- Startup in China uses gait recognition technology to identify people. Said to be accurate even with disguise and faked limps. Currently it is tested for criminal investigations, but it screams intrusion, especially in CN.
- Deep learning optimizes based on parameters and weights of a neural network through a stochastic gradient descent (backpropagation) which can be costly an inefficient. “CoDeepNEAT” assembles layers of neurons in an artificial neural network by mixing different functions in each layer, with convolutions of a CNN or a ‘cell’ of a recurrent neural network.
- To meet the challenge of AI for enterprise, IBM draws upon the capability maturity model. Fundamentally, AI deals with probabilistic than deterministic models, and requires a lot of “messy data”. IBM’s paper calls for AI to be customized to its business case. However, ground truth labels on data detracts from unsupervised machine learning. A Training Lead advocated to lead feature extraction contradicts automated feature extraction in deep learning. Lastly, there are no clear guidelines on how enterprises can keep their models unbiased.
- Machine learning with images and videos require a richly annotated / labeled data set. Researchers at Sun Yat-Sen use a neural network model to continuously compare guesses of multiple networks with one another, reducing the need for ‘ground truth’ from a labeled dataset. A convolutional neural network (CNN) extracts the 2D stick figure before a long short-term memory (LSTM) neural network specializing in retaining memory of a sequence of events extracts continuity of the body from multiple video frames to create a 3D model. The learning phase is self-supervised, correcting its comparison between 2D and 3D models.
QDNC Issue #6
- WeWork culls 3% of its workforce for ‘performance-related’ reasons. Although its bonds are trading at a lower offer price year to year, Softbank committed an additional $2 billion earlier this January, bringing its total investments up to $10 billion.
SpaceX successfully launches Crew Dragon in an uncrewed flight test. It is the first of SpaceX’s spacecrafts meant to transfer humans to and from the International Space Station. On 3 Mar morning, it is scheduled to dock to the station, then complete a return trip to Earth on 8 Mar. Musk indicates concern on how it would perform during its high-speed re-entry into Earth’s atmosphere.
- While the wedding industry brings in big money, tech companies find it hard to disrupt. One of the reasons acquisition cost, which recurs cyclically. The other is the inertia to use tech for merely one aspect of wedding planning. Of these, tech companies with carefully selected value propositions have thrived (albeit in limited numbers). The Knot – a wedding planning website, and now Zola: an e-commerce platform for couples and wedding goers to register for products from hundreds of brands. Success can partially be attributed to good timing: it came about the time when drop shipping was picking up.
- Tech innovations in law include: AI that reads through all case law and provides lawyers with a collection of cases required to develop legal positions (BakerHostetler); tech to enable data breach readiness, cybersecurity (Rajah & Tann Asia); operational automation like e-filing and e-billing, and data analytics (Jerrold Soh, Lex Quanta). Personally I find data security and analytics the most needed and willingly adopted aspects of legaltech. Note to read up more on this.
- Being an ex-avid user of ClassPass, I’ve always been curious about CP’s profit model and their value proposition to studios and users. This is a pretty comprehensive article that highlights how CP adjusts its offerings to increase its margins. I think as long as CP continues its path to data sophistication, and inform their pricing decisions based on solid data analysis, it has a great chance at survival and even profit.
- An interview with Michael Chui (MGI partner) on using AI for social good. Many of these revolve around disaster relief. He believes that, most of the time, the data exists – even in traditional forms (kept in a database). The challenge is in making it accessible for use.
QDNC Issue #5
- DBS digiPortfolio: a team of portfolio managers curates the portfolio and monitors the market, initiating rebalances where required. This is supplemented by automation of back-testing, rebalancing, monitoring, and display of all trade activities. It aims to attract non-investors and will be implemented end 2019. Developed in partnership with HK based robo-advice startup, Quantifeed.
- Razer opens Malaysian HQ, plans for it to be a fintech and talent development hub.
- A telemedicine cabin developed by French firm Health for Humanity (H4D) allows a doctor hundreds of miles away from a patient measure their pulse rate, temperature and blood oxygen level, making it easier to provide healthcare in remote areas.
- Intel creates the Cryogenic Wafer Prober, which allows researchers to test qubits on silicon wafers at temperatures of just a few Kelvins. This allows scaling up of silicon quantum computers manufacturing. Current QC processors are tested for months in a low-temp dilution refrigeration. The Prober can collects information such as sources of quantum noise, quality of quantum dotes, and materials that can be used to create the spin cubits within minutes.
- The growth of China’s P2P lending has drastically reversed, with total loan amount at 91.4 billion yuan, down 55.1% year on year. In 2018, fraudulent business models, poor loans, and company shut downs led to stricter governmental regulations on P2P lending. China still remains a favorite as a leader in fintech financing globally.
- DeepMind brings wind output predictions closer to accuracy. The algorithms were trained on historical weather data and the year’s worth of wind power output. The industry has been using AI for years to refine wind prediction, but difficulties remain – of these are unexpected external factors. DeepMind aims to predict wind output 36 hours in advance so energy sources can be scheduled in response, boosting value of wind energy.
QDNC Issue #4
McKinsey Insights has a compilation of amazing articles. Today’s list is – unintentionally – centred around fintech. Because McKinsey articles are so comprehensive, I find it hard to summarize its essence, but the quality content compelled me to highlight them anyway! I highly recommend reading the articles in full instead.
- Financial and regulatory risks associated with machine learning models push financial institutes to restrict ML usage to low-risk applications such as digital marketing. Increased risk likely due to increased model complexity, often require design decisions before training takes place. Six new elements to the validation framework of banks —interpretability, bias, feature engineering, hyperparameters, production readiness, and dynamic model calibration — could work to enhance model-risk management. This was tremendously insightful: definitely coming back for a second read to digest it again.
- With the trend of e-merchants and increasingly digitized payments in China, CreditEase steps in to the demand gap for small-business lending. They use AI to make wealth management and investment decisions for high net worth clients.
- Goes through the type of payment frauds, how advanced analytics are used for fraud detection, challenges, and what a successful model of fraud detection with analytics would look like: business-back, criminal-forward, intelligence-driven, customer-focused.
- Five ways to use data analytics in a smarter way. By aligning analytics to strategic vision; embedding analytics into decision-making and workflow;, developing advanced analytics teams and assets; invest in critical analytic roles; keep data ready for use case and user revolution.
QDNC Issue #3
- Project Leo is an AI-powered copywriter by Singapore based data innovation center under Dentsu Aegis. It leverages on Google’s machine learning and matches customer intent to generate Google search ads. Side note, Alibaba also uses NLG to generate its product copy on Tmall and Taobao.
- Singtel partners with China Mobile International, allows China Mobile enterprise customers to seamlessly deploy IoT devices onto Singtel’s network in Singapore without switching network. Additionally, Singtel partners with cloud computing platform, IoT applications can now migrate from device – network – cloud more seamlessly. Wondering what IoT applications they refer to. ‘Smart metering for utilities, powering connected health and industrial appliances, and smart city infrastructure’ among NB-IoT cited.
- A research team at RMIT designed a liquid metal catalyst that converts carbon dioxide gas back to carbon flakes, that can be re-buried underground. The added benefit is that carbon can hold electrical charges, effectively acting as a supercapacitor.
- Gradient Ventures (Google’s AI fund) invests $7M in Elsa, an English learning app that uses voice recognition to grade spoken language skills in non-native learners.
- Oracle’s suit against Google for the latter’s use of Java APIs in Google’s Android system. (Java is owned by Oracle.) Red Hat, Mozilla, Microsoft, Python, stand behind Google, citing that if APIs could not be freely copied, programmers cannot work collaboratively across software.
- Joint Enterprise Defense Infrastructure program (JEDI) is a 10-year contract up to $10 billion that would have the Pentagon using a single cloud provider. Pentagon was going to to use just one provider, likely Amazon. Oracle has sued former DoD staff, allegedly with ties to Amazon, but were also involved in the JEDI contract.
- Facebook‘s new privacy tool allows deletion of all history by its users, warned of damaging ad targeting without user data available.
QDNC Issue #2
- Current AI projects have humans tag data set. Superb AI generates customized tagged data for companies.
- Microsoft has launched a dedicated quantum programming language, Q#. IBM offers an integrated quantum computing system for commercial use, making up part of IBM Q System One that runs on the Q network. Programmers can use Qiskit to submit their quantum programs. D-wave launches LEAP, their real-time Quantum Application Environment (QAE).
- Proposed concept that gravitational phenomenon like formation of black holes and planetary systems have the same origin as quantum phenomenon (entanglement, tunnel effect).
- Radio Frequency (RF) transmitters can transmit at far longer distances than most assume. Leveraging this, Ubiquitilink aims to provide a satellite-based network service. Current prototype provides 2G service. Ways to lengthen RF transmission include: lowering orbital height, narrowing RF beam, lengthening wavelength.
- Warren Buffett sold his $2.1 billion stake in Oracle within a quarter, claiming he didn’t understand the space. Previous experience with in IBM cloud, which he ditched last year, was a blemish in his investing track.
- Investigative piece on content moderators in Facebook. Low wages, lack of employee empathy, and strict regulations make for psychologically stressful workplace. Employees reportedly influenced by the number of conspiracy theory videos watched, some buying into them from extensive exposure.
- AI text generator by Open AI – GPT2 – that appears to read context and style. Avoids previous limitations of losing the context mid-sentence or mid-generation. Generates text across fiction and non-fictional samples. Further reading.
Q.uark: Creativity in AI
Entrepreneur: Elon Musk Predicted Artificial Intelligence Would Be ‘Seriously Dangerous’ by 2019. How Close Is That to Reality?.
https://www.entrepreneur.com/article/323278
A great point brought up here – that AI could be narrow (trained to do a specific set of tasks) or general (intelligence that is fluid, creative, like humans are).
The current state of AI is, as the article outlines, feeding large amounts of historical data, letting the machine learn, and awaiting an output from this training.
Can we expect intelligence that extends beyond the limitations of its training set?
What kind of data and/or how do we structure these data for General AI to take shape?
The author’s stand is that we currently do not have a functional General AI which I agree with, but am more optimistic about.
One factor for my optimism is triggered by some questions he posed: there is not a single creative neuron in our minds, yet we are capable of creativity. How do we replicate that in AI? How about emotions?
Having some background in neuroscience I can chip in with a possibility.
Creativity is not purely determined by a single gene, one specific area of your brain. Creativity, in my opinion, is an interplay of natural tendency (to think, or not think, a certain way); synaptic connection, pattern recognition; motivators and parameters. Creative, thus, can be trained.
Synaptic Strength and Pattern Recognition: The more we are trained to associate certain elements and ideas together, the stronger our synaptic connections. This can go both ways. In one, stronger connections encourages pattern recognition. We associate disparate or related concepts quickly to create an new output from existing elements.
On the other, the more we associate related concepts, the harder it gets to break down these associations to think of new links – in other words, new perspectives and creativity.
Motivators: motivators to me are emotions, external stimulators like the environment. Great war poets and lyrical verses on love are stimulated by war, by heartbreak, suffering. How do we stimulate, for a machine without survival instinct and emotions, a motivation to create?
Parameters as replacement for motivators: Just as we teach infants, we train machines at the early stage with rules and parameters to define and structure learning. Since we have yet a viable model for motivators, could we train machines to break these parameters once they have matured in its learning? (As Audemars Piaget would say, we have to first master the rules to break them.)
Is there a specific way humans bend and break parameters to exhibit creativity?
QDNC Issue #1
- Chinese scientists use CRISPR to delete CCR5 genes in twins before conception, aim to create HIV resistance. The consequence is improved cognitive functions. Previous studies have found links between CCR5 and memory, learning.
- Adyen launches direct consumer to merchant payments powered by open banking, leveraging EU’s Payment Service Directive (PSD2).
- UOB previously partnered with Razer to launch Razer Pay, and recently teamed up with NETS and Grab. They have launched TMRW in Thailand – a mobile-only banking app that Gamifies Personal Finance.
- Quantum computing‘s benefits should not be limited to speed. Classical computers and QCs will serve entirely different purposes. We should think of ways QCs can be used, likely in ways that computing is not used for now.
What is Q.bit?
In a bid to keep up with the new, Q.bit is a personal project to aggregate news I read daily that are most intriguing or relevant. The topics I follow include:
technology — big data, artificial intelligence, machine learning, data science, data analytics, innovations
entrepreneurship — the startup scene, new innovations and businesses, agile methodologies
digital payments — e-payments, mobile wallet, payment gateways
quantum mechanics — quantum computing, quantum philosophy and theory
astrophysics — deep space, space technology
biotechnology — gene editing
psychology — cognitive psychology, behavioral psychology, neuroscience, neuropsychology
consumer behavior — marketing software, consumer insights, consumer data
This list is non-exhaustive and will continue to expand as my interests grow!
How it works
Everyday, I aim to post a self-written summary of at least 3 news articles I’ve read. The more the merrier of course! I would have – at bare minimum – skimmed through the article to understand the essence of it.
Why not just share the headlines? Because headlines are too often misleading. This project is both to force me to read the entire article.
Q Daily News Crunch / Q Weekly News Crunch
Aggregated news I’ve read and summarized.
Q.uarks
A quark (/kwɔːrk, kwɑːrk/) is a type of elementary particle and a fundamental constituent of matter.
Q.uarks will be more personal thoughts in response to what I’ve read. Things I wonder about, questions I have. They may come as a retrospect of insights and trends in the past month of news I’ve read.
Updates
Updates will include posts about this blog, or anything to do with my personal or professional development insights.
Why “Qbit”?
In quantum computing, a qubit (/ˈkjuːbɪt/) or quantum bit (sometimes qbit) is the basic unit of quantum information—the quantum version of the classical binary bit physically realized with a two-state device.
Hi, I’m Qing – have always been partial to the letter Q. And am a Quantum Mechanics fan! Thus Qbit 🙂 It’s bits of information I gather over time.
Q.bit 