Designing IA for AI - Information Architecture Conference 2024
CDAC presentation as part of Global AI Festival and Future
1. Dr. S D Sudarsan
Executive Director (C-DAC, Bangalore)
19 February 2024
Generative AI
Opportunities & Challenges
2. Centre for Development of Advanced Computing (C-DAC)
Vision
To emerge as the premier R&D
Institution for the design,
development and deployment of
world class electronic and ICT
solutions for economic & human
advancement.
Professionals
Locations
Year Established
1988 12 6000+
Premier R&D organization of the Ministry of
Electronics & Information Technology (MeitY) to
carry out R&D in ICT, Electronics & Associated
Domains
3.
4. C-DAC Bangalore
C-DAC, Knowledge Park
No.1, Old Madras Road
Byappanahalli
Bengaluru - 560038
C-DAC, Electronic City
No 68, 1st Phase
Electronics City
Bangalore - 560100
10. What is Generative AI? (Contd.)
Generative Artificial Intelligence (AI) is type of AI that are designed to generate novel, original content
including text, images, audio, code, and videos, in response to prompts.
GenAI
DL
ANN
Statistical
ML
12. GenAI – Taxonomy of Applications
Type Example Application
01 – Advisory AI that, given access to a
company’s financials, can provide
fiscal/tax guidance.
02 – Assistive An AI that can book all aspects of
business travel based on a calendar
invite.
03 – Cooperative Involves a “back and forth”
protocol where generative AI and
the user work collaboratively
towards a goal. (e.g. Github
Copilot)
04 – Augmentative AI that equips a UX designer to
produce a functioning app from
their design tool for smart policing.
05 – Digitally
Autonomous
An AI agent that acts as an elite
cybersecurity officer.
06 – Physically
Autonomous
A humanoid AI capable of
performing Policing.
Moving towards digitally and/or physically
autonomous systems… Source: nuvalence.io
13.
14.
15.
16.
17. Can you Recognize this Gentleman?
“This Person Does not Exist”
Generated by AI
https://thispersondoesnotexist.com/
18. What is the next scene?
GenAI assisted screenplay
18
22. GenAI and Cybersecurity Opportunities
By learning the patterns of network traffic,
generative models can flag unusual behavior
that might indicate a cyberattack.
Generative adversarial networks (GANs) models
can create realistic phishing pages to test an
organization’s defenses or identify phishing
websites in real time.
Simulating cyberattack to train organization's
incident response teams and practice responses
to different cyber threats.
Creates synthetic datasets for training the
models in a controlled and privacy-preserving
manner. Test and evaluate security solutions
without using real, potentially sensitive data.
Creates decoy assets or honeypots within a
network to attract attackers.
Anonymize sensitive data before it is shared or
analyzed, ensuring that privacy regulations are
followed while still allowing for meaningful
analysis.
Create visual representations of complex
cybersecurity data, making it easier for security
analysts to identify patterns and anomalies in
large datasets.
Identifies and blocks inappropriate or harmful
content in emails, chat messages, or social
media.
Generative models can analyze patterns in
passwords and generate recommendations for
stronger, more secure passwords.
Generates variations of known malware
samples to stay ahead of evolving malware
threats by creating new signatures and
improving detection capabilities.
25. Case Study 3: GenAI Attacks on Tuberculosis Detection using CT Scan
Ref: Li Y eta al, 2023
Invisible Data
Perturbation
Original CT scan Poisoned CT scan