2019 ~ aiCubicle

Saturday, December 7, 2019

HSBC’s plan to move USD 20 billion in assets to blockchain

December 07, 2019 Nikhil Verma

Investment bank HSBC is using a blockchain distributed ledger technology (DLT) to digitize transaction records of private investments, enabling clients globally to access the details of their assets online in near real-time. It plans to move USD 20 billion in assets that include equity, debt and real estate onto its new Digital Vault blockchain, a shift away from its current use of paper records to respond to client search requests.

Presumably, millions of potential investors and users will be on-ramped to blockchain interfaces and they likely won't even be aware of the backend technology

Digital Vault, developed by HSBC's Securities Services unit (HSS), is expected to eventually handle the custody of additional digital asset classes, enabling the bank to move more of the asset transaction lifecycle onto the ledger in the future. The bank has been involved with enterprise DLT firm R3 since at least 2015.

Stephen Bayly, HSBC's CIO for Securities Services, said the bank is responding to clients who have been requesting real-time visibility into their private transactions so they know when they will receive the coupon on a private debt transaction or to facilitate a records audit. Private assets are prime candidates for digitization and we see this platform as a key step on the journey as the model evolves. In future full transaction lifecycle could be stored on a ledger, including issuing digital tokens instead of paper certificates.

This could be a watershed moment for main-streaming of blockchain in securities management.

Robots in Finance Could Wipe Out Some of Its Highest-Paying Jobs

December 07, 2019 Nikhil Verma

Famed quantitative financial mathematician Marcos Lopez de Prado testified on Dec 06, 2019 before the U.S. Congress, together with four other panelists. The topic was "Robots on Wall Street: The Impact of AI on Capital Markets and Jobs in the Financial Services Industry".

Dr. Lopez de Prado is the Professor of Practice, Engineering School, Cornell University and Chief Investment Officer, True Positive Technologies. As per him the machine learning creates a number of challenges for the 6.14 million people employed in the US in the finance and insurance industry, many of whom will lose their jobs, not necessarily because they are replaced by machines, but because they are not trained to work alongside algorithms. The retraining of these workers is an urgent and difficult task.

Nasdaq runs more than 40 different algorithms, using about 35,000 parameters, to look for market abuse and manipulation in real time. According to another panalist, Martina Rejsjo, head of Nasdaq Surveillance North America Equities, Nasdaq Inc., the massive and, in many cases, exponential growth in market data is a significant challenge for surveillance professionals. Market abuse attempts have become more sophisticated, putting more pressure on surveillance teams to find the proverbial needle in the data haystack.

According to CFA Institute's senior director Rebecca Fender, 43% of CFA members and candidates expect their roles to change significantly in the next five to 10 years, according to a survey of more than 3,800 respondents. The three roles most likely to disappear are sales agents, traders and performance analysts.

While Charlton McIlwain, professor of media, culture and communication at New York University said that racial groups that are already extremely underrepresented in the financial services industry will be most at risk in terms of automation and the escalation of fintech development. This is especially true given the vast underrepresentation of African-Americans and Latinx in the adjacent technology sector workforce.

Bitcoin’s maximum supply is now well below 21 million

November 12, 2019 Nikhil Verma

In October 2019, the Blockchain Bitcoin surpassed the 18 million BTC milestone. In fact, there are only 3 million Bitcoins left that can still be mined in the future. However, the maximum number of Bitcoins that can ever be usable is now closer to 17 million than the 21 million to be created that was initially fixed by Satoshi Nakamoto. According to a study by Chainalysis, which was published at the end of 2017 and relayed at the time by Fortune.com, nearly 3.79 million Bitcoins had already been lost. However, there are still 120 years to go before all Bitcoins have been officially mined. This should take us to 2140.

But if the coins are lost at this rate, would it become a collector's item? or will the rarity of the coins boost it price and still be or practical use?

https://medium.com/altcoin-magazine/bitcoins-maximum-supply-is-now-well-below-21-million-33023a576bf1
https://www.unchained-capital.com/blog/geology-of-lost-coins/

Using your ECG tests, AI can predict if you will die within a year

November 12, 2019 Nikhil Verma

Researchers from Geisinger Health System in Pennsylvania analyzed the results of 1.77 million ECGs and other records from almost 400,000 patients. The team used this data to compare machine learning-based models that either directly analyzed the raw ECG signals or relied on aggregated human-derived measures (standard ECG features typically recorded by a cardiologist) and commonly diagnosed disease patterns. The neural network model that directly analyzed the ECG signals was found to be superior for predicting one-year risk of death. The neural network was able to accurately predict risk of death even in patients deemed by a physician to have a normal ECG.

Just a fun thought: While machines and AI algorithms are still benign, what will happen when machines become autonomous and superior to humans - will they "take action" to prove their superiority? It is scary if the result at stake is something like that above.

https://www.newscientist.com/article/2222907-ai-can-predict-if-youll-die-soon-but-weve-no-idea-how-it-works/
https://arxiv.org/ftp/arxiv/papers/1904/1904.07032.pdf

Human-Machine Singularity Series | Evolution of AI and it taking over our jobs!

November 12, 2019 Nikhil Verma

There are four levels of complexity of Artificial Intelligence:

Instructive - Algorithms where rules are codified manually
This has been in existence since the advent of computers and (the machine – i.e. computer) represents a dumb interpreter of human intelligence. This stage saw siloed digitization of data and process with aim to increase speed especially for repeatable tasks and also reduce errors. It quickly evolved into also giving structure to the data and interoperability between organizations. Machines just did whatever humans did much faster, without fatigue and with minimal errors. What could be done by thousands of people could be done by few machines in fraction of time. And indeed, this sparked an exponential decline in human effort required to generate the same degree of value. This stage is level-0 of AI as machines are not thinking – they do whatever you code.

Supervised - Algorithms that codify rules through guided learning
This is where AI has become main-stream today. The outcome is dependent on the past input-output sets (training data) which is not explicitly coded in the solution. The algorithms finetune their input-output logic using past data and are able to predict future outcomes using this “learnt” input-output logic.

Unsupervised - Algorithms that codify rules through exploration
Today, this is beginning to get into main-stream. The algorithm learns from the past input data and figure outs the inferences or outputs from that data, thus creating the input-output logic.

Generalized - Algorithms that codify rules which adapt to changes in environment
This is the cutting edge of artificial intelligence research. It includes the general-artificial-intelligence algorithms. The algorithm learns on the way – just like a child does – and figures out the best input-output logic. It is just few steps away from enabling Human-Machine Singularity where algorithm will continue to refine themselves to become the better of the two races. Who would then need humans! This will have an exponential impact on human employability.

Rise of AI and impact on architects

November 02, 2019 Nikhil Verma

90% of architects will lose their jobs as artificial intelligence takes over the design process, according to designer Sebastian Errazuriz (link below). Wallgren Arkitekter and BOX Bygg have created a tool that generates adaptive plans.

I do not disagree with the assessment. The profession combines art with optimization problems. Consider the task of desigining a house. Give it to AI - it can search the internet for retro or modern ideas, analyze the common design patterns that is in vogue, implement design features that complement climatic conditions, surrounding landscape and ofcourse maximize use of limited land - say between greens and concrete. While mathematics is solving for optimization, AI + Internet add the "art" in the whole process.

Reference:
Rise of artificial intelligence means architects are "doomed" says Sebastian Errazuriz
https://www.dezeen.com/2019/10/22/artificial-intelligence-ai-architects-jobs-sebastian-errazuriz/

Gartner's AI Hype Cycle update for 2019 - signficiant step towards becoming mainstream

November 02, 2019 Nikhil Verma

In Sep 2019, Gartner released an update to its AI hype cycle. Between 2018 and 2019, organizations that have deployed artificial intelligence (AI) grew from 4% to 14%, according to Gartner’s 2019 CIO Agenda survey.

AI is reaching organizations in many different ways compared with a few years ago, when there was no alternative to building your own solutions with machine learning (ML). AutoML and intelligent applications have the greatest momentum, while other approaches are also popular — namely, AI platform as a service or AI cloud services.

Conversational AI remains at the top of corporate agendas spurred by the worldwide success of Amazon Alexa, Google Assistant and others.

DeepMind's AI achieves grandmaster level in StartCraft II game

November 02, 2019 Nikhil Verma

DeepMind (a Google subsidiary) has designed an AI system, called AlphaStar that now outranks the vast majority of active StarCraft II players, demonstrating a much more robust and repeatable ability to strategize on the fly than before. This was quite a feat. StarCaft II is highly complex, with 10 to the power of 26 choices for every move. It’s also a game of imperfect information - and there are no definitive strategies for winning.

The achievement marked a new level of machine intelligence. AlphaStar used reinforcement learning, where an algorithm learns through trial and error, to master playing with all the games. The AI reached a rank above 99.8% of the active players in the official online league. DeepMind team modified a commonly used technique known as self-play, in which a reinforcement-learning algorithm plays against itself to learn faster. DeepMind famously used this technique to train AlphaGo Zero, the program that taught itself without any human input to beat the best players in the ancient game of Go.

The results have ben pubished in Nature on Oct 30, 2019 at https://www.nature.com/articles/s41586-019-1724-z