AI platform MioTech has secured investments from J.P. Morgan Asset Management and HSBC Asset Management’s Financial Technology Venture Capital Fund. The new capital will go toward accelerating talent acquisition, expansion into new international markets, as well as developing existing and new product lines. With increasing regulatory pressure and evolving investor, corporate and consumer demand, MioTech expects to capture fast-growing market needs for green tech and sustainability solutions with its proprietary technologies. MioTech’s ability to utilize AI to gather and assess unstructured ESG data sets will provide valuable inputs to fundamental research for J.P. Morgan Asset Management. This will enhance J.P. Morgan Asset Management’s consideration of material ESG risks and opportunities in its active investment processes and help support its global and China clients in making sustainable investment decisions. With valuable expertise into the ESG data and factors that are unique to China, MioTech’s capabilities will also help to advance the development of proprietary quantitative ESG scoring system.
RBC’s assistant NOMI gain 1.5 million active users with Forecast feature for bills reminder and cash planning
- Growing use of RBC’s digital assistant NOMI justifies the big Canadian bank’s four-year investment in advanced AI and machine learning for improving consumers’ financial lives. The technology enables several predictive features including a new weekly cash-flow tool. Not only are the number of users increasing, but they’re spending more time using features of NOMI. 37% of Americans are living on the financial edge and 14% are living paycheck to paycheck. Experts have been predicting that advanced technology could be a boon for consumers in this situation. RBC’s NOMI now addresses this need specifically. More than half of NOMI users say it is, and less than 1% have left RBC since starting to use the app. A newly added cash-flow forecasting feature could add to the momentum. The bank’s 17 million customers can now get the answers to everyday cash-flow questions like those above using NOMI. Built using machine learning and artificial intelligence software, NOMI Forecast tells RBC customers how much money they will have over a seven-day period based on their preauthorized payment withdrawals.
- The NOMI platform already provides customized insights to customers to create simplified budgets and automated savings. NOMI has about 1.5 million active users. The “active” designation is significant because tallies of users typically include the people who may have downloaded an app, but never actually used it. Further, 53% of NOMI users say the virtual assistant is a game changer for their finances, the bank states. This is partly because RBC has been steadily adding features to the platform. With NOMI Forecast, consumers are less likely to miss paying a bill or run short of cash before payday. NOMI Forecast addresses pain points that consumers have about finances. Over three quarters of Americans feel anxious about their financial situation and more than half (58%) feel that finances control their lives. NOMI Forecast gives customers back that control. NOMI Forecast uses technology from Borealis AI, a research center backed by RBC that is focused on using ML: and AI in financial applications. NOMI Forecast marks the first time Borealis has been integrated into a retail banking offering.
AI model optimization tech creates several computation routes, optimized for an inference device and dataset
Startup Deci is developing a platform to optimize machine learning models. By leveraging data science techniques, Deci developed products to accelerate AI on hardware by redesigning models to maximize throughput while minimizing latency. Deci achieves runtime acceleration on cloud, edge, and mobile through data preprocessing and loading, automatically selecting model architectures and hyperparameters (i.e., the variables that influence a model’s predictions). The platform also handles steps like deployment, serving, and monitoring, continuously tracking models, and offering recommendations where customers can migrate to more cost-effective services. Deci’s platform offers a substantial performance boost to existing deep learning models while preserving their accuracy. It designs deep models to more effectively use the hardware platform they run on, be it CPU, GPU, FPGA, or special-purpose ASIC accelerators. The accelerator is a data-dependent algorithmic solution that works in synergy with other known compression techniques, such as pruning and quantization. Deci’s solution is an engine — Automated Neural Architecture Construction, or AutoNAC — that redesigns models to create new models with several computation routes, optimized for an inference device and dataset. Each route is specialized with a prediction task, and Deci’s router component ensures that each data input is directed via the proper route.
Automated “cognitive data capture” tech focuses on the full document journey
- Startup Rossum, is a platform used by enterprises to automatically process and “understand” all their inbound documents. Rossum leans on deep learning to automate manual document processing and data entry. Powered by integrations with myriad enterprise resource planning (ERP), robotic process automation (RPA), and document management systems (DMS), Rossum uses “cognitive data capture” to read and understand the content of a document just like a human does. Rossum facilitates not only the data extraction, but also the data validation and correction through what it calls “spatial OCR (optical character recognition),” which learns to recognize different structures and patterns of different documents — an invoice number might be on the top left-hand side in one invoice, but somewhere completely different in another.
- Using a “human-in-the-loop” system to learn, Rossum also promises to improve over time. Rossum has transitioned from a “document understanding engine” into what it now calls itself an “all-in-one document gateway.” This includes receive, which is all about unifying all mail channels across paper and digital by connecting inboxes, scanners, and DMS systems; understand, which is concerned with data capture; and communicate and act, which automates workflows such as notifying the document’s senders that it has been received or whether there are any issues.
AI model optimization tech creates several computation routes, optimized for an inference device and dataset
Startup Deci is developing a platform to optimize machine learning models. By leveraging data science techniques, Deci developed products to accelerate AI on hardware by redesigning models to maximize throughput while minimizing latency. Deci achieves runtime acceleration on cloud, edge, and mobile through data preprocessing and loading, automatically selecting model architectures and hyperparameters (i.e., the variables that influence a model’s predictions). The platform also handles steps like deployment, serving, and monitoring, continuously tracking models, and offering recommendations where customers can migrate to more cost-effective services. Deci’s platform offers a substantial performance boost to existing deep learning models while preserving their accuracy. It designs deep models to more effectively use the hardware platform they run on, be it CPU, GPU, FPGA, or special-purpose ASIC accelerators. The accelerator is a data-dependent algorithmic solution that works in synergy with other known compression techniques, such as pruning and quantization. Deci’s solution is an engine — Automated Neural Architecture Construction, or AutoNAC — that redesigns models to create new models with several computation routes, optimized for an inference device and dataset. Each route is specialized with a prediction task, and Deci’s router component ensures that each data input is directed via the proper route.
Ally’s chatbot is able to answer 65% of customer questions without the need for human assistance
Sathish Muthukrishnan, chief information, data and digital officer for Ally Financial, said it is no longer enough for FIs to simply augment existing tools because smooth, instant access to online financial solutions has become the norm for consumers. A world where real-time, personalized banking services are table stakes means that FIs must find new ways to keep customers’ attention and differentiate themselves. This makes it critical for them to examine AI-supported tools or voice-activated technologies, such as smart speakers, as consumers seek ways to make their banking experiences as convenient as possible. Ally is leaning on various technologies to help keep customers engaged as the digital-first banking era takes shape. Sixty-five percent of customer questions that come to the company’s AI -supported digital assistant, Ally Assist, are currently being answered by the chatbot without the need for human assistance.
Privacy-preserving ‘Federated Learning’ tech does not require data transfer directly to a central database
- The Hong Kong Applied Science and Technology Research Institute (ASTRI) joins forces with tech-embracing companies to leverage a privacy-preserving technology, called “Federated Learning”, to develop AI models and output in the form of encrypted parameters that serve as a reference for financial institutions to conduct comprehensive credit analyses for micro, small and medium-sized enterprises (MSMEs) to help them get access to financing. Unlike traditional machine-learning methods, Federated Learning does not require data to be transferred directly to a central database, thus protecting privacy and mitigating the risk of data security breaches.
- Data partners and financial institutions can establish common credit evaluation models by combining their encrypted parameters. During the process, the collaborators do not have access to any consumer personal data, nor are the identities of the enterprises identified. Only when an enterprise applies for financing and is undergoing authorisation can the designated financial institution obtain the relevant parameters and conduct a credit evaluation.
Technology completely automates the entire real estate appraisal process while offering Compliance Guarantee
Global DMS’ EVO platform is revolutionizing the way lenders manage collateral by streamlining the entire appraisal process through workflow automation, programmable AI assignment, transparent communication, and LOS connectivity. EVO also supports key back end processes, including secure automated borrower credit card processing and compliant appraisal delivery. Global DMS’ EVO appraisal management platform completely automates the entire real estate appraisal process, while providing the industry’s only Compliance Guarantee Program. Best of all, EVO is completely user configurable—ensuring a fast, personalized, and pain-free implementation that can be measured in days. Lenders looking for a plug-and-play solution to address their numerous appraisal woes need to look no further, as EVO provides the ability to easily implement and change their valuation process without costly and time-consuming internal IT projects, no matter their business model—whether that’s managing several Appraisal Management Companies (AMCs) or their own appraisal panel directly.
Pruning, novel accelerator h/w, meta-learning and neural architecture help curb model training costs
- Microsoft and Nvidia have trained what they claim is one of the largest and most capable AI language models to date: Megatron-Turing Natural Language Generation (MT-NLP). MT-NLP contains 530 billion parameters — the parts of the model learned from historical data — and achieves leading accuracy in a broad set of tasks, including reading comprehension and natural language inferences. Since 2012, the amount of compute needed to train a model to the same performance on classifying images in a popular benchmark — ImageNet — has been decreasing by a factor of two every 16 months. Approaches like network pruning prior to training could lead to further gains. Research has shown that parameters pruned after training, a process that decreases the model size, could have been pruned before training without any effect on the network’s ability to learn.
- Called the “lottery ticket hypothesis,” the idea is that the initial values parameters in a model receive are crucial for determining whether they’re important. Parameters kept after pruning receive “lucky” initial values; the network can train successfully with only those parameters present. Network pruning is far from a solved science, however. New ways of pruning that work before or in early training will have to be developed, as most current methods apply only retroactively. And when parameters are pruned, the resulting structures aren’t always a fit for the training hardware (e.g., GPUs), meaning that pruning 90% of parameters won’t necessarily reduce the cost of training a model by 90%. Whether through pruning, novel AI accelerator hardware, or techniques like meta-learning and neural architecture search, the need for alternatives to unattainably large models is quickly becoming clear. Study shows that using 2019-era approaches, training an image recognition model with a 5% error rate would cost $100 billion and produce as much carbon emissions as New York City does in a month.
Diagnostics platform helps improve customer service chatbots with high-level analytics and visualizations
- AI-powered customer service SaaS startup Zowie launches Zowie Diagnostics, a free business analytics dashboard that aims to give companies a first-hand look at the value of AI-powered automation before taking the plunge on a customer service automation platform like Zowie or one of its many competitors. Once plugged into your company’s customer service portal — be it Zendesk, Intercom, Freshworks, or another one of the many Zowie-integrated platforms out there — Zowie Diagnostics will offer a comprehensive bird’s-eye view of your customer service operation, complete with high-level analytics, sleek data visualizations, and specific, actionable suggestions for how to improve things. Using proprietary AI smarts based in part on natural language processing, Zowie bulk-analyzes the contents of data sources like customer service tickets and chats to understand what customers are saying and how well the company is meeting their needs. In addition to top-level metrics like the average number of tickets per agent and typical response time, Zowie Diagnostics will also break down useful details like what percentage of customer questions are repetitive and what times of day the team is most overtaxed.
- More than just an analytics dashboard oozing with stats and charts, Zowie Diagnostics uses this bulk data analysis to call out specific customer experience shortcomings — and serve up recommendations for how to fix them. the ultimate goal of this new free offering is obvious: Propel more IT decision-makers in the direction of Zowie’s product for their conversational AI needs.