August 4, 1997: Skynet goes online August 29, 1997, 2:14am ET: - - PowerPoint PPT Presentation

▶
august 4 1997 skynet goes online august 29 1997 2 14am et
SMART_READER_LITE
LIVE PREVIEW

August 4, 1997: Skynet goes online August 29, 1997, 2:14am ET: - - PowerPoint PPT Presentation

Sep 21, 2023 100 likes •678 views

August 4, 1997: Skynet goes online August 29, 1997, 2:14am ET: Skynet gains consciousness August 29, 1997: Judgement Day Terminator (1984) What features does T-800 have? What features does T-800 have? Vision/Perception

slide-1
SLIDE 1
slide-2
SLIDE 2
slide-3
SLIDE 3

August 4, 1997: Skynet goes online

slide-4
SLIDE 4

August 29, 1997, 2:14am ET: Skynet gains ‘consciousness’

slide-5
SLIDE 5
slide-6
SLIDE 6
slide-7
SLIDE 7

August 29, 1997: Judgement Day

slide-8
SLIDE 8
slide-9
SLIDE 9
slide-10
SLIDE 10
slide-11
SLIDE 11
slide-12
SLIDE 12

Terminator (1984)

slide-13
SLIDE 13

What ‘features’ does T-800 have?

slide-14
SLIDE 14

What ‘features’ does T-800 have?

  • Vision/Perception
  • Speech
  • Motion/Planning
  • Object Manipulation
  • Self-Repair
  • Goal-directed Exploration
  • Social Skills
  • Ethics/Morality
slide-15
SLIDE 15

Back to our timeline..

slide-16
SLIDE 16

1960’s (1st Wave)

  • Single Layer networks
  • XOR problem killed research for two decades
slide-17
SLIDE 17

Mid-1980s (2nd Wave)

  • Multi-layer networks
  • Backpropagation algorithm
slide-18
SLIDE 18

2010s (3rd Wave)

  • Big Data

– O(M) labeled images

  • Big Compute
  • ‘Deep’ Learning
slide-19
SLIDE 19
  • 19 -
slide-20
SLIDE 20
slide-21
SLIDE 21

ImageNet Architecture

slide-22
SLIDE 22

Deep Learning for Computer Vision

slide-23
SLIDE 23
slide-24
SLIDE 24

Deep Learning for Segmentation

slide-25
SLIDE 25

Slide Courtesy of Nervana Systems

slide-26
SLIDE 26

Deep Learning for Speech

slide-27
SLIDE 27
slide-28
SLIDE 28
slide-29
SLIDE 29

IBM Watson

slide-30
SLIDE 30

DARPA Robotics Challenge

slide-31
SLIDE 31

MIT Cheetah

slide-32
SLIDE 32

Deep Learning for Object Manipulation

slide-33
SLIDE 33

DARPA Autonomous Driving Challenge

slide-34
SLIDE 34

Deep Learning for Self-Driving Cars

slide-35
SLIDE 35

Can Deep Learning Work for Science?

  • Similarities

– Tasks:

  • Pattern Classification
  • Regression
  • Clustering
  • Feature Learning
  • Anomaly Detection
  • Differences

– Unique attributes of Scientific Data

  • Multi-channel / Multi-variate
  • Double precision floating point
  • Noise and Artefacts
  • Statistics are likely different
slide-36
SLIDE 36

CAM5 0.25-degree simulation data

slide-37
SLIDE 37

Challenge: Multi-Variate Data

slide-38
SLIDE 38

Task: Find Extreme Weather Patterns

slide-39
SLIDE 39

Supervised Learning

  • Training Input: Cropped, Centered, Multi-variate

patches with Labels*

–Tropical Cyclone (TC) –Atmospheric River (AR) –Weather Front (WF)

*Labels are provided by TECA, which in turn implements

human-specified criteria

  • Output: Binary (Yes/No) on Test patches

– Is there a TC in the patch? – Is there an AR in the patch? – Is there a WF in the patch?

slide-40
SLIDE 40

Logistic Regression K-Nearest Neighbor Support Vector Machine Random Forest ConvNet Test Test Test Test Test Tropical Cyclone 95.85 97.85 95.85 99.4 99.1 Atmospheric Rivers 82.65 81.7 83.0 88.4 90.0 Weather Fronts 89.8 76.45 90.2 87.5 89.4

Supervised Classification Accuracy

slide-41
SLIDE 41

Semi-Supervised Learning

  • Objectives:

– Want to predict bounding box location for weather pattern – Might have few/no labels for several weather patterns; want to discover new patterns – Create unified architecture for all weather patterns

slide-42
SLIDE 42

Semi-Supervised Convolutional Architecture

Encoder Decoder Classification + Bounding Box Regression

slide-43
SLIDE 43

Reconstruction Results

slide-44
SLIDE 44

Classification + Regression Results

Ground Truth Prediction

slide-45
SLIDE 45

Deep Learning for Science

Modeling galaxy shapes Oxford Nanopore sequencing Decoding speech from ECoG Generating cosmology mass maps Clustering Daya Bay events LHC Signal/Background classification

slide-46
SLIDE 46

Deep Learning Stack

Technologies Deep Learning Frameworks Multi Node libraries Single Node libraries Hardware

MLSL MPI GRPC MKL-DNN Torch, Neon, CNTK, MXNet, … CuDNN MKL CPU GPU FPGA …

slide-47
SLIDE 47

Deep Learning: Open Challenges

  • 1. Performance and Scaling
  • 2. Complex Data
  • 3. Hyper-Parameter Optimization
  • 4. Scarcity of Labeled Data
  • 5. Interpretability and Visualization
slide-48
SLIDE 48
slide-49
SLIDE 49

Assumptions…

  • Communities will self-organize and conduct labeling

campaigns

– Active Learning systems can determine optimal strategies for seeking labels

  • Incorporation of domain science principles into

learning algorithms

– Solution spaces that satisfy physical constraints

  • Pattern Classification, Clustering, Anomaly

Detection are solved problems

slide-50
SLIDE 50

What is the role of humans?

HPSS

/project

Labels Patterns, Clusters, Anomalies Mechanisms, Hypothesis

slide-51
SLIDE 51

How close are we to creating a T-800?

  • Vision/Perception 👎 😁 🎊
  • Speech 😁 🚁 💦
  • Motion/Planning 👎 🚁
  • Object Manipulation 😑
  • Self-Repair ☹
  • Goal-directed Exploration 🙂
  • Social Skills 🤑
  • Ethics/Morality 😲
slide-52
SLIDE 52

How about the T-1000 and T-X?

slide-53
SLIDE 53

Time Travel?

slide-54
SLIDE 54

Time Travel? Let’s skip this for the time being…

slide-55
SLIDE 55

Conclusions

  • AI appears to be working

– Genuine breakthroughs in vision, speech, control – Wide range of commercial applications

  • Tremendous potential for scientific applications

– Low-hanging fruit, but hard questions are coming next

  • NERSC is at the forefront of Deep Learning for

Science

– Applications – Production stack – R&D on optimizations + scaling, methods

slide-56
SLIDE 56

Thanks!

  • Contact: prabhat@lbl.gov
  • Connect on LinkedIn
  • Internships, full-time opportunities