Uploaded by victor.thomas

watsonx overview - tech

advertisement
watsonx.
An overview
1
Introducing…
watsonx
Scale and accelerate the impact of AI with trusted data.
Train, tune and deploy AI across your business,
leveraging critical, trusted data wherever it resides.
IBM Marketing & Communications | 2023 | © 2023 IBM Corporation
2
Put AI to work with watsonx
Scale and accelerate the impact of AI with trusted data.
Train, validate, tune and
deploy AI models
Enterprise-ready next-generation
AI platform for builders brings
together traditional machine
learning and new generative AI
capabilities powered by
foundation models.
IBM Marketing & Communications | 2023 | © 2023 IBM Corporation
Scale AI workloads, for all
your data, anywhere
Enable responsible, transparent and
explainable data and AI workflows
Fit-for-purpose data store
optimized for governed data and
AI workloads, supported by
querying, governance and open
data formats to access and share
data.
End-to-end solution
encompassing both data and AI
governance to enable
responsible, transparent, and
explainable AI workflows.
33
IBM watsonx.ai
↓
an enterprise-ready next-generation
AI platform for builders bringing
together traditional machine learning
and new generative AI capabilities
powered by foundation models.
4
watsonx.ai
Train, validate, tune, and deploy AI models with confidence
The enterprise-ready next-generation AI platform
for builders, bringing together traditional machine
learning and new generative AI capabilities powered
by foundation models.
A proven platform for Machine Learning
Automated
Team
Collaboration Development
ModelOps
Decision
Optimization
+ new generative AI capabilities
Foundation
model Libraries
Prompt Lab
Tuning Studio
5
Model constituents- BluePile Data Lake
A curated, production-ready enterprise-relevant dataset to power IBM
foundation models
Legal
Financial
+
Regulatory
Sustainability
Cybersecurity
Code (e.g. Ansible)
100s TBs of raw data with provenance and carefully curated (e.g., hate, abuse, and profanity filters) to power
IBM foundation models
Our models today leverage about 0.5T tokens after filtering etc. We plan to scale this
to the 1-1.5T range over the next 3-6 months.
Prompting Laboratory
Available Now
Quickly iterate on zero and few-shot
prompting for natural language tasks
using free-form custom prompts.
Apply prompts to datasets at scale
using python libraries.
Coming Soon
Quickly bootstrap new prompts using
task-specific prompt templates.
Turn GUI prompts into code with 1-click.
Save and share prompts across your
organization.
7
IBM watsonx.data
↓
the only open data lakehouse
optimized for all governed data,
analytics and AI workloads
across hybrid-clouds.
8
Lakehouses are a new approach meant to combine the advantages
of data warehouses and data lakes
First generation lakehouses
still have key constraints
that limit their ability to
address cost and
complexity challenges:
• Single query engines set
up to support limited
workloads –typically just
BI or ML
• Typically deployed over
cloud only with no
support for multi-/hybrid
-cloud deployments
• Minimal governance and
metadata capabilities to
deploy across your entire
ecosystem
IBM Software | ©
9
IBM watsonx.data: The industry’s only open data store with multi-engine
support built for hybrid deployment of analytics and AI workloads.
IBM Software | ©
10
Optional
Overview of the key components of the IBM watsonx.data: multiple query engines, open
table formats and built-in enterprise governance
Access 100% of your data across
databases and data lakes
Your existing
ecosystem
Core Lakehouse functionality
Data warehouse
Data lake
Multiple engines such as Presto and Spark
that provide fast, reliable, and efficient
processing of big data at scale
Query
engines
Governance
and Metadata
Ecosystem infrastructure
Metadata store
Access control management
Built-in governance that is compatible with
existing solutions, including IBM Knowledge
Catalog
Vendor agnostic open formats for analytic
data sets, allowing different engines to access
and share the same data, at the same time
Data format
Optimize workload costs and
performance using multi-engine
functionality
Ensure governance and reduce
time to insight with centralized
metadata and access management
Access all of your data across databases
and data lakes
Storage
Cost effective, simple object storage
available across hybrid-cloud and multi-cloud
environments
Reduce storage costs and facilitate
data ingest
Infrastructure
Hybrid-cloud deployments and workload
portability across hyperscalers and on-prem
with Red Hat OpenShift
Deploy on any infrastructure and
optimize available resources
watsonx.data
11
Reduce your data
warehouse costs by
up 50%* by optimizing
workloads
Optimize workloads from your data
warehouse when you take advantage
of low-cost object storage and fit-forpurpose query engines
*When comparing published 2023 list prices
normalized for VPC hours of IBM watsonx.data to
several major cloud data warehouse vendors.
Savings may vary depending on configurations,
workloads and vendors.
12
Access all your data across
hybrid-cloud through a
single point of entry
An open data store built for hybrid
deployment of your analytics and AI
workloads
IBM Software | © 2023 IBM Corporation |
Confidential
13
Get started in minutes
with built-in governance,
security and automation.
Accelerate time to trusted analytics and AI
Connect to your existing analytics data and deploy fit-for-purpose
engines in minutes
Address enterprise compliance and security using built-in
centralized governance across your data ecosystem
Use foundation models to discover, augment, refine, and visualize
watsonx.data data and metadata
IBM
watsonx.data
Scale AI workloads,
for all your data, anywhere
IBM Software | © 2023 IBM Corporation
Reduce the cost of your
data warehouse by up to
50%* through workload
optimization across
multiple query engines
and storage tiers.
Access all your data
through a single point
of entry across all
clouds and on-prem
environments.
*When comparing published 2023 list prices normalized for VPC hours of IBM watsonx.data to several major cloud
data warehouse vendors. Savings may vary depending on configurations, workloads and vendors.
Get started in
minutes with built-in
governance, security
and automation.
15
IBM watsonx.governance
↓
an end-to-end solution
encompassing both data and AI
governance to enable responsible,
transparent, and explainable data
pipelines and AI workflows.
16
Trusted AI models are:
Fair
Address unfair bias
Are privileged groups at a
systematic advantage
compared to other groups?
Explainable
Easy to understand
outcomes/decisions
Why did the AI arrive at an
outcome? What changes
would alter that outcome?
Robust
Handle changing
model parameters
How do changing business
conditions affect
your models?
Bias detection
Ensuring fairness in model scoring
Actively monitor all model predictions and
continuous calculations on model fairness
•
•
•
•
Analyze deployed model predictions
for bias
Collect and aggregate bias data for
dashboards and alerts
Find correlations in non-feature data
Use a corrected model for
”de-biased” predictions
Model Explainability
Understand model outcomes
Explain model predictions
–
–
–
Show the most influential features
Explain in natural language
Available API for prediction explanations
What-if analysis
–
–
Experiment with values
Assess effects of changes to features
Drift detection
Handle changing scenarios
Measure the degree to which a model has
moved away from reality
–
–
Drop in accuracy – reality has
changed, as shown by the
scoring data
Drop in consistency – reality is the
same, the events vary
Drift monitoring and alerts
– Degradation of model performance
can trigger retraining and
redeployment
Automate operations to drive consistency,
efficiency and transparency at scale
Facts & Lineage
Governance
Validation
Orchestration
Automatically capture
model metadata
Manage AI model and
compliance risks
Evaluate AI
performance
Automate your AI flow
Can I explain key model
facts automatically? How
did my model change and
what data was it built
upon?
Does your AI meet
organizational and
regulatory standards? Are
your production
models monitored?
Is your model approved for
production use? Will it meet
business needs?
Can I automate my AI
lifecycle for consistency,
repeatability, and
flexibility?
IBM
watsonx.governance
Enable responsible,
transparent and explainable
data and AI workflows
IBM Software | © 2023 IBM Corporation
Trace and document the
origin of datasets,
models and pipelines —
so you can explain your
AI’s decisions, every
time
Monitor AI models for
fairness, bias and drift
— and taking action if
they go awry
Translate regulations
into automatically
enforced policies to
ensure compliance
22
Download