Getting Real With Artificial Intelligence
Louisa Tran, art by Emily Nunell | @emdrawsthings
We’re a fragile bunch, aren’t we? Always in constant danger of the threat of an Opal top-up and rising inflation — the price of beef tacos at my favourite Mexican restaurant increased 20% this year, vastly outstripping the 3.33% increase to my pay. Perhaps the greatest fear we face is the crippling concern that after years of study, we won’t have a job in our chosen field. Unfortunately, given how saturated the job market is — only set to worsen as students graduate in bigger numbers — I may have to hold onto my retail job for a little while longer.
In 2015, the Australian Graduate Survey (AGS) observed that 76% of UTS grads were able to score a full-time or part-time job in their chosen field within three months of graduating. Pretty sweet stats if you ask me. Well, since then a lot of things have happened. Kendall Jenner is now a thing, Trump became president, and just casually, robots have slowly started to replace us at work. A report written by the CSIRO and Australian Computer Society estimates that nearly half of Aussie jobs are at risk of computerisation and automation. Logic tells us that grads are especially affected, as we are the least experienced and thus have the least to offer — at least in the short-run. For those on the cusp of graduation, here’s an overview of the major developments occurring in business, law, medicine, and journalism.
Entry-level accounting has been particularly prone to automation as of late because the technology required to replace workers at this level is readily available, not too sophisticated or expensive to implement, and the cost savings are palpable. For example, the ‘Big Four’ firms have started deploying robotic process automation (RPA) robots or ‘bots’ to mimic workers who copy and paste information from one system to another to generate a report. To give you a better picture: a bot can scan an invoice, upload the data onto an Excel spreadsheet, log into another system such as MYOB, generate a revenue report, and finally, send it off as an email attachment from the firm’s account. Deloitte estimates that a bot can complete 15 minutes of work in one. Talk about a hot minute.
KPMG also uses robots in its tax division. At lower-levels, income taxes are concerned with classifying certain revenue and expenses as taxable income and allowable deductions. Through the use of ‘Machine Learning’ (ML) where systems are taught to learn on their own, robots have started differentiating deductible repairs from non-deductible improvements. Accuracy rates have been recorded as high as 99%. To explain this technology, say a computer was exposed to thousands of images of horses and cows. By incorporating a feedback loop that indicates whether it’s correct or not, it will be able to focus on the distinct features of the two animals to come up with a system that allows it to categorise the images correctly.
Audit is another area that has taken a significant liking to automation. Since time immemorial, auditors have faced the difficulty of wanting to provide 100% confidence in financial reports, but not having the time to test every single transaction that a company makes. To this end, auditors have only been able to sample transactions, and therefore each audit opinion that is issued is always adjusted for sampling errors. With the help of technology, auditors will be able to provide the confidence that investors and other users of financial statements are demanding, and in a shorter amount of time. The work of human auditors may soon be a thing of the past, while the number of grads getting their foot in the door is getting slimmer than the fat content in my mum’s soy milk: light.
Law is another profession that has historically been very kind to graduates. This is because the wealth of information that must be summarised and understood from cases, textbooks, and articles have always been enormous and require a large amount of people. Grads needed to be equipped with superior research skills and adept at fashioning the most nuanced of search terms to weed through a heck of a lot of irrelevant information to strike gold. Thanks to an artificially intelligent lawyer named Ross, researching in this manner may be another thing of the past. Ross works by allowing lawyers to ask questions in ‘natural language’, in much the same way we would ask a colleague a question. Ross then sifts through over a billion cases, articles, and other documents, and returns the exact passage a lawyer needs. Not only is this faster, it is also cheaper for the firm: previously the hours spent training grads to use legal databases such as LexisNexis were not billable to clients and, as such, the firm absorbed the costs.
LexisNexis has also created its own AI machine called Lex Machina — the Latin phrase for ‘Law Machine’ — is even more powerful than Ross as it uses natural language processing to, this time, determine which judges favour plaintiffs, the legal strategies of opposing lawyers based on the cases they are likely to use, and which arguments are likely to convince specific judges. US company Premonition takes this one step further by predicting the winner of a case before it goes to court based on a statistical analysis of verdicts in similar cases.
The work of doctors in diagnosing and suggesting treatment plans for patients, much like lawyers, is largely dependent upon how much information they can read, understand, and — most importantly — remember. Recent developments in data analytic capabilities have only made this more difficult as now, more than ever, there is an unprecedented amount of important information to keep track of. From the same technology that powers Ross, IBM Watson has handed doctors a lifeline. Watson works by allowing doctors to ask questions in natural language. Through this process, Watson learns the symptoms a patient presents, and compares this against data from clinical trials, 23 million journal articles, and other relevant information to form diagnosis. Amazingly, Watson boasts 90% accuracy in lung-cancer diagnoses, significantly higher than the 50% accuracy that human doctors average. Further, Watson can complete 160 hours of research in less than 10 minutes. If this isn’t amazing enough, just wait until you hear how Watson is helping doctors tailor treatment plans for cancer patients. Currently, it is very expensive to sequence a patient’s entire genome for mutations. Generally, patients will opt instead for a ‘panel test’ wherein only a section of the genome is tested, that which contains the subset of genes scientifically most likely to lead to mutations, and thus cancer. Much like the auditing example from before, as efficient as sampling can be, it can also lead to sampling errors. In the case of a 76-year-old man with brain cancer, Watson sequenced his entire genome and uncovered mutations that weren’t identified in the panel test. This meant that particular drugs and trials that could potentially save the patient were also identified. In June of this year, it was announced that IBM Watson will be brought to the land down under for the first time.
Journalism is markedly different from the aforementioned professions as there is a greater deal of creativity, at least in the general sense, from the outset. Based on this understanding, journalists are often lulled into a false sense of security when it comes to AI. AI proves however, that nothing is beyond its reach. In 2007, a North Carolina-based company created an artificial writer, Wordsmith. Wordsmith picks elements from a data set and structures a human-sounding article, capable of emotive language, and varied diction and syntax. In 2015, this service was made public, and since then more than 1.5 billion pieces have been written. The only current drawback is that the content must be data driven. Where qualitative, descriptive statements are needed, but cannot be matched with data, Wordsmith finds itself out of its depth.
Journalism has also changed in recent times due to the amount of information now available. Where before, with limited resources, journalists were tasked with reporting only on bigger stories that would captivate broader audiences, raw local data from governments, local authorities, and public services is being used — in tandem with services like Wordsmith — to produce the local stories that have up until this point, gone unreported.
Hope lost for humanity?
So, the real question is, are we still relevant in this shiny new age where robots can complete the same tasks more efficiently and economically? The answer is a resounding yes. Let me tell you why. Harking back to the above examples, it’s pretty obvious that technology cannot complete the tasks from A to Z by itself. We are required both at conception — to create the parameters within which systems will work, and at the end — to provide insights and negotiate with clients and patients. This is where our real value lies.
For example, in business, a robot is fantastic at crunching the numbers, but number crunching does not in itself avail the company of needing to pass its own judgement. Further, it does not do anything in the way of helping a company balance the competing needs of stakeholders and maintain its corporate social responsibility. As a robot is neither ethical nor capable of bias, it will never completely eradicate the need for human agents in this profession.
In law, professionals are still required to advocate on behalf of their clients. A robot can provide the facts, but unlike a lawyer it cannot appeal to the emotions of a judge, and this is important because — unlike robots — we are never swayed completely by numbers. We consider emotions such as forgiveness and compassion — in law, the outcome is not about the bottom line, but about somebody’s wellbeing and the interests of the public. It is much better that lawyers are assisted by technology so that together, they can achieve cheaper, but still human outcomes for their clients.
In medicine, it is important for a doctor to work on their ability to empathise. It is all well and good for a machine to create a tailored treatment plan, but it takes a human being to sit down with a patient and motivate them to continue taking their medicine, or to encourage them to keep fighting. The body has an amazing way of healing itself, especially where human resilience is involved. No matter how advanced IBM Watson becomes, hospitals will never be devoid of human doctors.
In journalism, AI allows reporters to spend less time understanding data sets and more time pursuing leads provided by AI analysis. This also allows for higher-order reporting. And sometimes, it’s just nicer to have somebody explain the news to you rather than reading it online. Think about the weather for example; we have all the data on the screen, but we still appreciate Tim Bailey getting us all fired up about the weekend temperatures. Heck, the Bailey Barometer is a national treasure and losing it would be a complete travesty.
A final mention must be made. Technology is flash, and it’s exciting but what it isn’t is flesh, and flesh begets feeling. Think about the last time you felt something, big or small. Maybe you wanted to prove somebody wrong. Maybe you fell in love. Maybe you wanted to take a chance on something or somebody. These are all motivations behind actions that have shaped the course of history and in a world with human agents, sometimes being irrational and motivated by our core emotions is not only important, but leads to a better outcome. To this end, a robot will never be able to replace what we do naturally, and what we do so easily; that is to be vulnerable, to be leaders, and form meaningful connections with one another. If anything, technology has led us towards our core competencies, and will force us into becoming better versions of ourselves at a faster pace than we could have ever imagined. This can only be a good thing. Hand-in-hand with technology, we’ll be better equipped for our futures, and achieve so much more together than we ever could have alone. So chin up, graddies. We’re going to be alright.