At the first of three showcases at VentureBeat’s Transform 2019 AI event, five companies — CognitiveScale, Applied Brain Research, integrate.ai, AI Foundry, and D-ID — took the stage to tell their stories and share their latest news.
CognitiveScale creates enterprise AI software — Cortex — that’s intended to bring about “practical, scalable, trusted AI solutions” for organizations. Its latest product, Cortex Certifai, is an auditing tool for AI that uses AI. “We have been hearing from numerous industry analysts and influencers that AI that is ethical, transparent and explainable is among the top three considerations when it comes to taking decisions on AI implementation in an organization,” the company told VentureBeat.
Because so much of AI is a black box, auditing AI is a unique challenge. Cortex Certifai checks for model and system vulnerability and creates an AI Trust Index, and it doesn’t require granular access to models and thus protects an audited company’s IP. It’s also designed to be model-agnostic and user friendly, thereby opening up the tool to more users.
From the Transform 2019 stage, chief revenue office ML Maco announced that CognitiveScale is working with the government of Canada on an air cargo project. They needed to check for cargo risk — currently a time-consuming manual task that is subject to errors — and required a way to understand the data and determine whether the information on a given shipment’s label is reliable, incomplete, and so on. “When you can do that,” Maco said, “You can be sure the system is fair, robust, and explainable.” To address those issues, CognitiveScale developed Cortex Certifai.
Applied Brain Research
Applied Brain Research works on the edge with Nengo, its AI OS designed after the human brain that developers can use to add intelligence to everything from cards to drones to robots. Nengo is interoperable with both neuromorphic chips and more traditional CPUs and GPUs, but its magic is extreme efficiency, which it achieves without a lot of horsepower.
The next step for the company is commercializing the SpiNNaker 2 neuromorphic chip to create a complete edge platform. Pegged for release in 2021, the company describes it as “next generation, low-power, real-time,” and it’s designed to bring power efficiency and processing speed to edge devices. In the future, Applied Brain Research expects to be able to integrate multiple networks to achieve sensorized speech, gesture, and images all on one chip.
Privacy is a huge challenge in AI right now, and it butts up against the need for companies to have AI-powered insights for customer intelligence. Integrate.ai unveiled a new Customer Intelligence Platform that promises predictive analytics for specific business objectives like improved digital customer interactions, and it does so without using or exposing any personally identifiable information (PII). The new tool adds to integrate.ai’s SaaS platform and is designed to build trust with end customers.
As founder and CEO Steve Irvine explained, it’s difficult to compete with large tech companies; you need to leverage an existing tech stack. The Customer Intelligence Platform runs on top of a company’s existing technologies, and it provides a simple dashboard interface that’s readable by anyone on a company’s team, not just IT or engineers. You can set targets, and the tool will show you a number of variations so you can see different permutations on how to achieve your business goals, with pros and cons clearly outlined. Then, you can use the monitor tool to see how accurate the predictions were after projects are complete.
AI Foundry is focused on classifying documents, extracting any pertinent data, and inputting that information into a core system of record. Instead of having humans perform this task manually or use current OCR methods, the company wants to save time and reduce errors using AI.
AI Foundry recently built an “Intelligent Process Automation Mortgage Document Library.” The library leverages machine vision to recognize incoming documents with a purported 95% accuracy. Although the tool is specifically for mortgages, AI Foundry hopes to apply the technology to all document-intensive industries.
Since AI Foundry launched its cloud app, it’s secured 10 new customers who have achieved ROI in about six months.
D-ID, as its name implies, de-identifies photos so facial recognition technology can’t work on them. The company replaces all personally identifiable information, such as faces and license plates, with anonymized computer-generated data. Existing solutions rely on blurring or pixelation, while D-ID anonymized faces retain a natural complexion and key attributes like age, gender, gaze, and so on.
Adding to its de-identification capabilities, D-ID discussed its news smart anonymization capability. It’s designed for video — in all its forms, but primarily as it pertains to applications like smart cities and companies that store people’s user-generated video. The assumption is that those companies should, but likely won’t, protect your face data. D-ID can replace your actual face with a computer-generated avatar that retains demographic information but strips out your vulnerable “faceprint.”