Artificial intelligence in the post-deep learning era
STI Sequences, Innovation Networks, and Accelerators
1. STI Sequences, Innovation
Networks, and Accelerators
Modeling, measuring and visualizing innovation as sequences
of connected activities and networks of connected actors
C. Scott Dempwolf, PhD
Assistant Research Professor
& Director
July 7, 2015
UMD – Morgan State Joint
Center for Economic Development
2. What is network analysis?
Network analysis models,
measures and visualizes systems of
human interaction in ways that
include specific nodes (people,
organizations, places, documents,
industries, etc.) and specific ties
(relationships between nodes).
York County, PA industries and occupations 2008
with sales and employment flows
Military armored
vehicle manufacturing
Motorcycle manufacturing
Snack food manufacturing
Printing
Air conditioner manufacturing
Motor vehicle parts manufacturing
Production
3. Innovation as a sequence of activities
Research Question
What are the sequences of STI
and commercialization
activities involved in
developing new ideas
through basic research and
translating them into new
products or services in the
marketplace?
Pending NSF application with UMIACS
Professor Ben Shneiderman
EventFlow Screenshot
Innovation is a process
comprised of a sequence of
activities leading to a new
product in the marketplace
Research Publication Invention Proof-of-Concept Commercialization Product
4. Seeing innovation activities as networks
Each source document generates an activity network
Data
Sources
Activity
Type
Activity
Networks
Funding
Agency
Research
Institution
PI / Inventor PlaceCompany
Commercializatio
n docs
Research Publication Invention Proof-of-Concept Commercialization Product
NIH
NSF
NASA
StarMetrics
Citations
Bibliometrics
Patents
Copyrights (future)
SBIR / STTR Phase
1
Accelerators
Web
SBIR / STTR Phase
2
CrunchBase
Trademarks
UPC / SKU
SBIR / STTR
Web
Product doc
Commercializatio
n docs
Commercialization
Docs
Commercializ
ation docs
Commercializ
ation docs
POC Docs
Commerciali
zation docs
Commerciali
zation docs
Patent Docs
Commercializ
ation docs
Commercializ
ation docs
Publication
Docs
Commercializ
ation docs
Commercializ
ation docs
Research
Grant Docs
$$
Document
Ties
5. Seeing innovation activities as networks
Combining activity networks yields an innovation process network
Document
Ties
Innovation
Process
Network
Activity
Networks
Funding
Agency
Research
Institution
PI / Inventor PlaceCompany
Commercializatio
n docs
Product doc
Commercializatio
n docs
Commercialization
Docs
Commercializ
ation docs
Commercializ
ation docs
POC Docs
Commerciali
zation docs
Commerciali
zation docs
Patent Docs
Commercializ
ation docs
Commercializ
ation docs
Publication
Docs
Commercializ
ation docs
Commercializ
ation docs
Research
Grant Docs
$$
Patent
New Product
Journal
Publication
Prototype
6. Emerging
Theory and
Research
Illinois Battery Cluster 2010 - 2014
There are two main components of
the innovation network…
• A Research component comprised
mostly of researchers and research
institutions
• An Industry component
comprised mostly of investors,
serial entrepreneurs, and
supporting infrastructure for
commercialization
… and a “Bridge” connecting them;
the bridge includes accelerators,
POC centers, TTO’s, etc. and their
connections to brokers in each
component.
7. The Innovation Ecosystem
and the Valley of Death
Adapted from Jackson, 2011
This is a network
representations of
the two sides of the
so-called valley of
death
8. Accelerators as Bridge Structures
• Accelerators and specifically founders
are brokers
• Their social capital is more valuable to
startups than the money invested
• That social capital is limited to specific
markets, technologies and investors
where they have experience and
relationships
• The accelerator environment is designed
to create and recreate social capital in
three networks: cohort, capital and
market
Accelerator founder and
venture capital / market
network
Accelerator and startup
cohort network
9. Accelerators as Bridge Structures
Visualizing CrunchBase Data
77 accelerators and random sample of 77 angel investors
not affiliated with an accelerator, and their networks
When clustered, many accelerators emerged as central
nodes while only a few unaffiliated angels were central
The accelerator subnetwork was 8.5 times larger than
the unaffiliated angel network and exhibited more
opportunity for brokerage
Accelerators invested 33% less per startup in angel
funding ($100K vs $150K) and 50% less overall ($1.3B vs
2.6B) than unaffiliated angels. Combined their startups
raised an additional $41B in subsequent funding rounds
and acquisitions
Data source: https://www.crunchbase.com/
10. Emerging Theory and Research
(working hypothesis)
Bridging structures (accelerators, TTO’s
and inter-mediaries) between the two
components are sparse and vary by
technology and geographic region
Regions with denser, more connected
bridging components will be
characterized by faster innovation
sequences and more innovation
sequences leading to new products.
Illinois Battery Cluster 2010 - 2014
12. Accelerators: A Network Perspective
Visualizing Global Capital Flows
Specific places are recognized as national
and global hubs for innovation.
These places both attract more capital
investment and in turn make more
investments both locally and globally
Venture capital is clearly sticky in well
known places – the SF Bay area, Boston /
Cambridge, New York, Chicago,
London, LA…
Data source: https://www.crunchbase.com/
13. Why networks & technology matter
• Startups need to be seeded into
strong clusters
• Clusters have strong connections to
markets, supply chains and talent
pools
• Clusters form around technologies
• Capital flows around technologies
14. Policy Considerations
• Networking & social capital are central components;
relationships are essential
• Networks are specific and unique to the technology field
• Current successful accelerators are built around
technologies with relatively low cost entry; short
prototyping timeframe; high scalability; low geographic
& supply chain constraints
• Accelerators are in the business of picking (probable)
winners
Editor's Notes
About me
Founder and Chief Technology Officer, Tertius Analytics
C. Scott Dempwolf is an Assistant Professor in the Urban Studies and Planning Program at the University of Maryland, directs an EDA-funded University Center and teaches a planning studio. His research and publications examine relationships between innovation, economic development and industrial land use. Scott returned to academia after 20+ years of community and economic development practice, earning his PhD in Urban Planning at UMD; a Masters from Temple and a Bachelor’s from MIT.
Some familiar examples include manufacturing supply chains, or the input-output table or transaction table that shows flows between industries.
Innovation is a process comprised of a sequence of activities
This sequence includes research, publication, invention, proof-of-concept, and commercialization leading to a new product in the marketplace
Current research (NSF pending) models and analyzes these sequences using new EventFlow software to provide first-ever statistical analysis of +/- 5,000 innovation sequences
What are the most common sequence patterns? How many activities are involved? How long do they take?
Do sequences and their characteristics vary by technology focus or geographic region?
How many and which activities get public funding, and how much? Are there correlations between the characteristics of public funding and sequence outcomes?
Are there gaps or lags in the sequence patterns? If so, where in the sequence and how long do they last?
Each activity in an innovation sequence has an activity network associated with it – the network of people and organizations involved in funding and completing the activity
People and organizations in activity networks are also connected to place nodes (place of residence, place of work or establishment location) at the time of the activity
People and organizations may also be connected to document nodes through publication authorship
People and organizations may be connected indirectly (weak ties) based on, for example, similarities in the technologies that they are working on
Activities in an innovation sequence are connected by citations (backward linkages)
Activities may also be indirectly connected (weak ties) through shared people or organizational nodes
Aggregating the activity networks for all the activities in an innovation sequence produces an innovation process network
Activity networks and innovation process networks are embedded in and drawn from a larger innovation network
Activity networks vary based on the type of activity – the people and organizations involved in research and development activities tend to be different from those involved in commercialization activities
Adding the investment dimension
This slide was originally presented at the Industry Studies Association Conference in May, 2014.
This slide was originally presented at the Industry Studies Association Conference in May, 2014.
It does not represent the “two main components” but rather just shows the commercialization components represented in the CrunchBase data.
Activity networks vary based on the type of activity – the people and organizations involved in research and development activities tend to be different from those involved in commercialization activities