INSIGHTS News

AutoCIO, the world’s first AI-driven asset management platform delivering active, customisable investment strategies at scale, has appointed Tim Wong as its first Chief Product Officer (CPO).

Tim will take ownership of the strategy and execution of AutoCIO’s product roadmap, developing and scaling the platform in line with evolving client needs. He will lead the integration of AutoCIO into different investment management systems and WealthTech platforms, through AutoCIO’s new Application Programme Interface (API) product.

The product strategy comes in response to a rapidly increasing demand for AI investing solutions that can deliver customised, alpha-generating investment strategies at a fraction of the normal cost and time. AutoCIO’s API solution will mean the platform can be seamlessly integrated into existing order execution and investment management processes.

The expansion of AutoCIO’s integration across the investment management sector comes at a time when the asset management industry looks for new solutions to boost competitiveness, with fees across the sector declining by 4% in 2021^[1]and margins dropping by 14% in three years^[2]. Tim brings more than a decade of experience in product management and strategy consulting. Most recently, he spent four years at Finbourne Technology, where he oversaw product.

development for investment data management and portfolio management software for the asset management industry. Previously, he was a Senior Product Manager at Amazon Prime, leading on the development, launch, and scaling of Prime benefit programmes.

Tim Wong, CPO at AutoCIO, said:

“Asset management is one of the few industries yet to be meaningfully disrupted by technology. Yet it is an industry that is facing a fast-changing landscape, driven by relentless downward pressure on fees, and growing demand from investors for customised products that align to their values and objectives.

“AutoCIO is well positioned to help investment managers leverage AI to meet these challenges, by enabling cost-efficient product development, alpha generation and delivery of customisable investment strategies at scale. I’m excited to be joining a world-class team, to continue strengthening AutoCIO’s value proposition, and to bring our offering to new audiences across the investment landscape.”

Yasin Rosowsky, CEO of Arabesque, added:

“The demand for customisation will be one of the defining industry trends of the decade, and will require technology solutions that can generate hyper-customised strategies on a different scale to anything we have seen before. In the coming years, AI will play a pivotal role in enhancing or replacing traditional products, tools and approaches that are no longer as relevant as they once were.

“As we look to bring AutoCIO to the forefront of the investment management technology landscape, I am thrilled to welcome Tim to our team. His experience and vision will be key to the adoption of AutoCIO across a greater mix of investment clients, reaching far more investors and volumes of capital.”

About AutoCIO

AutoCIO, a platform that allows financial institutions to generate active investment strategies, has doubled the size of its team since 2019.

Users can build customised funds by selecting their preferences across thousands of variables including risk, geography, industry and sustainability.

This enables users to generate hyper customised funds within hours, and at scale, enabling clients to create millions of unique strategies. For each fund generated, users can also analyse a back-test of its performance and compare this against other strategies and benchmarks.  

AutoCIO’s AI engine is designed to analyse patterns and behaviours in financial markets, with a range of machine learning algorithms used to forecast stock performance. The engine processes more than 30 million data points daily, and evaluates the movement of over 25,000 equities.

^[1] Cooper, Brendan. “2021 Fee Analyzer Report: Investment Metrics.” invmetrics.com. Accessed August 11, 2022. https://go2.invmetrics.com/report/2021-fee-analyzer-report.

^[2] Heredia, Lubasha. “Global Asset Management 2021: The $100 Trillion Machine.” BCG Global. BCG Global, July 9, 2021. https://www.bcg.com/publications/2021/global-asset-management-industry-report.

INSIGHTS News

ESG Data Convergence Project metrics for the Private Equity market launch on ESG Book

– Launch of ESG Data Convergence Project metrics on ESG Book enables private equity GPs and LPs worldwide to access meaningful, performance-based, comparable sustainability data from private companies.

– Supported by over 100 leading private equity firms representing $8.7 trillion USD in AUM and more than 1,400 private companies, the ESG Data Convergence Project standardises ESG reporting for private markets.

– Through ESG Book, private companies can disclose information against the ESG Data Convergence Project’s reporting template as well as on a range of sustainability frameworks globally.

– Delivered through cloud technology, ESG Book enables companies to be custodians of their data through a secure digital platform, provides framework-neutral ESG information in real-time, and promotes transparency.

ESG Book, the central source for accessible and digital corporate sustainability data, today announced that private companies can now disclose sustainability data on the platform through a new set of metrics designed to standardise ESG reporting for private markets.

The launch of ESG Data Convergence Project metrics on ESG Book enables private equity general partners (GPs) and limited partners (LPs) worldwide to access meaningful, performance-based, comparable sustainability data from private companies.

Supported by over 100 leading private equity firms representing $8.7 trillion USD in AUM and more than 1,400 private companies, the ESG Data Convergence Project was unveiled in September 2021 as a solution to the private market’s historically fragmented approach to collecting and reporting ESG data. Through ESG Book, private companies are able to disclose information against the Project’s reporting template, as well as on a wide range of sustainability frameworks globally.

Delivered through cutting-edge cloud technology, ESG Book makes sustainability data more widely available and comparable for all stakeholders, allows companies to be custodians of their own data through a secure digital platform, provides framework-neutral ESG information in real-time, and promotes transparency. The platform is supported by international organisations including the International Finance Corporation (IFC), and runs according to principles from the UN Global Compact, which encourages companies to adopt sustainable policies.

Dr Daniel Klier, CEO of ESG Book, said: “The lack of a comparable, consistent ESG data collection and reporting framework has long been an issue for the private equity industry in assessing the sustainability progress of portfolio companies. The ESG Data Convergence Project is a vitally important initiative to help address this problem, and through ESG Book, we are now able to offer a market-leading solution that makes it easier for private companies to securely disclose their sustainability data against meaningful performance metrics. In doing so, this will increase transparency across private markets, reduce data gaps, and drive material ESG improvements.”

On ESG Book, private companies can map their data against the Private Equity ESG Data Convergence Project’s key metrics, alongside other frameworks including the Sustainability Accounting Standards Board (SASB), Global Reporting Initiative (GRI), Sustainable Finance Disclosure Regulation (SFDR) and Task Force on Climate-related Financial Disclosures (TCFD). This will accelerate data comparability across companies, including publicly listed corporations, with a suite of analytics and benchmarking tools allowing GPs and LPs, together with portfolio companies, to measure progress on ESG performance.

Today’s announcement comes as capital flows towards ESG strategies across private markets continues to rapidly grow, with an estimated $3.1trillion¹ of sustainable investments – one-third of private capital under management – now committed worldwide.

1) ‘ESG in Alternatives Report 2021’, Preqin

Notes to Editors

ESG Book is the new central source for digital corporate sustainability data, and is supported by a global alliance of leading organisations including the International Finance Corporation (IFC), Global Reporting Initiative, Bridgewater Associates, HSBC, Deutsche Bank, HKEX, Allianz and Swiss Re. ESG Book was developed by Arabesque, a technology company supporting the financial industry in sustainability and artificial intelligence. It counts many of the world’s leading financial organisations among its clients, which collectively manage over $120 trillion in assets.

INSIGHTS Research

Systematic credit has drawn more limelight over the years as electronic trading of various credit instruments gained in volume and share of the market. The increased availability of high-quality data and growth of liquidity has made it possible for us at Arabesque AI to consider an expansion into credit instruments.

Both the valuations of corporate credit and equity are dependent on the overall health of a company. Therefore, they share many performance drivers, which we have already implemented into our equity analytical models. This makes it an interesting use-case for us to investigate our models’ transfer-learning capabilities. Through various proofs-of-concept over the past year, we have demonstrated the ability to build analytical models for corporate credit bonds.

Naivety and Challenges

From a theoretical, machine-learning perspective, where we have built a strong pipeline of models for our equity predictions, the application to the credit universe is a simple problem. It can be solved by creating a new dataset and a new label which we then use to train and evaluate a baseline model from our existing pipeline. However, the reality is a lot more challenging, due to 1) heterogeneity of data in clustering and across time, 2) high dataset imbalances, 3) trustworthiness of data and last, but not least, 4) entity and exchange mapping challenges. Let’s briefly look at each of these challenges.

Heterogeneity of data: When we consider equities, we can naively group them by the geographies they trade in, the sectors they conduct business in, etc. Ultimately, most of these instruments are non-preferential shares, otherwise known as common shares. Hence, they are comparable in a way. Corporate credit is awash with little details that makes it hard to compare. Some bonds are callable or puttable, which gives either the issuer the right to redeem the bond before the maturity date, or it gives the holder the right to demand the paying back of the principal amount before the maturity date. In stocks, options are separate financial products and therefore don’t need to be considered in pure stock-price forecasting. Further, the maturity dates of bonds are not aligned, one company can issue various types of bonds, such as secured bonds or convertibles and, to make it even more complicated, some European bonds are eligible for the ECB’s asset purchase programme. Hence, the grouping of “similar assets” for training is a harder task in bonds if one wishes to adjust for all these granularities.

To make matters worse, equities can almost always be assumed to be perpetually existing unless in the case of corporate events. On the other hand, bonds can almost always be assumed to expire at some point in time, except in the occasional case of perpetual bonds. This means that the universe refresh rate is exceedingly high. This presents many challenges for machine learning algorithms, not least limited to inconsistent dataset sizes or the unknown extent of survivorship bias vs. maturity effect. Datasets, therefore, need to be asset-agnostic to a certain degree and carefully constructed to maintain comparability.

High dataset imbalance: In equities we can either frame the problem as a price prediction or a returns prediction, either of which can be calculated by the prices of the equities (split/dividend-adjusted, which are still just intrinsic datapoints of the equities). In bonds, we can either frame the problem as a price prediction or a credit spread prediction. The former is a bond datapoint and the latter a combination of the bond yield versus the risk-free rate, typically a US Treasury bond. Here, we are implicitly predicting for “interactions” between two different assets— the bond and the risk-free rate. Moreover, when we train for a target label of a minimum spread widening/narrowing, we find stark class imbalances. These are more pronounced than the same setting in equities of minimum return requirements. The imbalance often calls for the need of readjusting the loss function where for trading cost reasons we would value one class over the other. For example, it is easier going long on a bond than to short a bond compared to the equities world.

Trustworthiness of data: The challenges above are compounded by the deteriorating quality of data in bonds of lesser-known issuing entities or lower credit ratings. In a trading landscape where OTC trading still contributes a significant share of the liquidity, bid/ask data and volume recorded from electronic markets are sometimes misleading, and worse, untradeable. This not only influences the training of the models but also the executability of credit trading signals. Often, this means sanity checking the data manually. The trustworthiness of data also feeds back to design on the type of trading decision horizons and therefore the target labels for the credit model.

Mapping of entities: Many commercial data providers carry their own asset mapping IDs. As bonds are issued by firms that, most of the time, also have issued their own shares means that we have an incentive to link equity IDs to the bond IDs. The mapping is important for understanding where the bonds lie on the capital structure and what credit risks they bear. This is less of a problem when one sources data from the same data provider but quickly becomes a tedious task when mapping across databases.

Measuring the quality of a systematic credit model

For any system, there must be a way to conduct quality checks. For machine learning systems, we can rely on metrics such as accuracy, error sizes, f1-scores etc. However, these might not be sufficient for models that produce forecasts for more illiquid holdings. On longer holding periods, it is important to understand the models from a fundamental analyst’s perspective. This means 1) understanding the behaviour of different machine learning systems and algorithms, 2) understanding the contribution and importance of different input features, and 3) understanding the variability of model outputs.

Model response to datasets: We know that different algorithms respond differently to the same dataset. Training an ARMA model will yield different outcomes as a Gaussian process model. Therefore, we need to monitor the performance of each model for the same dataset on their out-of-sample prediction power. Given known issues with input data and potential clustering of erroneous data, it is also important to understand how the algorithms respond to corrupted data at various segments of the datasets, i.e., response to adversarial attacks. As different models have different data requirements, i.e., i.i.d. variables for some statistical models, and large enough datasets for neural nets, we also investigate the models’ performance when varying sizes of datasets. However, this sometimes results in overgeneralizing and glossing over key differentiating features of bonds. Understanding these aspects is key to choosing models given our aforementioned challenges of persistently wide datasets in credit space.

Feature importance: As we vary the models, the large number of data points we feed into the models makes it hard to differentiate which really contain information, and which are simply noise. We can select features by comprehensively searching through the perturbation of features to identify gains in e.g., accuracy. But this is extremely computationally expensive and only works for one instance of the {model, dataset} set when we could possibly have multiple datasets over the years and different clusters. We can map the feature importance easily when using an XGBoost model through LIME/SHAP algorithms, but these are not necessarily applicable to the other models; the same goes for statistical tests on model coefficients. A hack is to combine a leave-one-out algorithm with a blanket blackbox model representing the entire system of models to map from a subset of features to our produced signals.

Variability of model outputs: Models produce signals that can change as fickle as I change my mind when choosing flavours in an ice cream parlour. A common way to deal with this is to smooth signals over time through moving averages. For systematic credit strategies, however, we need to intuitively understand the fickle signals – if we smoothen the signals, surely that means we cannot be that confident about the models’ decisions? To deal with the volatile signals, we can look at measuring the uncertainty of predictions via inductive conformal prediction which also nicely avoids the need to consistently retrain models.

About Arabesque AI

​Arabesque AI was founded in 2019 as part of the Arabesque Group. We developed our proprietary Artificial Intelligence Engine to forecast stock prices globally on a daily basis. On top of our AI, we built and launched AutoCIO, a platform that creates bespoke investment strategies. Using AI and Big Data, AutoCIO offers hyper-customization, enabling investors to align their investment and sustainability criteria. At Arabesque, AI is not only a buzzword. We advise over $450mn on our platform, proving that AI is ready to be used in practice.

INSIGHTS News

• AutoCIO is powered by Arabesque’s proprietary AI Engine, which uses the equivalent processing power of tens of thousands of computers run on Google Cloud

• Arabesque’s latest offering is an industry solution to customise highly specific active equity strategies tailored to investors’ sustainability objectives and values

• AutoCIO uses AI to forecast stock performance on a universe of 25,000 equities daily, and can create millions of active equity investment strategies

• Over USD 400mio of investment strategies are successfully powered by AutoCIO

• The new technology coincides with a growing industry need to leverage automation and AI for product development, alpha generation and delivering a differentiated client experience

• Arabesque’s technology is the latest evolution and the natural next step for the industry from active to passive and now to autonomous asset management

Arabesque has today unveiled its Autonomous Asset Management offering for the creation of highly customised and sustainable active investment strategies, powered by an artificial intelligence technology that can generate and operate millions of active equity strategies.   

Developed by Arabesque AI, ‘AutoCIO’ enables asset managers and investment professionals to configure and build hyper-customised active strategies that can be tailored to each investor through more than a thousand different personalised investment options. The launch comes as the asset management industry increasingly looks to leverage technologies like automation and AI for cost-efficient product development, alpha generation and delivering a customised and differentiated client experience. With over USD 400 million currently powered by Arabesque’s AutoCIO, the platform offers investors an unprecedented degree of customisation through a streamlined web app that can generate a vast arrange of bespoke strategies, with AI used to forecast stock performance across 25,000 equities daily.

Speaking about today’s announcement, Georg Kell, Chairman of the Arabesque Group, said:

“Artificial intelligence will play a pivotal role in the customisation of active investing in the coming years, with pressure growing to innovate both in terms of technology and client centricity. Whilst the market is increasingly demanding sustainable products that align with the objectives and values of investors, asset managers are currently unable to offer customisable, active solutions at scale. Investment firms face a fast-changing landscape where many traditional products, tools and approaches are no longer relevant as they once were.”

“AutoCIO is a game-changing solution that can enable asset managers to deliver an enormous range of highly customised ESG investment strategies in a cost-efficient and scalable way. Sustainability issues are fast becoming a global priority, and new technologies like this will empower many more investors to participate.”

Powering AutoCIO is Arabesque’s proprietary AI Engine, which identifies and analyses patterns in data on a large scale to discover subtle relationships that can be translated into alpha opportunities. The AI Engine is capable of processing billions of data points each day for its stock signal outputs, using the equivalent processing power of tens of thousands of computers, and is run on carbon-neutral Google Cloud infrastructure. As new data is inputted, the AI Engine re-learns what is driving stock returns and aims to improve over time, removing human biases and reducing the potential for errors.

On the launch of AutoCIO, Dr Yasin Rosowsky, CEO of Arabesque AI, said:

“Asset managers today increasingly need to personalise products and services at scale to focus on customised strategies that incorporate investors’ sustainability objectives and values. We use the power of AI to build systems capable of handling the complexity of financial data and enable scalable investment process design for a wide variety of use cases in an efficient and cost-effective way. This is not a robo or passive investment solution, but fully active asset management, powered by AI. “We are excited to be bringing AutoCIO to the market and provide asset managers and investment professionals with a scalable, digital tool to build actively-managed, customised solutions that meet their clients’ sustainability goals.”

INSIGHTS News

It is with great sadness that we announce the passing of John G. Ruggie, at the age of 76.

John Ruggie touched many minds and hearts, and leaves behind a towering legacy of thought leadership in human rights and social innovation. As an intellectual, he greatly influenced global affairs at the intersection of a globalizing world, with a particular concern for societal implications. As a social innovator, and as policy advisor to former Secretary-General of the United Nations, Kofi Annan, John played a critical role in launching the UN Global Compact in 2000, which has since become the world’s largest corporate sustainability initiative. He was also a driving force behind successful institutional reform and renewal at the UN, for which the United Nations as a whole were awarded the Nobel Peace Prize in 2001.

That same year, John left the United Nations for the Harvard Kennedy School of Government, where he served as Berthold Beitz Research Professor in Human Rights and International Affairs, although he kept in close touch with the UN organization. In 2005, Annan appointed Ruggie as the UN Secretary-General’s Special Representative for Business and Human Rights, tasked with proposing measures to strengthen the human rights performance of the global business sector. In 2011, the UN Human Rights Council, in an unprecedented step, unanimously endorsed the “Guiding Principles on Business and Human Rights” that John developed through extensive consultations, pilot projects and research. The ‘Ruggie Principles’ have since made their way into numerous national legislations and have helped shape the progress of human rights in business in the 21^st century.

John was a Fellow of the American Academy of Arts & Sciences and received numerous awards from academic and professional societies for his contributions to social science, public policy, and the development of international law. During the last years of his life, John chaired the board of the non-profit organization Shift, and served on the Board of Arabesque where his wit and principles-based approach to markets guided strategies and operations.

Georg Kell, Chairman of Arabesque and founding Executive Director of the UN Global Compact, today said:

“John Ruggie was my mentor and friend for nearly three decades. I can think of no other person whose mind and heart could span so many geographies and disciplines. During his years at the UN, it was not difficult to locate him in the building – one had to just follow the sound of his unique laughter, which cheered up diplomats and bureaucrats alike! And more recently, it was energizing to engage with John on the role of new technologies for a more sustainable world. He had a beautiful and curious mind, but most of all a kind heart.”

“John’s career as an academic and a social innovator is an inspiration to all young people who look for purpose. His principles-based approach to all matters of decision-making is a role model for us all.”

John, you will be greatly missed. We will not forget you, and we will continue to grow your legacy. The good fight for fairness, respect, understanding and reason will go on.

INSIGHTS News

Regulators around the world are keeping busy with rolling out numerous ESG disclosure and reporting recommendations to help markets avoid greenwashing and ensure the long-term risk management of sustainability factors. Here is this month’s round-up of sustainable finance regulatory updates:

European Union

In June, the European Commission launched the EU Taxonomy Compass. The online tool, accompanied by a downloadable Excel file, digitises the contents of the Taxonomy, focusing on the EU Taxonomy Climate Delegated Act as a first step. The Taxonomy Compass is searchable by activity and related criteria across all taxonomy economic sectors, allowing users to identify and monitor the applicable Taxonomy screening criteria to relevant investment or economic activities.

The European Banking Authority (EBA) issued a report on ESG risk management and supervision for credit institutions and investment firms. The document provides guidelines on the definition, management and governance of ESG factors and the integration of ESG risks into the regulatory and supervisory framework for regulated institutions. Importantly, the report features a non‐exhaustive list of ESG factors, indicators and metrics (Annex I) that could serve as an indication of the quantitative metrics credit and investment firms would need to report.

UK takes further steps towards introducing its own Green Taxonomy

In November 2020, the UK government announced plans to develop a UK green classification system for sustainable activities. On June 9, the UK government launched an independent Green Technical Advisory Group (GTAG) to advise on the standards for defining green investment. The GTAG will oversee, among other things, the delivery of a “Green Taxonomy”—a common classification system defining the criteria investments need to meet before they can be labelled environmentally sustainable. 

United States

On June 16, the U.S. Congress, by a narrow 215-214 vote, passed H.R. 1187, the Corporate Governance Improvement and Investor Protection Act. The act contains 11 titles pertaining to different ESG, climate, and sustainability disclosure topics. The legislation would require public companies to annually disclose certain ESG metrics and “their connection to the long-term business strategy” and now moves to the US Senate. This package includes a number of bills authored by several Members of the Financial Services Committee, featuring:

• Mr. Vargas’ bill, “The ESG Disclosure Simplification Act,” which requires public companies to disclose certain ESG information to shareholders as well as the impact of the ESG policies on their strategies;
• Mr. Foster’s bill, “The Shareholder Political Transparency Act,” which requires public companies to submit quarterly reports to the SEC on any and all political expenditures, including dark money;
• Ms. Velazquez’s bill, “The Greater Accountability in Pay Act,” which sheds light on pay disparities, helping to close the gender and racial pay gap.
• Ms. Axne’s bill, “The Disclosure of Tax Heavens and Offshoring Act,” requires disclosures that discourages companies’ use of tax havens and encourages repatriation of taxes to the United States; and
• Mr. Casten’s bill, “The Climate Risk Disclosure Act,” requires disclosures that encourages companies to plan for the impact of climate change on their company.

Singapore

Singapore’s Green Finance Industry Taskforce (GFIT), led by the Monetary Authority of Singapore (MAS) issued the following:

• White paper on scaling green finance in the real estate, infrastructure, fund management, and transition sectors.
• Framework to help banks assess eligible green trade finance transactions
• Detailed implementation guide for climate-related disclosures by financial institutions

This follows the Singapore Green Taxonomy proposal issued earlier this year which was subject to consultation by the Green Finance Industry Taskforce (GFIT), convened by MAS, in addition to the handbook on implementing environmental risk management for asset managers, banks, and insurers.

India

The top 1000 Indian companies by market cap will be required to use a specific format and follow new ESG reporting requirements when making their sustainability disclosures, as announced by the Securities and Exchange Board of India (SEBI). The Business Responsibility and Sustainability Reports (BRSRs) will feature disclosures relating to ESG risks, sustainability performance, environmental metrics, and social factors.

Central Banks

Central bankers around the world are focusing on introducing dedicated green policies as climate change and ESG continue to have a significant economic impact:

• The Central Bank of Japan (BOJ) introduced a new policy initiative, with a new lending facility to be launched focusing on climate change. The facility will be launched later in the year.
• Brazil’s central bank Banco Central do Brasil is stepping up its ESG regulatory framework. To do so, it is putting together a list of ESG risks that banks will have to incorporate into their credit models.
• The European Central Bank (ECB) gathered in Frankfurt to examine inflation targets and the possibility of introducing climate risk disclosures for asset purchases. Interestingly, the ECB discussed central bank policy and sustainability, with a view to adjusting monetary policy in line with sustainability considerations.

The EU’s Renewed Sustainable Finance Strategy

On 6th July, the EU unveiled the latest update to its Sustainable Finance Strategy, focusing on six sets of actions:

1. Extend the existing sustainable finance toolbox to facilitate access to transition finance
2. Improve the inclusiveness of small and medium-sized enterprises (SMEs), and consumers, by giving them the right tools and incentives to access transition finance.
3. Enhance the resilience of the economic and financial system to sustainability risks
4. Increase the contribution of the financial sector to sustainability
5. Ensure the integrity of the EU financial system and monitor its orderly transition to sustainability
6. Develop international sustainable finance initiatives and standards, and support EU partner countries

Particularly noteworthy are the proposal for a European Green Bond Standard, as well as the adoption of the EU Taxonomy Delegated Act that clarifies what information should be disclosed by corporates and financial institutions under Art. 8 of the EU Taxonomy:

• The European Green Bond Standard proposal, also adopted on 6th July 2021, will create a high-quality voluntary standard for bonds financing sustainable investment. The Commission proposed a Regulation on a voluntary European Green Bond Standard (EUGBS). This proposal will create a high-quality voluntary standard available to all issuers (private and sovereigns) to help financing sustainable investments.

• Finally, the Commission adopted a Delegated Act on the information to be disclosed by financial and non-financial companies about how sustainable their activities are, based on Article 8 of the EU Taxonomy. The Delegated Act specifies the content, methodology and presentation of information to be disclosed by large financial and non-financial companies on the share of their business, investments or lending activities that are aligned with the EU Taxonomy.

INSIGHTS Research

There is much excitement in the machine learning community surrounding language models (LMs), neural networks trained to “understand” the intricacies of language, semantics, and grammar. These have revolutionised natural language processing (NLP). In this newsletter we’ll go over what they are, some examples of what they can do, and ethical implications to their use, that as a community we must consider.

LMs transform sentences into numerical (vector) representations, which are subsequently used as inputs to a more traditional machine learning model, such as classification or regression. They do this by modelling the statistical distributions of words in sentences; they are trained to predict the most likely words at a given position in a sentence given the surrounding context. The LM does a lot of heavy lifting in terms of finding useful and relevant representations of language, using the most efficient representation of the meaning of a sentence using a handful of real numbers.

Leveraging this approach, the introduction of BERT (Bidirectional Encoder Representations from Transformers) in 2018 by Google researchers [1] constituted a serious paradigm shift, outperforming the previous state-of-the-art LMs on eleven language modelling challenges. BERT was 7.7% better than the competition on the GLUE sentence modelling metric, which evaluates models’ ‘understanding’ of test sentences, which had previously been dominated by a type of recurrent neural network called an LSTM (Long-Short Term Memory). A lot of the success can be attributed to the introduction of a powerful new neural network architecture known as a Transformer, which has been widely adopted in other NLP frameworks, computer vision, and time-series modelling. Transformers are now a state-of-the-art neural architecture, bringing performance gains over traditional sequence models due to computational and data efficiency.

A major advantage of using LMs is that only a relatively small amount of labelled data is required to solve a supervised learning problem. Raw, unlabelled data is used to train the LM, for example the text of Wikipedia articles or Reddit posts – you just need a very large corpus of human-written text. Once the LM “understands” language, you can fine-tune a specific task with a handful of manually labelled examples to get good results. For example, only a handful of Amazon reviews, labelled by their ‘star’ rating, are required to train a product sentiment classifier if a LM is used. BERT is now used in production for almost every Google search query [2].

After BERT, which had 340 million parameters and was trained on about 16 GB of text, there was a quick succession of bigger and bigger LMs (with strictly increasing improvements in sentence “understanding” metrics as well as intended task performance). A few examples include:

• (2019) XLNet (340 million parameters, 126 GB of data)
• (2019) MegatronLM (8.3 billion parameters, 174 GB of data)
• (2020) T-NLG (17 billion parameters, 174 GB of data)
• (2020) GPT-3 (175 billion parameters, 570 GB of data)

GPT-3 is broadly the state-of-the-art, and is essentially an LM trained on the entire internet. It is an extremely powerful tool, and as a result of being publicly available, has been used for some mind-blowing applications (see here for some examples: https://www.educative.io/blog/top-uses-gpt-3-deep-learning).

GPT-3 based models can also do things like generate code (HTML, CSS, JavaScript) for websites that have been given a sentence prompting, as well as generate narrative stories, and even ask ‘Albert Einstein’ to teach us about astrophysics [3]:

Me: Teach me about black holes.

Albert Einstein: A black hole is formed when a star has burnt all of its fuel. There are two ways this can happen. If the mass of the star is less than a certain limit, known as the Chandrasekhar Limit, then it can end in a Type 1a supernova explosion. When the mass of the star is above that limit, however, it can collapse in on itself. The reason for this is that the star is being pulled in on itself by its own gravity. It is being accelerated due to its own gravity. When a force is applied to something that is being accelerated, its momentum increases. The more massive the object is, the more force it takes to keep it from slowing down.

Me: what is Chandrasekhar Limit?

Albert Einstein: The Chandrasekhar Limit is 1.44 solar masses. An object with a mass less than this limit will not collapse into a black hole. If it is heavier than this limit, then its gravity will not be strong enough to fight against itself and it will collapse into a black hole.

Me: who is Chandrasekhar

Albert Einstein: Chandrasekhar was a scientist who calculated this mass limit.

Me: When did Chandrasekhar die?

Albert Einstein: He died in 1995 at the age of 84.

While this is clearly extremely impressive technology, there are ongoing concerns about the environmental and ethical consequences of such powerful software. Training a BERT model (by now considered a relatively small LM) has been estimated to consume as much energy as a trans-American flight [4]. This is exacerbated by the fact that the model is often trained a few times to trial different hyperparameters. Cloud computing companies generally use some renewable energy sources and/or carbon credit offsetting, but the majority of energy used is non-renewable [5].

Furthermore, and potentially more worrying, Bender et al [5] note that the datasets used to train massive LMs vastly overrepresent racist, misogynistic, and white-supremacist views, which they suggest is a result of the predominance of this sort of text on the English internet. Machine learning models cannot be separated from their training data, and essentially replicate the patterns observed in training. McGuffie & Newhouse [6] show that it is relatively easy to use GPT-3 to generate large quantities of grammatically coherent, racist, or extremist text which can then be used, for example, to swiftly populate forums and message boards, with the intent to radicalise human readers.

The AI community has yet to agree on approaches for addressing such problems, but the consensus will likely involve a push towards better curated training data for powerful models. For example, Google have pushed this forward in image-based training data by releasing the ‘More Inclusive Annotations for People’ image dataset. This changes labels of humans within images from (person, man, woman, boy, girl) to (person), with secondary gender labelling of (predominantly feminine, predominantly masculine, or unknown) and age labelling of (young, middle, older, or unknown) [7]. On the NLP side, the ‘Translated Wikipedia Biographies’ dataset aims to provide a mechanism for assessing common gender errors in machine translation, such as an implicit grammatical assumption that ‘doctor’ refers to a man [8].

In this month’s Arabesque AI newsletter, we’ve discussed language modelling, some powerful examples of its use, and highlighted a handful of concerns toward their use. There’s no doubt that LM technology is extremely powerful and effective at the task it has been trained to perform, but as a community we must be aware of potential ethical caveats, as well as the evolution of real-world dangers.

Dr Tom McAuliffe – with thanks to Dr Isabelle Lorge (both Arabesque AI)

References

[1] Vaswani, A., N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. “Attention is all you need.” In NIPS. 2017.

[2] https://searchengineland.com/google-bert-used-on-almost-every-english-query-342193 (accessed 26/06/21)

[3] https://news.ycombinator.com/item?id=23870595 (accessed 26/06/21)

[4] Strubell, Emma, Ananya Ganesh, and Andrew McCallum. “Energy and policy considerations for deep learning in NLP.” arXiv preprint arXiv:1906.02243. 2019.

[5] Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big??.” In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, pp. 610-623. 2021.

[6] McGuffie, Kris, and Alex Newhouse. “The radicalization risks of GPT-3 and advanced neural language models.” arXiv preprint arXiv:2009.06807. 2020.

[7] Schumann, Candice, Susanna Ricco, Utsav Prabhu, Vittorio Ferrari, and Caroline Pantofaru. “A Step Toward More Inclusive People Annotations for Fairness.” arXiv preprint arXiv:2105.02317. 2021.

[8] https://ai.googleblog.com/2021/06/a-dataset-for-studying-gender-bias-in.html (accessed 26/06/21)

INSIGHTS News

• Bril to join as CEO of Arabesque Asset Management after the end of his tenure as CIO of the United Nations Joint Staff Pension Fund (UNJSPF).
• Hasselgren joins Arabesque as Head of Nordics and Head of Europe for Corporates and Sovereigns. Formerly served as Global Head of Sustainability and Impact Investment at Danske Bank, and Global Head of RI Strategy and ESG Integration at Institutional Shareholder Services (ISS).
• Appointments are part of a series of significant new hires across Arabesque as the Group expands its sustainability and AI solutions to meet growing demand.

5 March, 2021, London, Frankfurt and Stockholm – Arabesque has today announced the appointment of Herman Bril, outgoing CIO of the United Nations Joint Staff Pension Fund (UNJSPF), as the new CEO of Arabesque Asset Management, and Ulrika Hasselgren, former Global Head of Sustainability and Impact Investment at Danske Bank, as the Group’s new Head of Nordics as well as Head of Europe for Corporates and Sovereigns. Ms. Hasselgren joins with immediate effect, with Mr. Bril’s role commencing in July 2021, and both will be partners at Arabesque.

The appointments of Bril and Hasselgren, who bring with them a wealth of experience from leadership roles at prominent financial institutions, comes as Arabesque expands its services to meet rapidly increasing demand for sustainability and AI technology solutions across global markets.

Mr. Bril will be responsible for Arabesque’s global asset management business and its growing suite of AI-driven sustainable investment products and solutions, including the firm’s new net-zero climate investment strategy that launches this year for clients. He will be based at the Group’s new headquarters in London.

In addition to overseeing Arabesque’s range of ESG data and insights services for corporate and sovereign clients in Europe, Ms. Hasselgren will lead the group’s activities and expansion in the Nordic region and will be based in Stockholm.

Today’s announcement marks the latest in a series of significant new hires at Arabesque, as it looks to shape the industry of sustainability through its investment solutions, AI and financial technology expertise.

With over 25 years of experience in international financial institutions in multiple countries, Mr. Bril possesses broad-level expertise encompassing asset management, pension funds, life insurance, investment banking, derivatives trading and treasury operations. Joining UNJSPF as Chief Investment Officer in 2016 in New York, Bril successfully developed and implemented the fund’s sustainable investment strategy, with assets under management increasing from USD 52 billion to USD 82 billion during his tenure over nearly five years.

Speaking about today’s announcement, Mr. Bril said:

“Technology and data are playing a key role in reshaping sustainable investing, spurring market transformation away from industrial-era concepts towards future-fit models and new horizons.”

“Powered by the market-leading ESG data assets of S-Ray and the group’s AI technology, Arabesque can help drive change through its autonomous, sustainable investment products and solutions.”

“I am thrilled to be joining Arabesque’s exceptional team of sustainable investment engineers, and scaling-up the delivery of competitive ESG strategies that reflect client values whilst delivering market returns.”

Prior to his role at UNJSPF, Herman Bril served as Group CFO and Managing Director at Cardano Risk Management, and was Senior Vice President, Head of Treasury and Capital Management at Aegon NV, and Head of Asset Management and Chief Investment Officer of Syntrus Achmea Asset Management, where he was responsible for Dutch pension funds with assets under management totalling Euro 45 billion.

Ulrika Hasselgren is a senior business leader and pioneer in sustainability, with more than 20 years in the finance industry working with investors and asset managers across the Nordics, Europe and North America. Most recently, Ms. Hasselgren served as Global Head of Sustainability and Impact Investment and Member of the Wealth Management Leadership team at Danske Bank in Copenhagen. In this role, Hasselgren created a vision and strategic roadmap to embed sustainability into the core of the bank’s investment business and led the journey for ESG quant integration and active ownership across its investment processes, products, and advisory.

Upon joining Arabesque, Ulrika Hasselgren today said:

“Over the coming few years, technology will transform the global marketplace, with artificial intelligence shaping the rapid evolution of financial services from online banking to asset management, and ushering in the next generation of ESG products.”

“The emergence of technology has also enabled the integration of sustainability within business and investment processes at a greater scale than ever before, and Arabesque will play a key leadership role in this transition through its unique range of AI-based ESG solutions.”

“I’m excited to be joining Arabesque’s fast-growing team with its collaborative and entrepreneurial culture, and I look forward to bringing innovative new data, insights and investment products to the Nordic market.”

Before her role at Danske Bank, Hasselgren joined the Institutional Shareholder Services (ISS) leadership team as Global Head of RI Strategy and ESG Integration, after Ethix SRI Advisors, the company she co-founded in 1999, was acquired by ISS in 2015. Ms. Hasselgren has developed sustainability policies, investment strategies and products, created standard-setting research methodologies, and developed tools and education programs for asset owners and asset managers. She has served on the boards of Eurosif, Dansif, and Swesif, and chaired The LuxFlag ESG Label Eligibility Committee.

Georg Kell, Chairman of the Arabesque Group and founding Executive Director of the United Nations Global Compact, added:

“A more sustainable financial system is starting to emerge, with technology and digitalisation the key drivers shaping the future of ESG investing.”

“I am thrilled to welcome Herman Bril and Ulrika Hasselgren to Arabesque, with their experience of successfully developing and implementing sustainable investment strategies at various financial institutions. As the company scales up its products and services worldwide to meet market demand, their vision and leadership will be invaluable.”

Established in 2013, Arabesque is a leading provider of technology solutions for sustainable finance, with three integrated businesses offering ESG investment strategies, data and insights for financial decision-making, and AI investment technology.

INSIGHTS Opinion

Towards successful AI applications in the financial industry

Technological advancements within Artificial Intelligence (AI) continue to play a major role in shaping global competitiveness and productivity as AI becomes more mainstream. The presence of AI in pressing global social issues such as privacy issues and Covid-19 testing, has led to an increased focus by policy makers and the public.  Examples include:

• The European Union (EU) aims to establish a reputation for “trustworthy AI” that is “made in Europe” with a strong emphasis on ethical and human-centric approaches, aligned with human rights values and democratic principles. 
• In December 2020 the UK House of Lords Liaison Committee called on the government to better coordinate its AI policy and the use of data and technology by national and local governments.
• UK regulators have set up the first meeting of the AI Public Private Forum with the aim to assess the impact AI has on financial services.
• EU adoption of a new digital finance law last September.
• SEC’s Strategic Hub for Innovation and Financial Technology (FinHub) is currently running virtual meetings with industry representatives to better understand how AI is used to, for example, build new products, offer new services, create efficiencies, and enhance regulatory compliance.

Due to AI’s self-learning nature, challenges around model explainability (i.e., the ability to discern the model’s reasoning), data integrity and privacy need to be addressed for successful applications of AI.  Arabesque AI offers these solutions for consideration in an effort to narrow the knowledge gap between practitioners, policy makers, and the public.

• Model Risk Management:

Challenge: Model explainability presents a unique challenge for AI applications, particularly in the space of Deep Learning, where non-linear dependencies can occur. For complex models, such non-linearities operate as black boxes making it difficult to explain model reasoning.

Proposed solution: In our opinion a comprehensive risk model should focus on model development, evaluation and validation as well as ongoing testing and monitoring; paired with transparency of research conducted to address the importance of explainability. AI models learn underlying distributions from the data used to train them and thus particular focus should be placed on tail-end events (e.g., unprecedented market conditions), in order to assess how well these models can generalise to data not seen or under-represented in the training process.

• Data Integrity:

Challenge: Last month’s newsletter highlighted issues around data bias for successful AI applications that require large amounts of data to train and validate.

Proposed solution: It is our belief, continuous monitoring and evaluation of input data is imperative. Building models trained to detect data bias or abnormal data can help in building more robust AI pipelines. Furthermore, recent studies report that more diverse teams of researchers are better equipped to spot data bias, highlighting the need for some level of human oversight of AI applications.

Currently we are observing a trend to include alternative data sources into AI pipelines e.g., social media, news or geo-spatial data. Data source verification is a vital step in the data pipeline to guarantee reliability, especially as we are seeing a rise in media manipulation campaigns. In our belief, a robust data ingestion, verification and storage pipeline will help safeguard AI applications and lead to more robust AI deployments.

• Privacy

Challenge: Recent publications highlight how AI can use customer data such as credit card transactions. As such data sources get incorporated into AI applications, data privacy and security considerations need to be addressed.

Proposed solution: In our opinion, the protection of personal data is a key responsibility of AI adopters and paramount to building trust in the technology. Various techniques exist within AI that safeguard personal data. This should be incorporated within all stages of model development, from idea generation to the public released of the product itself. Relevant legislation includes European General Data Protection Regulation (GDPR) that mandates strict conditions whilst using personally identifiable data.

While AI holds tremendous potential to radically transform the financial industry, it also presents unique challenges, risks and regulatory considerations. It is vital to continuously assess AI strategies on firm level as well as industry and national levels. Only by building trust in robust AI applications can the transformative power of AI be leveraged for good. 

Arabesque AI

More on Arabesque AI:

Arabesque has been built on the two disruptors of finance, sustainability and Artificial Intelligence. Utilizing advancements in technology, as an organization, Arabesque seeks to deliver transparent, sustainable, innovative solutions for our clients; whether through our SRay® data services, investment solutions or, most recently, our AI research. Arabesque AI was established in late 2019, with a minority stake owned by the asset manager DWS, with the mission to build a world-leading, AI-driven, investment technology company that offers its clients high-performing, efficient and individually customizable investment strategies. The investment philosophy underpinning the use of AI, is that the discernible structure in financial markets is highly complex and varies over time, markets, and asset classes. AI can thus be used to build systems capable of handling this complexity and of enabling scalable investment process design for a wide variety of use cases in an efficient and cost-effective way.

INSIGHTS News

• A HSBC Group General Manager and Global Head of Sustainable Finance Dr Daniel Klier to join Arabesque in June 2021

• Dr Klier will lead the firm’s global growth strategy, with sustainability and technology set to transform financial markets over the next decade

• He will be responsible for the expansion of S-Ray’s ESG services worldwide, whilst delivering new investment and technology solutions to Arabesque’s global clients

• Appointment is part of a series of significant new hires across Arabesque in the coming months as the group expands its sustainable finance offerings

12 February, 2021, London and Frankfurt – A HSBC Group General Manager and Global Head of Sustainable Finance, Dr Daniel Klier, will join as the new CEO of Arabesque S-Ray GmbH and President of the Arabesque Group, effective in June 2021.

Dr Klier’s engagement is a key part of Arabesque’s growth strategy, as the company expands its services to meet rapidly increasing demand for sustainability and AI technology solutions in financial markets.

He will lead the expansion of Arabesque S-Ray’s ESG data, advisory and research services worldwide, alongside bringing to market innovative new tools for capturing, analysing and reporting sustainability data.

He will also be scaling up Arabesque’s global group business, as it looks to shape the industry of sustainability through its investment solutions, AI and financial technology expertise.

Speaking about today’s announcement, Dr Daniel Klier, Arabesque S-Ray’s incoming CEO and President of the Arabesque Group, said:

“Finance is entering one of the most profound decades of change in its history, with sustainability and technology driving transformation of the global marketplace.”

“Arabesque is uniquely positioned to lead this major transition, combining cutting-edge AI applications, market-leading data assets and ESG investment propositions.”

“I am excited to be joining an incredible team that is supported by a world-class board of sustainability experts, together with some of Europe’s largest financial institutions as shareholders, and look forward to driving Arabesque forward as a pioneer in sustainability.”

Established in 2013, Arabesque is a leading provider of technology solutions for sustainable finance, with three integrated businesses offering ESG investment strategies, data and insights for financial decision-making, and AI investment technology.

Arabesque S-Ray has strategic partnerships with shareholders including Allianz X, Commerz Real AG, DWS Group, Helaba Digital, and the German federal state of Hessen. In 2020, it was announced that DWS had taken a minority stake in Arabesque AI.

Daniel joined HSBC in 2013 as Group Head of Strategy in London, where he oversaw the development of HSBC’s global strategy, capital allocation, and the implementation of the Group’s strategic projects.

Since 2017, he has led HSBC Group’s efforts to become a global leader in sustainability, as Global Head of Sustainable Finance, covering the Group’s business activities around climate change and other ESG-related topics, climate risk management and stakeholder engagement, resulting in numerous industry awards. He has also chaired various influential groups on sustainable finance, including the Bank of England Climate Risk working group, the Institute of International Finance (IIF) Sustainable Finance working group, and the UK Finance Sustainability Committee. Prior to joining HSBC, Daniel was a Partner at McKinsey & Company.

Georg Kell, Chairman of the Arabesque Group, and founding Executive Director of the United Nations Global Compact, said:

“Technology is revolutionising sustainable investment, and democratising finance as a force for good to shape a more prosperous future for our society and the planet.”

“I am thrilled to welcome Daniel to Arabesque, and believe that by adding his experience, leadership and vision to the team, we could not be better placed to meet our mission of mainstreaming sustainability across capital markets.”