Accelerating Transformation Artificial Intelligence Uli Homann CVP Cloud & AI June 2019
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Accelerating Transformation
Artificial Intelligence
Uli HomannCVP Cloud & AI
June 2019
Great AI Needs Great Compute
DNN Processing Units
EFFICIENCY
10
FLEXIBILITY
Soft DPU(FPGA)
Control Unit (CU)
Registers
Arithmetic Logic
Unit (ALU)
CPUs GPUsASICsDPU
CerebrasEyerissGoogle TPUGraphcoreGroqIntel NervanaMythicWave ComputingetcMicrosoft
BrainWave
Baidu SDADeephi TechEff. Speech EngineTeradeepetc
Great AI Needs Great Data
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
012
020406080
100120140160180
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
Zett
abyt
esData growth projectionsIDC predicts Global DataSphere will grow from 33 Zettabytes (ZB) in 2018 to 175 ZB by 2025
Figure 1: Annual size of the Global DataSphere175 ZB
Source: Data Age 2025, sponsored by Seagate with data from IDC Global DataSphere, Nov 2018
KilobyteMegabyteGigabyteTerabytePetabyteExabyteZettabyte
DNA storage
Encoding
AGTCTAG
0011010111000100111010001111
Sequencing
Random access
(read)
AGTCTAG
Decoding
0011010111000100111010001111
Synthesis
(write)
Preservation
Joint project between Microsoft and University of Washington
>1,000x denser than tape
2,000 years in ~10˚C
Station B at Princeton
15
Great AI Needs Great Data But How Much do you Really Need
There are three main types of damage detected
ALGAE CRACKS RUST
Visualising the data
Recent Successes of Reinforcement Learning
“Closed-world” systems make simulating
millions of trials easy.
20
It’s not easy!
Atari Games ~ 200M frames= ~77 Days @ 30 frames/sec Alpha Go ~ 130M self-plays
Pacman ~ 150M frames
Great AI Needs Great Values
Today’s debate
Fairness
Reliability & Safety
Source: Rand Corp. Driving to Safety; Kara & Paddock
Miles & years needed for to demonstrate autonomous vehicle reliability
Privacy & Security
We knew that if we could identify them in their second trimester, there’s a good chance we could capture them for years
““
Inclusiveness
Economic Graph Research
Transparency HasAsthma(x) => LessRisk(x)
Accountability
“We can’t afford to look at this future with uncritical eyes.”– The Future Computed
Great AI Needs Great People
Microsoft Research AI(MSR AI)
Pool talent from multiple areas of AI
Natural language
Perception Learning Reasoning
[ UBIQUITOUS COMPUTING][ AMBIENT INTELLIGENCE]
Related Documents
