1

B057509

Digital Business

Neil Pollock

6th April 2018

2260 words

“Will artificial intelligence develop unchallenged or will some

B057509
 2

“Will artificial intelligence develop unchallenged or will some forms of expertise resist
automation?”

Artificial Intelligence (AI) has the potential to profoundly change the way tasks are

performed and to be as transformative to the economy as the rise of computers [CITATION

The17 \l 1033 ]. Indeed, a host of industries are already experiencing the effects of the

technology in their business, including manufacturing, health care, transportation, education

and many more to come (Brynjolfsson & Mcafee, 2017). The technology is improving

outcomes in human health, safer driving, better learning and general enhancement of people’s

lives (Stone et al, 2016). Yet, as with any new technology which has had an impact on the

industry including the internal engine, telephones or personal computers, the fear of how the

technology will impact jobs have started to emerge – will it overhaul entire industries and

will this erase or create more demand for human labour? [CITATION The17 \l 1033 ].

Who is right in this debate – the pessimists claiming that AI will indeed take the majority or

all the jobs or the optimists who insist it will drive the creation of new jobs, more than the

ones that will cease to exist [ CITATION The16 \l 1033 ] ? This paper will argue that the truth lies

somewhere in between – automation will spread and labour markets have a will continue to

be disrupted by it requiring employees to learn new skills faster than ever before and work in

collaboration with AI, old occupations will perish, but new ones will open up. AI provides a

world of opportunity that will change the way humans operate in all sectors, and ultimately it

is the ability and responsibility of humans to program and create AI with which they will be

able to work hand-in-hand in the foreseeable future.

There is no denying that the move towards AI has already begun and it will continue to

substitute and outperform currently existing operations across many sectors. Currently

B057509
 3

machine learning is the branch of AI which has been most massively used by companies with

applications in augmented reality, online ad targeting or self-driving cars (The Economist,

2017.). It has already aided and complemented the digital transformation in more ways than

one could have imagined. Another type of AI which is substituting white collar workers is

deep learning – for example, computers have effectively provided expert medical advice in

the field of radiology (The Economist, 2016.). Although AI could be used for even the

smallest of tasks across many occupations, which are most at risk of being automated?

According to a study by Osborne and Frey (Frey & Osborne, 2013) nearly 47% of US

workers (out of a total of 702 occupation being tested) have a high potential of being

automated or only partly so according to different studies (Arntz, et al., 2017). Those include

occupations such as transport and logistics workers (i.e. UBER autonomous vehicles), office

support or sales and services (i.e. cashiers, telemarketers). While, the percentage has been

estimated at a slightly lower 35% within the United Kingdom, where the “creative economy”

is providing a barrier to automation (Frey & Osborne, 2013). This is not the case in Japan,

where technology and automation is arguably more advanced than elsewhere with up to 49%

of jobs threatened (Frey & Osborne, 2013). Although these figures are based on an extended

survey and one could argue there are limitations, going forward they are likely to become

more real.

While AI has had and will most certainly continue to have enormous effects, its spread is

limited by a variety of factors and is, therefore, not as fast or as wide as many have expected.

AI has had enormous advances in fields such as perception, cognition and problem solving;

often surpassing human’s performance. A stage has been reached where Polanyi’s Paradox,

our inability to fully explain what we know, has been overcome by machine learning which

does not require the previously ubiquitous and requiring full self-understanding the “coding”

B057509
 4

of our knowledge into the machine (Brynjolfsson & Mcafeee, 2017). The examples which

can be given include software that has become normal and customary to so many, includes

Siri (Apple), Alexa (Amazon), Cortana (Microsoft) and Google Assistant (Google). These all

use speech recognition and the use of machine learning to provide a ‘bot’ system that will

hold conversations and provide the user with information upon request. Whilst “deep” neutral

nets in image recognition exist with examples being Google’s DeepDream software or

reverse image search, Amazon Rekognition or the use of such services within corporate

offices as opposed to ID badges (Byrnjolfsson & Mcafee, 2017). As regards to the latter, AI

is being used for optimisation (inventory optimisation and product recommendations as

utilised by Amazon), malware detection (Deep Instinct) or online ad placement improvement

for fast-moving consumer goods companies (Brynjofsson & Mcafee, 2017). AI’s

understanding is often limited to a particular task and subsequently represents examples of

“narrow AI” (Shead, 2017). Consequently, the lag behind human’s (or even baby’s or

animal’s) wide general intelligence and machines are not even close (Shead, 2017). Humans

use what is called “unsupervised learning” while machines do not. That said – machines do

have access to large sets of data which they can recognise and label (consider AI image

search algorithms which recognised skin cancer better than twenty-one dermatologists)

(Bloomberg, 2017). However, it is this source of intelligence that is a principle factor to why

the spread of AI may well be limited.

While there is a specific set of jobs or parts of occupations which can be taken up by AI,

notably the repetitive ones, there are also such that will resist automation. This is due to the

employment of human interactions and requirement of human-necessitated aptitudes. In fact,

automation has been largely overestimated by neglecting heterogeneity of tasks in certain

jobs (Arntz, et al., 2017). The most obvious predictor to whether a certain job is likely to be

B057509
 5

impacted and replaced by AI is whether it has repetitive elements within it or is considered

entirely routine, including both routine manual labour and routine cognitive tasks (Lacity &

Willcocks, 2016). Two large groups which conduct non-routine work include high-value

white collar employees, i.e. senior management, management consultants, business analysts,

architects or designers; and blue collar, unskilled workers employed in occupations such as

cleaning (Stewart, et al., 2015). AI does not have the social abilities humans possess – while

it can for example assess someone’s mental state, it cannot work or do the psychologist’s or

social workers’ job of altering or changing their mental state due to its inability to express

empathy, compassion, solidarity, motivation and inspiration (Brynjolfsson & Mcafee, 2017).

On that account, a set of jobs which require social interaction are particularly unlikely to ever

be replaced by AI including non-routine doctor consultation, therapy or personal training.

And indeed, as reported by Deloitte (Stewart, et al., 2015) an unprecedented shift to “caring”

jobs have already been observed in the UK – as compared to 1871 when they represented a

mere 1.1% of total employment; in 2011 that number has risen to 12.2%, with more specific

numbers such as a 909% and 580% increase in nursing and teaching assistants respectively.

Creative and inventive occupations are also unlikely to be as impacted by AI – innovators,

entrepreneurs, creators, scientists and artists. As Picasso observed when contemplating the

future of computers “But they are useless. They can only give you answers” – those

occupations are taken up by people finding the right questions to ask and previously

unimagined opportunities and territories to explore (Brynjolfsson & Mcafee, 2017) – a task

which cannot be automated. Ethical aspects of jobs cannot be taken by AI as well – those

cover racial, ethnic and gender prejudices or privacy concerns (Stone et al., 2016); an

example is self-driving cars which need to make the ethical decision between lives of

pedestrians and the lives of people within the vehicle – such consideration will serve to

expand the ethics human-operated field within AI (Kosbie, et al., 2017). There have been

B057509
 6

cases with autonomous vehicles, most recently Uber (Financial Times, 2018), where people

have died as a result of AI not proving up to the standards required, therefore in this sphere

the development of unchallenged AI will have to be somewhat resisted for the time being.

Furthermore, there must be ethical principles considered within the development of AI; such

as no robots being designed solely to kill or harm humans, and a robot should always be

recognisable from a human (Lacity & Willcocks 2016). Although AI offers so many

opportunities, it is unlikely ever to surpass humans within various occupations.

While AI is rapidly expanding it will also serve to create new jobs as has happened with

breakthrough technologies in the past. The proliferation of the computer, necessitated the

creation of computer science occupations such as software developers, computers analysts,

and network system or database administrators (Kosbie, et al., 2017). Hence, while it can be

argued that further automation via AI will redefine the job market and the economy, increase

efficiencies and reduce costs for companies, it will also increase the demand for new human-

led AI-related jobs (The Economist, 2016). It is much easier to imagine jobs being lost

because of AI as compared to those that will emerge as result of AI (Stone et al., 2016). Yet,

in the past, new technologies have consistently led to more jobs being created as compared to

those that cease to exist. This aligns with a 2018 prediction of Gartner, a technology research

company, that by “By 2020, AI will become a positive net job motivator, creating 2.3M jobs

while only eliminating 1.8M jobs” (Gartner, 2018). Tasks which can be automated but

complement an already automated function which via AI can be performed more quickly,

cheaply and efficiently; will naturally increase the demand for connected non-automated

tasks (The Economist, 2016). For example, fast and accurate analysis of legal documents

through AI has increased demand for legal and paralegal work with legal clerks in the US,

rising 1.1% on an annual basis between 2000 and 2013; better automated recommendations

B057509
 7

via e-commerce have increased online shopping and employment in the wider retail sector; in

the medical profession, automation of routine tasks has enabled doctors to refocus on the

human-part of the job and therefore, increased both their expertise, capacity and demand for

the service they are providing (Mitchell, 2017). In the case of self-driving cars, the more time

consumers have will enable them to consume more goods and services, thus increasing the

demand in other parts of the economy (The Economist, 2016). Yet, for a smoother transition

to a new AI-enabled economy many workers will need to be trained to refocus on non-routine

tasks (The Economist, 2016). Until there is a general acceptance that AI has arrived and is not

going anywhere; people are yet to recognise that it will become a positive net job motivator.

While not many fields of work will resist automation, humans will work hand in hand with

AI and effectively complement each other. Zuboff (1985) suggested that technology should

not only be used to automate quicker and accurate production; but also presents an

opportunity to inform the work process to improve operations and increase innovation long-

term. Complementary work will allow the most efficient and beneficial outcome to be

achieved in all occupations going forward. Through robotic process automation services, AI

has enabled previously labour-intensive jobs to be conducted rapidly and accurately leaving

more room for unstructured aspects of a job to still be carried out by humans (Lacity &

Willcocks, 2016). One such example is Xchanging, a London based start-up, which has

substituted several humans analysing up to 500 insurance claims in a few days to a work

necessitating only 30 minutes of AI analysis and the assistance of a few humans. In the case

of Xchanging, when the jobs required have teams of AI and humans working alongside one

another, the performance has been noted of a higher standard with AI being quick to conduct

repetitive tasks and humans taking up a problem-solving approach and social interactions

(Lacity & Willcocks, 2016). Complementary work includes chatbots and customer-service AI

B057509
 8

services which have to be built, trained and updated by humans to have real life dialogues

and interactions; autonomous vehicles might require human labour to deal with emergencies,

non-routine operation or handling packages and deliveries (The Economist, 2016). The

complementary aspects of AI enhance the appeal for many businesses and occupations to

adopt it in some form without resistance.

In conclusion, AI will substitute and alter occupations currently performed by humans – it

will also inevitably eliminate jobs but create new ones too. It is a sphere that is constantly

advancing and is yet to take full fruition and thus, the full repercussions are still not clearly

known. While the benefits are unlikely to spread evenly, individuals can seek ways to expand

their existing skillsets in occupations requiring new skills, or such that are unlikely to be

substituted by AI – involving interaction and inherently human abilities such as exhibiting

empathy, care, creativity and motivation. This paper argues that although artificial

intelligence is developing and will continue to develop rapidly; there are certain factors,

elements and occupations that will resist automation. Whilst AI has the potential to

profoundly change the way tasks are performed (The Economist, 2017), its development is

likely to be restricted somewhat. It is the responsibility of humans to create AI with which

they will be able to work hand in hand in the foreseeable future.

B057509
 9

Bibliography

Arntz, M., Gregory, T., & Zierahn, U. (2017). Revisiting the risk of automation. Economics Letters, 159,
157-160.

Bakhshi, H., Downing, J. M., Osborne, M. A., & Schneider, P. (2017). The future of skills.
Employmenty in 2030. London: Pearson and NESTA.

Bloomberg. (2017, June 29). Diagnosing Skin Cancer with Google Images. Retrieved February 2018,
from https://www.bloomberg.com/news/articles/2017-06-29/diagnosing-skin-cancer-with-
google-images

Brynjolfsson, E., & Mcafee, A. (2017). The Business of Artificial Intelligence: What it can - and cannot
- do for your organisation. Harvard Business Review.

Financial Times. (2018). Uber halts self-driving tests after death in Arizona. Financial Times.
Retrieved March 2018, from https://www.ft.com/content/b765c2fc-2b9a-11e8-a34a-7e7563b0b0f4

Frey, C. B., & Osborne, M. (2013). The future of employment: How susceptible are jobs to
computerisation. Oxford: Oxford Martin School, University of Oxford.

Gartner, 2018. Gartner top strategic predictions for 2018 and beyond. [online] Available at:
https://www.gartner.com/smarterwithgartner/gartner-top-strategic-predictions-for-2018-and-
beyond/ [Accessed March 2018]

Kosbie, D., Moore, A. W., & Stehlik, M. (2017, June 5). How to prepare the next generation for jobs in
the AI Economy. Harvard Business Review.

Lacity, M. C., & Willcocks, L. P. (2016). A new approach to automating services. MIT Sloan
Management Review , Fall (ISSN 1532-9194).

Lacity, M. C., & Willcocks, L. P. (2016). Service automation; robots and the future of work. Print.
Steve Brooks Publishing.

Mitchell, T. &. (2017, April 20). Track how technology is transforming work. Nature, 544, 290 - 292.

Peter Stone, R. B.-B. (2016). Artificial Intelligence and Life in 2030. One Hundred Year Study on
Artificial Intelligence. Report of the 2015 Study Panel. Stanford, California: Stanford
University.

Shead, S. (2017, October 27). Facebook's AI boss: "In terms of general intelligence, we're not even
close to a rat". Retrieved February 2018, from http://uk.businessinsider.com/facebooks-ai-
boss-in-terms-of-general-intelligence-were-not-even-close-to-a-rat-2017-10

Stewart, I., De, D., & Cole, A. (2015). Technology and people: The great job creating machine .
London: Deloitte LLP.

The Economist. (2016, June 25). Automation and anxiety. Will smarter machines cause mass
unemployment? . Retrieved February 2018, from
https://www.economist.com/news/special-report/21700758-will-smarter-machines-cause-
mass-unemployment-automation-and-anxiety

B057509
 10

The Economist. (2017, August 26). Artificial intelligence will create new kinds of work. Retrieved
February 2018, from https://www.economist.com/news/business/21727093-humans-will-
supply-digital-services-complement-ai-artificial-intelligence-will-create-new

Zuboff, S. (1985). Automate/informate: the two faces of intelligent technology. Organizational

B057509