Daily Tech Digest - May 07, 2021

Four Things to Do if Your Bank is Eyeing Digital Assets

The evolution of money toward digital assets is affecting bank and fintech organizations globally. Companies should proactively think through adjustments now that will enable them to keep up with this rapid pace of change. At the start of this century, when mobile banking apps first began appearing and banks started offering remote deposit captures for checks, organizations that were slow to adopt these technologies wound up being left behind. The OCC guidance explicitly authorizing the use of digital assets should alleviate any doubts around whether such currencies will be a major disruption. ... A crucial determinant in how successful a bank will be in deploying digital asset-related services is how well-equipped and properly aligned its technology platforms, vendors, policies and procedures are. One of the primary concerns for traditional banks will be assessing their existing core banking platform; many leading vendors do not have blockchain and digital asset capabilities available at this time. This type of readiness is key if bank management hopes to avoid significant technology debt into the next decade.

How do Decision Trees and Random Forests Work?

There are two types of decision trees: classification and regression. A classification tree predicts the category of a categoric dependent variable — yes/no, apple/orange, died/survived, etc. A regression tree predicts the value of a numeric variable, similar to linear regression. The thing to watch out for with regression trees is that they can not extrapolate outside of the range of the training dataset like linear regression can. However, regression trees can use categoric input variables directly, unlike linear regression. While the Titanic decision tree shows binary splits (each non-leaf node produces two child nodes), this is not a general requirement. Depending on the decision tree, nodes may have three or even more child nodes. I’m going to focus on classification decision trees for the rest of this article, but the basic idea is the same for regression trees as for classification trees. Finally, I’ll mention that this discussion assumes the use of the rpart() function in R. I’ve heard that Python can’t handle categoric variables directly, but I’m much less familiar with Python, especially for data analysis. I believe that the basic theory is the same, but the implementation is different.

Why financial-services firms need to change with the times

Rapidly evolving technology, regulatory constraints, and relentless pressure to hit short-term financial targets may be hindering firms from making needed investments to upskill their employees. These employees also face critical skills gaps in areas such as empathy, resilience, adaptability, and creative problem-solving. Turnover is a factor as well — firms may resist investing in bespoke training initiatives that increase the market value of their people, who then leave and take their enhanced skills profile with them. Such programs are expensive and have an uncertain ROI. ... The challenge to upskill so many people is so significant that firms may not be able to solve it by working independently — though many have started that journey. For example, in 2017, Citigroup announced a partnership with Cornell Tech to develop digital talent in the New York City labor market. But a market-based, go-it-alone approach may be too slow, or risk leaving small firms behind. It behooves industry-wide associations and trade groups to create the right foundation to help all firms in a country to close the skills gap, leading to faster progress at a sector level.

The Rise and Rise of Digital Banking: How Fintech is Set to Disrupt Brick and Mortar Banking

Industry insiders have long been concerned about the role fintech have been playing in the world of banking and whether or not they will ultimately replace traditional financial institutions. This fear was exacerbated by the recent introduction of the People’s Bank of China Fintech Development Plan which looked to accelerate the accommodation of digital financial services in the country. But could fintechs actually spell the end of traditional banking? To address this properly, let’s address what finance actually is. The purpose of finance is to realise the optimal distribution of capital across time and space amid uncertainties and to serve the real economy and maximise social utility. One big barrier to this can be found in adverse selection through a lack of information and the emergence of ethical issues. Finance should exist to identify and price risks. All technologies that are developed should be intent on helping to better understand customers and their willingness, and ability, to pay – while pricing them accurately. With this in mind, traditional banks have an advantage in terms of capital costs, while fintechs are competitive in terms of operating costs.

Quantum computing could be useful faster than anyone expected

For most scientists, a quantum computer that can solve large-scale business problems is still a prospect that belongs to the distant future, and one that won't be realized for at least another decade. But now researchers from US banking giant Goldman Sachs and quantum computing company QC Ware have designed new quantum algorithms that they say could significantly boost the efficiency of some critical financial operations – on hardware that might be available in only five years' time. Rather than waiting for a fully-fledged quantum computer, bankers could start running the new algorithms on near-term quantum hardware and reap the benefits of the technology even while quantum devices remain immature. Goldman Sachs has, for many years, been digging into the potential that quantum technologies have to disrupt the financial sector. In particular, the bank's researchers have explored ways to use quantum computing to optimize what is known as Monte Carlo simulations, which consist of pricing financial assets based on how the price of other related assets change over time, and therefore accounting for the risk that is inherent to different options, stocks, currencies and commodities.

Cloud Native and Kubernetes Observability: Expert Panel

The concept of observability is really agnostic to where you’re running your workload, but the added complexity of multi-tenancy, cloud-native workloads, and containerization lead to a rising need for observability. Single-tenant monoliths can be easier to make observable because all the functionality is right there, but as you add more services and users there’s a chance that a bug will only manifest for one particular combination of services, versions of those services, and user traffic patterns. The most important thing to be aware of is when you’re about to grow your previous solutions, and to be proactive about adding the right instrumentation and analysis frameworks to achieve observability before it’s too late. When you stop being able to understand the blast radius each change will have, and when you stop being able to answer the questions you have about your system because the underlying data has been aggregated away…that’s the point at which it’s too late. So be proactive and invest early in observability to both improve developer productivity and decrease downtime.

How To Take Full Advantage Of GPUs In Large Language Models

Typically, training models use weak scaling approaches and distributed data parallelism to scale training batch size with a number of GPUs. Though this approach allows the model to train on larger datasets, it comes with a trade-off; all parameters must fit on a single GPU. This is where parallelism comes into picture. Model parallel training overcomes this limitation as it partitions the model across multiple GPUs. Previously, general purpose model parallel frameworks such as GPipe and Mesh-TensorFlow have been proposed for the same purpose. While gPipe divides groups of layers across different processors, Mesh-TensorFlow employs intra-layer model parallelism. Other methods of model parallelism such as tensor and pipeline parallelism have been proposed too. Unfortunately, wrote the researchers at NVIDIA, naive usage leads to fundamental scaling issues at thousands of GPUs. Expensive cross-node communication or idle periods waiting on other devices are few reasons. Moreover, the high number of compute operations required can result in unrealistically long training times without model parallelism.

Optimal Feature Discovery: Better, Leaner Machine Learning Models Through Information Theory

From the perspective of information theory, both the prediction target and the features in a model are random variables, and it’s possible to quantify in bits the amount of information provided about the target by one or more features. One important concept is relevance, a measure of how much information we expect to gain about the target by observing the value of the feature. Another important concept is redundance, a measure of how much information is shared between one feature and another. Going back to the coin flip example, there could be different ways to obtain information about the bias of the coin. We could have access to a feature that tells us the rate of heads based on the design of the coin, or we could build a profile feature that tracks the number of heads and tails, historically. Both features are equally relevant in that they provide equal amounts of information, but observing both features doesn’t give us more information than observing either one, hence they are mutually redundant.

There’s a revolution coming in voice profiling and the warning signs are loud and clear

When conducting research for my forthcoming book, The Voice Catchers: How Marketers Listen In to Exploit Your Feelings, Your Privacy, and Your Wallet, I went through over 1,000 trade magazine and news articles on the companies connected to various forms of voice profiling. I examined hundreds of pages of US and EU laws applying to biometric surveillance. I analysed dozens of patents. And because so much about this industry is evolving, I spoke to 43 people who are working to shape it. It soon became clear to me that we are in the early stages of a voice-profiling revolution that companies see as integral to the future of marketing. Thanks to the public’s embrace of smart speakers, intelligent car displays and voice-responsive phones – along with the rise of voice intelligence in call centres – marketers say they are on the verge of being able to use AI-assisted vocal analysis technology to achieve unprecedented insights into shoppers’ identities and inclinations. In doing so, they believe they will be able to circumvent the errors and fraud associated with traditional targeted advertising.

Linux Foundation launches open source agriculture infrastructure project

The Linux Foundation has lifted the lid on a new open source digital infrastructure project aimed at the agriculture industry. The AgStack Foundation, as the new project will be known, is designed to foster collaboration among all key stakeholders in the global agriculture space, spanning private business, governments, and academia. As with just about every other industry in recent years, there has been a growing digital transformation across the agriculture sector that has ushered in new connected devices for farmers and myriad AI and automated tools to optimize crop growth and circumvent critical obstacles, such as labor shortages. Open source technologies bring the added benefit of data and tools that any party can reuse for free, lowering the barrier to entry and helping keep companies from getting locked into proprietary software operated by a handful of big players. ... The AgStack Foundation will be focused on supporting the creation and maintenance of free and sector-specific digital infrastructure for both applications and the associated data.

Quote for the day:

"Leadership appears to be the art of getting others to want to do something you are convinced should be done." -- Vance Packard

No comments:

Post a Comment