Artificial intelligence and Jobs Andrea Renda 23 January 2020 Ge - - PowerPoint PPT Presentation

Artificial intelligence and Jobs

Andrea Renda 23 January 2020

“Ge Genera rative Advers rsari rial N l Network rk p pri rint, on can canvas as, 2 2018, sign gned ed with GAN N model del l loss fu func nction i n in n ink b nk by the he publishe her, fr from a a ser eries es of el elev even en u uniqu que i e images ages, pu publ blished ed by y Ob Obvio vious A s Art, Pa Paris, is, w wit ith orig igin inal gil ilded wood f d fram ame. e.

Th The St Stor

• re

1.

Th The b blog

• gger

Th The D Driver

Th The P Pilot

• t

Th The b ban anker

Th The g govern rnment

Th The p platform rm

the rise of digital platforms

• The rise of platforms depends on key features of the digital economy:
• Digital information goods
• Massive data availability
• Rising computing capacity (Moore’s law)
• Artificial Intelligence
• End-to-end Internet architecture
• Modularity
• Network neutrality and absence of regulation for intermediaries

23

Moore’s Law Modularity Scale without mass

Platformization and Superstar Firms AI & data-driven economy Blockchain /DLTs

Foundations

e2e architecture

Collaborative economy p2p platforms

• Digitization and information goods
• End-to-end design (originally neutral)
• System goods and modularity/granularity
• Transition from goods to services (“age of access”)
• Increased virtualization (“softwarisation”)
• Multi-sidedness, network FX and “platformization”
• Competition for eyeballs (“attention merchants”)
• Ever-changing architecture and shifting of entry possibilities
• Big data, machine learning and data-crunching algorithms
• Dynamic pricing and price discrimination

Moore’s Law Modularity Scale without mass

Platformization and Superstar Firms AI & data-driven economy Blockchain /DLTs

Foundations Evolution

Regulatory challenges

Distancing from liability Technology is “plastic” Problems in enforcement DRMs face problems Shifting to a less open architecture Jurisdiction and applicable law Liability attribution and apportionment Problems in verification of information Possible clash b/w data protection and DLTs Problems of regulating crypto and ICOs Anti-money laundering issues Enforcement through Smart contracts Emphasis on mark-up pricing is misplaced Multi-sidedness v. cost-based regulation Privacy and data protection Free flow of data and security AI ethics, transparency and liability Data spaces Cybersecurity and IoT Market definition Finding of dominance Open API regulation Level-playing field across layers Unfair P2B practices

e2e architecture

Collaborative economy

Contractualisation of labour relations Imbalances of bargaining power Difficulties in tax collection Legal liability issues

p2p platforms

Universal service provision Privatization of enforcement and Type I errors

Antitrust, zero-priced goods and cross-layer competition End of cost-based regulation in multi- sided markets Servitisation and Contractualisation Privatisation of governance (e.g. gig economy, smart contracts) Distancing from liability Technology plasticity and delegation

• f enforcement

Centripetal forces, territoriality and jurisdiction: towards new forms of rebalancing revenues Tension between decentralization, immutability and data protection Hybrid data governance arrangements AI, ethics and privacy

Highlights

ESTIMATES OF WORKERS ENGAGED IN ONLINE LABOUR PLATFORMS

• 1% - 5% of the adult population in the European Union (EU) has participated at some time in paid

work in the platform economy (European Parliament 2017)

• Europe – 9-22% (last year); 6-15% (last month); 5-12% (last week) Huws et al. (2017)
• 10% of the adult population in 14 EU Member States have used online platforms for providing

labour service (Pesole et al. 2018) Global Estimate

• 46 million registered workers on 142 freelance platforms (Oxford Internet Institute, 2019)

DIGITAL LABOUR PLATFORMS: LOCAL APP- BASED – RIDE HAILING

• Uber (Registered in USA)
• Operating in 68 countries
• North America – 2 countries
• Europe – 18 countries
• Latin America – 15 countries
• Asia-Pacific – 10 countries
• Africa – 7 countries
• CEE – 7 countries
• Netherlands, USA, UK, India – Dispute settlements
• Go-Jek (Registered in Indonesia)
• Operating in 5 Asian countries
• Easy Taxi (Registered in Brazil)
• Operates in 9 Latin American countries
• Bolt (Registered in Estonia)
• Operates in 37 countries
• CEE – 12 countries
• Europe – 9 countries
• Africa – 6 countries
• Middle East – 3 countries
• Asia-Pacific – 1 country
• Latin America – 1 country

Global Players Regional Players

VENTURE CAPITAL INVESTED

Source: Kenney and Zysman (2019) Unicorns, Cheshire cats, and the new dilemmas of entrepreneurial finance.

VENTURE CAPITAL IN DIGITAL LABOUR PLATFORMS

Sector Total Funding Amount (USD) Countries / Funding Amount Ride hailing and sharing (237) $77 billion US (40% Platforms - $31 Billion) China (5% Platforms - $30 Billion) Same Day Delivery (116) $21 billion US (45% Platforms - $3 Billion) China (8 Platforms - $13 Billion) Freelance (170) $1 billion US (40% Platforms - $552 Million) China (2 Platforms - $431 Million)

Source: Crunchbase Platform

STYLIZED DEPICTION OF A MULTI-SIDED PLATFORM, FLOWS OF RESOURCES, AND ITS ECOSYSTEMS

Source: Kenney, Bearson and Zysman (2019)

DOMINANT PLATFORM FIRMS

Platform Firm Platform Firm Model Revenue, 2018/9 ($ Million) Employment 2018 Apple Innovation 265,595 132,000 Amazon Transaction 232,887 647,500 Google Information/Innovation 136,819 98,771 Microsoft Information/Innovation 125,843 144,100 Facebook Transaction 55,838 35,587 Paypal Transaction 15,451 21,800

Source: Kenney, Bearson and Zysman (2019)

A r A recap ap…

The new technology stack: more than AI

5G

Source: U.S. White House (2016) Source: Patel. et al. (2017)

Source: Zelros AI

Source: Accenture/Frontier Economics (2017)

• The amount of data created in

the past two years is greater than the total amount created in the history of humankind. AI could soon outperform the capacity of the human brain and help reverse challenges to global growth such as aging populations and declining

• utput per worker.
• The widespread use of AI-

enabled technologies could double the economic growth rates of many advanced countries by 2035*. AI is predicted to drive global GDP gains of US$15.7 trillion by 2030

#.

Source: Accenture/Frontier Economics (2017)

The diffusion challenge: six growing divides

• Between innovation-leading and lagging countries
• Between cities and rural areas
• Between leading firms and “zombie” firms
• Between rich and poor
• Between education “haves” and “have-nots”
• Between “gifted” and “non-gifted”?

Source: Winick (2018)

The impact is unknown and endogenous. The real story is about polarisation, and the deterioration of working conditions

A big divide

• Whether v. when
• Dystopian v. Utopian
• Anecdotal v. Statistically Representative
• Static v. Dynamic
• Human-centric v. Neutral
• Cyborg (human-enhancement) v. external robots

Evidence: 50 shades of polarisation

• Acemoglu and Restrepo (2019): Wrong Kind of AI?
• Quality-reducing AI adoption is possible
• Public policy needs to encourage automation without quality reduction
• Acemoglu, Lelarge and Restrepo (2020):
• Manufacturing firms that adopt robots experience an increase in employment.

However, manufacturing firms experience decreases in employment when competitors adopt robots. In aggregate, negative effects dominate positive ones.

• Ashford (2020):
• What if AI and ICT do not create enough well-paying and permanent jobs to

relieve chronic under- and unemployment? Need for a Plan B.

The impact on jobs is endogenous

• Policy choices are essential
• GDP focus v. SWB (or SDG) focus: tacking inequalities and employment at once
• Protecting consumers v. empowering users
• Tax v. Universal Basic Income?
• Key areas
• Education (incl. entrepreneurial skills)
• Infrastructure
• Immigration
• Welfare reform
• Basic research and mission-oriented research
• Public management innovation

The issue of “value”

• Financialisation and digitalisation have led to

the emergence of new ways of doing business, in which value creation and value extraction are increasingly separated (Mazzucato 2018)

• These trends have brought important benefits,

but also dilute corporate responsibility for the sustainability of the economy, society and the environment

• EU industrial policy today is timidly seeking to

re-allocate entitlements as close as possible to where value is generated

1H 2018:

Source: WEO 2019

https://www.businessinsider.com/manufacturing-output-versus-employment-chart-2016-12

AI for what?

Possib

• ssible p

pol

• licie

cies ( s (Ash shford et al

• al. 202

2020)



Advance “Technology Bargaining” between employers and unions/workers; increase unionization



Increase the minimum wage, but reign in unsustainable consumption and encourage sustainable production



Provide a guaranteed basic income/institute a negative income tax, but reign in unsustainable consumption



Engage in Keynesian spending in the short term/Provide guaranteed government jobs in the long term



Pay those doing unpaid work, such as child-rearing/care of the elderly



Shorten the work week without a reduction in wages, but reign in unsustainable consumption



Change the effective taxing of labor and pollution/energy by taxing pollution/energy and reducing the tax on labor



Change the taxation of income and wealth



Increase the participation of workers in employee stock ownership plans (ESOPS) in their places of work



Adopt the recommendations of Marjorie Kelly in The Ownership Society in creating “B-corporations” to invest in public services



Institutionalize the central tenets of binary economics allowing people to gain income from collectivizing their financial capital currently restricted to elitist banks and investment cartels



Increase the contribution of employment to productivity improvements by designing work back into the production process and the delivery of services

Concluding rermarks

• AI will have a pervasive impact on almost all jobs
• The impact is endogenous - one should resist the temptation to

make predictions: but absent proactive public policy, jobs may not be there…

• Skills strategy
• Tax incentives
• Ethics and regulation (humans “in the loop”, “on the loop”, “in charge”)
• Liability rules
• AI for good v. AI competitiveness/dominion: can the geopolitical

Europe manage to sustain its higher standards?

Artificial intelligence and Jobs

Andrea Renda 23 January 2020