Unraveling Multimodality with Large Language Models.pdf
from data to artificial intelligence, from laboratory to industry
1. From Data to AI, from Laboratory to Industry
a lesson learned from an Indonesian University
widyawan@ugm.ac.id
2. • Director of IT of UGM, Yogyakarta
• Assistant professor in Information
Technology
• Member of Forum Masyarakat Statistik
Bapennas
• Co-founder of Datains, a Big Data Analytic
company
3. • UGM is situated in Yogyakarta, Java island
• a state owned university, established in 1949
• 20 schools, +250 study program
• offer vocational, undergrad and postgrad
• + 55.000 students, +2500 faculty member
4. “In God we trust, all others must bring data.”
W. Edwards Deming
5. Agenda
• Data and AI
• University Research in AI
• Use Case
• from Research to Startup
10. Why Data Silos
• Structural
• Government and organization increasingly divided into units and teams
• Often left to implement their own process and software
• Creates data separation
• Ex.: BPS, Dukcapil, MDP, Data Kesehatan
• Social
• No incentives for sharing data
• People do not share data to assert control and power
• Technological
• Legacy system
• Vendor lock-in
11. How
• Stop silos mentality
• Clear regulation in data transparency
• Encourage open data
16. current analytic technology
data decision action
descriptive
what happened?
diagnostic
why did it happened?
predictive
What will happen?
prescriptive
what should I do?
machine human human still doing
most of the process
17. descriptive analytics is still the most successful
http://meetings2.informs.org/analytics2013/execforum.html
18. • Data and AI
• University Research in AI
• Use Case
• from Research to Startup
19. Publication in AI
• Keyword: artificial intelligence, machine learning, deep learning, neural
network
• 3557 paper out of 897,406 papers (1951-2020)
No Country No of Papers
1 USA 193613
2 China 188337
3 UK 57827
39 Thailand 3715
40 Norway 3621
41 Indonesia 3557
42 New Zeeland 3461
43 Pakistan 3398
44 South Africa 3215
No University No of Papers
1 UI 409
2 ITB 406
3 UGM 221
24. Funding
• Government has spend money
• Many international support
• Funding from commercial entity is limited or non-existent
25. • Data and AI
• University Research in AI
• Use Case
• from Research to Startup
26. Deep Learning with
Edge Computing
• Convolutional Neural Network
• CPU based – Open Vino
• Vehicle detection, counting
• Road-scene segmentation
Credit to:
Computer
Vision
28. Segmentation and Classification of
Cervical Cells using Deep Learning
IEEE Access, Nia K., Khalid H., Lukito, Widyawan, Lutfan .L, Anton, S. Tedy M.
30. Sentiment, Radicalism and Hoax Analysis
• Lexicon, Log
regression
• SVM, Random
Forest, Naive
Bayes
• LSTM
• CNN
31. • Data and AI
• University Research in AI
• Use Case
• from Research to Startup
32. Technology Readiness Level (TRL)
research at universities product sold by companies
TRL
Resources
innovation commercialization
AI Startup
investor
government
existing research
resources
industry
34. Take Out
• Encourage open data for better AI innovation
• Indonesia is still in the middle of the table of AI publication league
• To be innovative, Indonesian company need support AI research more
• AI use case: edge computing, computer vision, NLP
• AI startup should be encouraged and supported