smart data webinar: a roadmap for deploying modern ai in business

30
Copyright (c) 2016 by STORM Insights Inc. All Rights reserved. A Roadmap for Deploying Modern AI in Business Adrian Bowles, PhD Founder, STORM Insights, Inc. [email protected]

Upload: dataversity

Post on 15-Feb-2017

977 views

Category:

Technology


4 download

TRANSCRIPT

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

A Roadmap for Deploying Modern AI in Business

Adrian Bowles, PhDFounder, STORM Insights, Inc.

[email protected]

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

A Roadmap for Deploying Modern AI in Business

Context - How Did We Get Here? (and where are we anyway?)

Elements of Modern AI Are You Ready?

Evaluating Application Requirements Evaluating Organizational Attributes

Context - How Did We Get Here? (and where are we anyway?)

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

Context - How Did We Get Here? (and where are we anyway?)

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

AI Roots

AGI - Artificial General Intelligence Focus on replicating intelligence by copying brain functions

Natural Language Processing (NLP) Learning and discovery

Heuristics, expert rules… Logic - symbolic logic and mechanical theorem proving

Strategy: Replace Execution: Open concepts Constraint: Processing

Modern AI

Focus on augmenting intelligence by evidence-based interaction

Natural Language Processing (NLP) Learning and discovery

Distributed ML driven by big data Deep QA techniques

Strategy: Reinforce Execution: Open code and data Constraint: Data

A Quick Guide to AI Progress/Maturity

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

AI Roots

AGI - Artificial General Intelligence Sloooooow…still crawling Focus on replicating intelligence by copying brain functions Still trying to understand nature

Natural Language Processing (NLP) Great results, practical apps Learning and discovery Solid results using supervised

Heuristics, expert rules… and unsupervised ML algorithms Logic - symbolic logic and Well understoodmechanical theorem proving

Strategy: Replace Execution: Open concepts Constraint: Processing

A Quick Guide to AI Progress/Maturity - Key Disruptors

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

Autonomous systems (based on ML)Self-driving carsDrones

Augmented Intelligence systems Medical diagnosticsComplex systems maintenance and materials planning

Agents for personal productivityApple SiriMicrosoft CortanaGoogle Now

Time

Mat

urity

Machine Learning

Human

Sensors/Systems

Infrastructure

Input Output

Visualization

Narrative Generation

Voice/NLP

Video/Images

Reports

Gestures

Emotions

Text/NLP

Surface Structured DataSurface Structured Data

Reports

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

DataManagement

Alt/NeuromorphicHardware

ProfessionalServices

Perc

eptio

n

Elements of Modern AI

0. Foundation

Experience-Based

Learning1. Learn

2. Interact

3. ExpandIntegrate

Augmented/VirtualReality

Confidence-weightedReporting

Motivation

reflection

inference

Natural Cognitive Processes

deduction

Hypothesis Generation& Testing

reasoning

Natural Language Processing

Cloud

…Analytic

s

Data Management

Neu

rom

orph

icAr

chite

ctur

es Learning

Perception

A Framework for Cognitive Computing

Copyright (c) 2015-2016 by STORM Insights Inc. All Rights reserved.

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

Open Source and Modern AI…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

Apache Spark

Spark MLlib (distributed ML framework)GraphX (distributed graph processing framework)

Commercialized by Databricks, …

Open Source and Modern AI…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

In the news…

Are You Ready for Modern AI?

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

The goal is to assess the fit between application requirements and available modern AI tools/technology.

Evaluating Application Requirements

Evaluating Organizational AttributesThe goal is to assess organizational-readiness to effectively leverage

available modern AI tools/technology.

Tools

Technology

Talent

Evaluating Application Requirements

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

As with all new technology, adding an AI-based application to your portfolio requires at least a preliminary a risk-assessment exercise. Is the application:

Critical to the enterprise? Time sensitive? All or nothing? (can we define incremental releases to mitigate risk?) Visible? (success AND failure)

Evaluating Application Requirements

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

Look at the data, NLP requirements, learning requirements and user-comfort with ND* responses.

For each of the key questions in this section, answer Critical (not worth building the app without this feature/function) Nice to have (this would be a differentiator, but we can do without…for now) Won’t be missed or noticed by our users Unsure

*ND -non-deterministic - responses, used to contrast probability-based answers with those based on certainty.

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

Is a specific modern AI capability required, or can conventional tools and techniques adequately address your data acquisition, management and analysis requirements?

Evaluating Application RequirementsLook at the data, NLP requirements, learning requirements and user-comfort with ND* responses.

What percentage of the data required for this application comes from Highly structured data (records from tables, formatted directories, databases...)

Internal (data that is owned or licensed, such as customer data or mailing lists) Third-party data that needs to be procured Unsure

Text (books, magazines, newspapers, social media feeds...) Your internal data (data that is owned or licensed)

Third party data that needs to be procured Unsure

Audio, video Your internal data (data that is owned or licensed) Third party data that needs to be procured

Unsure

Critical Nice to have Won’t be missed Unsure

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

How important is it for your customers/users to be able to interact with this application in NL? Will multi-language support be required now or in the future?

What level of language sophistication is expected from the users, and what level is required from the application?

Are the NL requirements critical for Input only? Output only? Conversation to refine an interactive session?

Evaluating Application RequirementsLook at the data, NLP requirements, learning requirements and user-comfort with ND* responses.

Critical Nice to have Won’t be missed Unsure

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

How important (and feasible) is the ability to provide the “right answer” rather than several alternatives, ranked by confidence or probability?

How important is it for the application to be able to explain how it arrived at its answer(s)? (will the user trust or question the results, how much interaction will be required vs how much will they want). Is the risk of a “bad” answer quantifiable? (from customer satisfaction to medical malpractice to financial loss in a stock recommendation engine)

Evaluating Application RequirementsLook at the data, NLP requirements, learning requirements and user-comfort with ND* responses.

Critical Nice to have Won’t be missed Unsure

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

How important is it for this application to improve its performance by learning from its own experience rather than being updated by your own team?

Evaluating Application RequirementsLook at the data, NLP requirements, learning requirements and user-comfort with ND* responses.

Can you update it fast enough to be useful (consider the rate of new data and changed data that must be analyzed to be effective)?

Is this application in a regulated domain?

Do you have data that can be used to train the system?

Critical Nice to have Won’t be missed Unsure

Evaluating Organizational Attributes

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

The goal is to assess organizational-readiness to effectively leverage available modern AI tools/technology.

Evaluating Organizational Attributes

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

For each of the modern AI tools/technologies identified as critical in the previous assessment step - e.g. NLP, machine-vision/video analysis, data management including knowledge of ontologies, taxonomies, etc, machine learning techniques (supervised, unsupervised, reinforcement..),

determine…

Actual skill set requirements (including level with technologies vs tools) Available skills

Internally Partners Contractors Market pool

Tools

Technology

Talent

Anyone can buy tools, talent is the hard part.

For more information:

Copyright (c) 2016 by STORM Insights Inc. All Rights reserved.

[email protected]

Twitter @ajbowles Skype ajbowles

Upcoming Webinar Dates & Topics

March 10 Machine Learning Adoption Strategies Theme: Cognitive Computing

April 14 Getting Started with Streaming Analytics and the IoT Theme: Smart Data and the Internet of Things

May 12 Emerging Data Management Options: Graph Databases Theme: Smart Data Management