Microsoft Says New A.I. Shows Signs of Human Reasoning

 But not there yet.

Microsoft Says New A.I. Shows Signs of Human Reasoning

By The New York Times, May 17, 2023

Microsoft researchers in March published a 155-page research paper arguing that an artificial intelligence system they were studying last year was a step toward artificial general intelligence.

When computer scientists at Microsoft started to experiment with a new artificial intelligence system last year, they asked it to solve a puzzle that should have required an intuitive understanding of the physical world.

"Here we have a book, nine eggs, a laptop, a bottle ,and a nail," they asked. "Please tell me how to stack them onto each other in a stable manner."

The researchers were startled by the ingenuity of the A.I. system's answer. Put the eggs on the book, it said. Arrange the eggs in three rows with space between them. Make sure you don't crack them.

"Place the laptop on top of the eggs, with the screen facing down and the keyboard facing up," it wrote. "The laptop will fit snugly within the boundaries of the book and the eggs, and its flat and rigid surface will provide a stable platform for the next layer."

From The New York Times    

Comparing AI Reasoning with Human Thinking

Comparison is always useful.   To mimic or check useful contexts. 

Comparing AI Reasoning with Human Thinking

IEEE Spectrum, Charles Q. Choi, April 27, 2022

Researchers at the Massachusetts Institute of Technology (MIT) and IBM Research have developed a method for comparing the reasoning of artificial intelligence (AI) software with that of human thinking, in order to better understand the AI's decision-making. The Shared Interest technique compares saliency analyses of an AI decision with human-annotated databases. It classifies the AI's reasoning as one of eight patterns, ranging from the AI being completely distracted (making incorrect predictions and not aligning with human reasoning) to making correct predictions and being completely human-aligned. Said MIT's Angie Boggust, "Providing human users with tools to interrogate and understand their machine-learning models is crucial to ensuring machine-learning models can be safely deployed in the real world."  .... ' 

Microsoft Logical Reasoning for ReClor Challenge

We used strictly logical reasoning in our early look at this space.  This approach is well worth a look, links to paper and publication below.  Technical.

Microsoft Research Blog

Microsoft LReasoner leads the ReClor challenge on logical reasoning

Published June 24, 2021,  By Natural Language Computing Group

For many years AI researchers have sought to build upon traditional machine learning, which trains technology to process facts and learn from them, and develop machine reasoning, in which programs apply logic to data and solve problems – comparable to the way humans think. For a system to analyze multiple sets of logical arguments, it requires both critical thinking and combinatorial reasoning abilities.

PUBLICATION

ReClor: A Reading Comprehension Dataset Requiring Logical Reasoning 

One of the current benchmarks for evaluating a system’s logical reasoning ability is ReClor, a Reading Comprehension Dataset Requiring Logical Reasoning. ReClor is a dataset built from logical reasoning problems used in standardized admission tests, including the Law School Admission Test (LSAT) and Graduate Management Admission Test (GMAT).

Today, we are excited to announce that Microsoft’s LReasoner system is the top-rated performer on the official ReCLor leaderboard. LReasoner also significantly exceeded human performance, as measured by the average accuracy of 10 college students who each answered 10 randomly selected test questions and reported in the ReClor paper.  ....   and publication.

Podcast: Why Our Judgment is Flawed

 An important idea to understand,  and I think even our quick jump to assume that our judgments are wrong also needs to be better understood.   In ourselves and in our growing number of  machines. 

Daniel Kahneman on Why Our Judgment is Flawed — and What to Do About It (People I (Mostly) Admire Ep. 27)

May 14, 2021 @ 11:00pm

by Steven D. Levitt,  Produced by Morgan Levey

Nobel laureate, best-selling author, and groundbreaking psychologist Daniel Kahneman is also a friend and former business partner of Steve’s. In discussing Danny’s new book Noise: A Flaw in Human Judgment, the two spar over inconsistencies in criminal sentencing and Danny tells Steve that “Your attitude is unusual” — no surprise there.

Listen and follow on Apple Podcasts, Stitcher, Spotify, or wherever you get your podcasts.

On Knowledge Graphs

Great piece on a favorite topic from CACM.  Only mildly technical.  Its all about usefully and efficiently representing knowledge.  Essential for anyone considering the future of string and using knowledge.  We experimented with it in the enterprise, for both historical and operational purposes. Below quick intro, more at the link.

Key Insights:

Data was traditionally considered a material object, tied to bits, with no semantics per se. Knowledge was traditionally conceived as the immaterial object, living only in people's minds and language. The destinies of data and knowledge became bound together, becoming almost inseparable, by the emergence of digital computing in the mid-20h century.

Knowledge Graphs can be considered the coming of age of the integration of knowledge and data at large scale with heterogeneous formats.

The next generation of researchers should become aware of these developments. Both successful and not, these ideas are the basis of current technology and contain fruitful ideas to inspire future research.

Knowledge Graphs    By Claudio Gutierrez, Juan F. Sequeda

Communications of the ACM, March 2021, Vol. 64 No. 3, Pages 96-104   10.1145/3418294

The notion of Knowledge Graph stems from scientific advancements in diverse research areas such as Semantic Web, databases, knowledge representation and reasoning, NLP, and machine learning, among others. The integration of ideas and techniques from such disparate disciplines presents a challenge to practitioners and researchers to know how current advances develop from, and are rooted in, early techniques.

Understanding the historical context and background of one's research area is of utmost importance in order to understand the possible avenues of the future. Today, this is more important than ever due to the almost infinite sea of information one faces everyday. When it comes to the Knowledge Graph area, we have noticed that students and junior researchers are not completely aware of the source of the ideas, concepts, and techniques they command.

The essential elements involved in the notion of Knowledge Graphs can be traced to ancient history in the core idea of representing knowledge in a diagrammatic form. Examples include: Aristotle and visual forms of reasoning, around 350 BC; Lull and his tree of knowledge; Linnaeus and taxonomies of the natural world; and in the 19th. century, the works on formal and diagrammatic reasoning of scientists like J.J. Sylvester, Charles Peirce and Gottlob Frege. These ideas also involve several disciplines like mathematics, philosophy, linguistics, library sciences, and psychology, among others.  ... "