SD Times Open-Source Project of the Week: IBM’s Uncertainty Quantification 360
IBM recently announced the open source release of its new toolkit, Uncertainty Quantification 360 (UQ360). This comprehensive toolkit redirects the focus of AI to measuring for levels of uncertainty, which can ultimately save developers and users time and effort.
Released at the 2021 IBM Data and AI Digital Developer Conference, Uncertainty Quantification 360 was created to provide developers with cutting edge technology and algorithms that serve to quantify the uncertainty of learning models while still in the development process. UQ360 is built to improve the communication between the AI and the person using it.
According to IBM, when professionals have access to high-quality uncertainty estimates, they can make important decisions in collaboration with AI, resulting in the best possible outcome. Uncertainty Quantification 360 has the intelligence to measure its own uncertainty so that the person using it is able to react accordingly rather than relying solely on the reports of AI.
The UQ360 Toolkit brings ease to users by providing the estimation, measurement, improvement and communication of UQ all in one place. It was also developed with a common interface and versatility in mind, making it effective to use with several different platforms. In addition, the UQ360 Toolkit places an emphasis on communication and allows for developers to choose an appropriate style of communication to best serve their efforts.
This technology can effectively function in several important fields, such as: medical technology and self-driving cars. With areas such as these, too much uncertainty can lead to fatalities, and that’s what UQ360 works to avoid.
In terms of the medical field, AI can help to identify the proper candidates for a drug trial while UQ360 works in the background measuring the level of certainty with which the AI is making these life-changing decisions. If it is determined that the uncertainty estimate is too high, doctors know that the candidates must be looked over again.
The same applies to self-driving cars and communicating the confidence of the AI with the driver so they can be certain that the car is making the right decisions. In these situations, using an AI without a proper uncertainty estimate can result in life altering technology based mistakes, according to IBM.
Credit: Source link