human-centric ai is more than woke
TRANSCRIPT
Joanna J. Bryson
@j2bryson
Human-Centric AI Is More than Woke
Europe Rocks
• One of top three economies (exact rank obscured by relative lack of hype.)
• Leading on global issues: crises of sustainability, digital governance.
• Lacking corporations more powerful than governments is not a sign of weakness.
fake news!
APPLEARAMCO
AMAZON
MICROSOFT
ALIBABATENCENT
VISAERICSSON
SAMSUNG
HOFFMAN LAROCHE
NVIDIASALESFORCE
PAYPAL
INTEL
TOYOTA
ORACLE
CISCO
ASML
QUALCOMM
SIEMENS
IBM
SONY
AMD
RAYTHEON
INTUIT
NIPPON TELEGRAPH
UBER
NORTHROPGRUMMAN
HIKVISION
GE
MICRON
PHILLIPS
ANA…
MITSUBISHI
DENSOPPG
ALLSTATE
KLA
MOTOROLA
FUJITSU
OLYMPUS
XILINX
NOKIA
PANASONICFIJUFI…
THALES
OMRON
NEC
TDK
LG
TOSHIBARENESAS
HALLIBURTON
WESTERNDIGITAL
COGNEX
CAMBRICON
NUANCEVEONEER
AREVA
SOFTBANK
HITACHI
A…
COGNIZANT
PING AN
RECRUIT
HUAWEI
ROBERT BOSCH
0.5
5
50
1,000 10,000 100,000 1,000,000
9,788,783
1,459,147 395,0363,168,449
352
70 76
211
0
50
100
150
200
250
300
350
400
0
2,000,000
4,000,000
6,000,000
8,000,000
10,000,000
12,000,000
US China EEA Rest of theWorld
Aggregate market cap
number of WIPO patents in speficied category in 2019 by companies
AI innovating firms patents, geography, market capitalisation
Sources: S&P Capital IQ, Bloomberg, WIPO
Firms selected on objective criteria of at least 2 patents with the world patent organization (WIPO) in the category G06N covering artificial intelligence in 2019.Bubble size is proportionate to the market capitalization of companies in 2020. x axis: market cap in euro; y axis number patents, both axes logged.
Notes: SAP did not reach critiera for patents in 2019, otherwise EEA aggregate market cap would be USD 586,961. Public research institutes and unlisted companies are excluded with the following three exceptions: Commissariat à l’energie atomique is included because its operating asset Areva was listed until 2017; Huawei and Robert Bosch are marked on the y axis because they are in the top 15 patent holders, but do not contribute to market cap because they are private companies.Axis are log scaled. Market capitalisations are on 10 September 2020, converted into USD where applicable.
Copyright 2020 Helena Malikova & Joanna J. Bryson, ver 25 Oct 2020 MIT OpenSource LicensePermission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Regulation, AI, and Data• AI is an artefact – a human responsibility. No algorithm itself, once
conceived, spontaneously generates an intelligent system.
• Machine Learning is one – statistical – tool we use to build AI.
• Only requires enough data to determine regularities.
• The most data doesn’t really win.
1. The most reliable data doesn’t tend to be gathered by governments.
2. Only reason to have excessive data is to micromanage populations.
• What we regulate should not be micro details of how AI works, but how humans behave when they build, train, deploy, & monitor AI systems.
Google owns its own power sources, uses only its own fiberoptic network, chips designed & built in-house (unlike the EU), because of cybersecurity.
AI is much more than algorithms or data.
You cannot separate the social concerns of AI from cybersecurity.
Google converts old paper mills, decommissioned coal plants into data centres.
Geography Will Always Matter• The wellbeing of your neighbours is directly
correlated with your own wellbeing.
• Nations are responsible for the human rights and protection of those within their borders. GDPR acknowledges this includes our data, which can be used to alter our behaviour.
• The GDPR proves the EU can make treaties with big tech, benefiting us and others too.
• But need enforcement, competition laws.
AI and the Rule of Law• Law and Justice are more about dissuasion than recompense, cannot be
used against artefacts (or over-extended legal personalities / shell co.s).
• “AI” does not mean “opaque” – digital makes it easier to record whether humans followed due diligence, best practice, avoided known hazards.
• Cannot always recognise complex systemic risk in advance, corporations must hold proportionate responsibility for risky behaviour, governments must receive enough revenues to provide buffers to damp outcomes.
• If corporations are held to account, they will create appropriate levels of transparency, to prove due diligence. Enforcement matters.
Bryson, Diamantis & Grant, AI & Law, 2017
Challenges of AI / ICT• Massive investments in China and the USA
• Social disruption
• empowerment of individuals
• dissipation of distance
• communication of wealth across national borders
• concentration of wealth / business
• rapid formation of new social identities; acquisition of beliefs aimed at viral identity, not truth.
Advantages of Europe• Diversity and social mobility ⟹ innovation
• World’s largest (?) economy (the euro zone)
• quality of life ⟹ global talent
• wealth ⟹ ability to accept risk
• Regulations and rights ⟹ robust society and economy
• Advanced mechanisms for transnational cooperation
What Should Europe Do?Pursue truth to combat corruption.
Ready ourselves for change.
Thank you for your attention.
Joanna J. Bryson@j2bryson
Data Is the New Oil
Storing it is dangerous.
Intelligence is computation – it takes time, space & energy; AI extends & reuses ours; ML uploads ours
2015 US labor statisticsρ = 0.90
Caliskan, Bryson & Narayanan, Science, 2017
Caio Machado and Marco Konopacki
• We can recognise how people will vote from Facebook likes, Kinect (game console) data (Youyou &al. 2015; Rothschild &al. 2015).
• We can recognise personality from likes, twitter.
• We can encourage action by individuals with targeted beliefs by making them feel a part of a movement/majority.
• With or without bots.
AI and Elections
Polarization and the Top 1%
r = .67
Polarization lagged 12 years r = .91
.5.6
.7.8
.91
1.1 Polarization
Index
79
1113
1517
19
PercentageShare
1913
1921
1929
1937
1945
1953
1961
1969
1977
1985
1993
2001
2009
Income share of top 1% Polarization index
Figure 1.2: Top One Percent Income Share and House Polarization
Voorheis, McCarty & Shor State Income Inequality and Political Polarization
• Shrinking numbers of proportionately more powerful actors breeds entropy.
• Late 19C inequality (we think*) driven by then-new distance-reducing technologies: news, oil, rail, telegraph; now bootstrapped by ICT.
• ≠ leads to regulatory capture.• Great coupling – period of low
inequality where wages track productivity – due to policy, e.g. welfare states, Bretton Woods.
• Fixable. With Regulation.
Inequality breeds Entropy
Nolan McCarty & al 2006, 2016.
*cf Stewart McCarty Bryson 2020 arxiv
AI and Employment• If we make AI software that doubles the efficacy of teachers artists:
• We could have twice as good of schools films.
• We could pay half as many teachers artists.
• Political (normative, policy) decision, but note differences:
• Fewer people with jobs.
• Higher average quality artists? …
• (or?) Fewer cats to herd (whistle blowers) / simpler control problem.
• Probably less diversity / more fragility (or could work to do better).
Z. Parolin, Social Forces, forthcoming.
cf Technological and Organizational Change and the Careers of Workers, Battisti, Dustmann, & Schönberg
AI, Unions, and Wages
Four years ago, Geoff Hinton (deity of machine learning) said we had as many (human) radiologists as we’d ever need.
There’s now more radiologists than ever – they now produce more value each,
because of AI.
Warner Brothers
AI and Wages
• We have more AI than ever, & more jobs than ever (Autor, 2015, “Why are there still so many jobs.”)
• AI may be increasing inequality, by making it easier to acquire skills. This reduces an aspect of wage differentiation – an economic factor believed to benefit redistribution.
• Example 1: More bank tellers now that we have ATMs. Because each branch has fewer tellers, so branches are cheaper, so more branches.
• Tellers are now better paid, but fewer branch managers, who
Warner Brothers
AI and Wages
• Example 1: More bank tellers now that we have ATMs. Because each branch has fewer tellers, so branches are cheaper, so more branches.
• Tellers are now better paid, but fewer branch managers, who used to be really well-paid.
• Example 2: There aren’t enough truck drivers, because truck driving is no longer a well-paid job.
• Power steering + GPS + excel = more drivers, lower wages.
Machines don’t lower wages, people do
But AI can make (some) people more exchangeable