to store state?
• Are multiple peers contributing?
• Can I sacrifice performance?

If the answer is no to one or more questions - do not use; use other things. He also noted that putting blockchain security on top of not secure things - does not make any sense.

Blockchain is neat for cryptocurrencies but outside digital transactions not that much; use older technologies and compare blockchain with other things before selecting it.

Blockchain talk was followed by an open panel discussion “The state of DevOps”. Among the discussed issues were the education of DevOps, gaps between understanding of DevOps on CEO and employee level; tools issues, like missing common levels of visibility and standardization. There is the common misconception that DevOps allows to automate everything and remove people replacing them with middles iRobot Roombas, but the idea is to build Ironman's suite for people and increase their capacity and this should be the way forward.

The possible next wave of startups will be DataOps

Another panel discussion I attended was  “Cybersecurity panel: in search of Holy Grail”. Latest trends show that security is a big issue, and everybody is looking for one solution to make everything secure. Recently many developers started to look at two-factor authentication as Holy Grail. It really works very well, and everyone should provide it, but it’s not a magic pill. Everybody should make daily exercise to increase security, use encryptions. There also should be regulations by the government and better relationships between government and private companies. UK government already started to regulate some IOT, so it’s started. Also, regulations and enforcement should go together.

Next presentation “Democratizing speech technologies” was from Katharina Borchert, Mozilla. She was talking that the Internet is currently owned by big companies and our content belongs to somebody else and monetized by them. When we have monopolies creativity suffers, all products are oriented on the mass market, so the speech recognition is done only for the most popular languages and does not care about emerging markets, about people who cannot read and write and could definitely benefit from using voice. It will be so for a long time because countries, where illiteracy is high, are poor countries and investment will not pay off.

Mozilla decided to solve this problem and created the open and free voice library and tools but it needs our help. There is a website Common Voice where you can read some weird sentences and in such way donate your voice. You can also validate others voice. This is the way how Mozilla collects training data, but training data cannot provide it alone, so they created Project Deep Speech, open source speech to text engine everybody can work with. Project is started in English but there are technologies to collect and start other languages 

"Leveraging ubiquitous computation intelligence" by Stephen Wolfram was a presentation devoted to Wolfram language, another thing I would like to try to use. Wolfram is a general multi-paradigm computational language developed by Wolfram Research and is the programming language of the mathematical symbolic computation program Mathematica and the Wolfram Programming Cloud. It emphasizes symbolic computation, functional programming, and rule-based programming and can employ arbitrary structures and data.

The following talk – “The next-generation stack: A future without massive security breaches”. Dominic Williams (DFINITY). Technology stack is too complex to make it secure and this requires a huge budget. We all depend on third-party tools and APIs. DFINITY wants to solve this problem with building thick internet, and their solution is named Internet Computer. The Internet Computer is a tamperproof open cloud platform created by data centers that are running the secure ICP protocol. It is hackproof and allows easily to build new systems and solutions.

Building more reliable internet by entering chaos. Kolton Andrus (Gremlin) was devoted to chaos engineering - planned experiments designed to reveal the weakness in the system. We cannot rely on systems because they will fail, they are fragile, so we need to break thing on purpose, proactively kill everything with selected blast radius. Gremlin helps with setting up these experiments, automating them, finding gaps in monitoring. There is a free and paid (expensive) version.

“The technology behind large enterprise” Alan Boehme (P&G). Challenge that a lot of big companies suffer - legacy technology. Corporations are organized in different ways, and they need flexibility they don’t have. There is also politics and government impact. 15 years ago, P&G outsourced everything and it was a success. It gave them the opportunity to update backbones of networks; use different cloud solutions, big data.

They are also very concerned about diversity and bias. They created S.H.E browser extension to remove gender bias in the search result. 

The last #FulSTK talk I attended was “What does trust, and safety look like” from Alex Starnos (Stanford University). Twenty years ago, most of the security problems were introduced by software glitches, now by people. User authentication with passwords invented in the 70s for mainframe time-sharing and its outdated. It does not fit the modern world. Nowadays we need to think about how our product will be misused and how people will harm other people using it. 

Next day I attended Binate.io - big data and AI conference. There were too many discussions about moral and AI, moral and Big Data. I understand that it is a hot topic but I also would like to hear about other things. There were talks that showed how big data can solve worlds problems – defining criminal activities patterns, stopping human trafficking and how data, AI, Machine learning are as good as people who wrote it: bias, stereotypes, false positives. The main message was that it is Personal responsibility of each developer to think how the algorithm you wrote can harm people.

As it's mainly entrepreneur conference there were multiple startups worth mentioning but these three are my favorites

https://goodbuygear.com/ - selling old stuff

https://www.joinlane.com/ - one place for all things workspace

https://www.hugo.team/ - collecting meeting notice. I see the startup that tries to solve this problem every year, but this one looks promising.