Error in judgement- humanity’s over-reliance on AI

Error in judgement- humanity’s over-reliance on AI

Humanity has evolved a lot since our old days living in the wilderness in caves. We have created great things- the Great Pyramids of Giza, the Colosseum, as well as modern wonders such as rockets, space satellites and the Internet.    

Ever since the earliest civilizations humans has striven to optimize our daily work and chores, by creating new tools via advancements in how we manufacture, design and implement new tools to create whole new production systems that rationalize labour. This has been a constant process throughout humanity’s history.  These improvements were not always but mostly related to engineering feats, where a new tool would be invented to make a particular task more efficient. However, we have now reached a point, where certain industries and sectors that longed depended on human in terms of skill, contact and know-how are under severe threat. Why? Because in our ever increasing pursuit of improvement we humans have started going down the path of replacing human decision-making and making it reliant on algorithms in a lot of areas that touch upon our daily lives.

You might think that this preposterous? Surely, no-one would let a piece of code make life and death decisions. Well you are wrong.  An algorithm is used to optimize the connection of your phone based on its’ location, distance from the tower, whether there is space to accommodate that phone and other factors.  Remember that automated customer service system that answers your questions (annoying) and still cannot reach a human? Yes, that is also part of that process.

Algorithms are entering our daily lives even more and could potentially help us run everything from building cars, to running airports, public transport to being used in decision making for cancer treatment.

However, doesn’t this over-reliance on technology come at a cost-a very big cost? The answer is yes. When the code fails it fails badly with dire consequences not just for people but also for the economy.

Below are some examples, where companies relied on AI and their failure cost them dearly.

AI failures

• Medicine and the process of providing a viable diagnosis and treatments is a prime example, where advancements into AI could be used. However, the technology still has a long way to go as it shown when a major tech company’s AI solution failed the mark and was giving incorrect and sometimes dangerous treatment to patients, which posed an extreme risk for everybody.
• Another interesting example involved Twitter and a famous software company, which had designed an AI chat-bot. The idea was the use Twitter as a platform, where the chat-bot would learn how humans interact and communicate and become more proficient in human interaction. However, things quickly turned ugly when the AI chat-bot started writing abusive and inappropriate comments, which forced the team responsible for the system to shut it down.
• A prominent technology company decided to give an AI-driven recruitment a go with some horrible results. The idea behind the system was to make the selection process more straightforward by selecting the top candidates for any role. However, the system failed miserably as it was being biased towards female candidates and it actually penalized them for being female.
• A tech company was trying to create facial recognition software to capture and identify criminals more easily. However, it failed after they used the system on professional athletes and 1 in 6 got falsely classified as a criminal.
• Facial recognition is notoriously difficult to achieve and another example highlights this more than others. A team of researchers was able to fool popular payments systems, which relied on facial recognition, by using 3d print-outs and in some cases just 2d images.
• Finally, perhaps one of the most famous examples of AI is the development of autonomous vehicles. There were very high expectations in the beginning of 2010. However, in 2020 the future still looks very unclear even though a lot of companies have entered the segment. Autonomous vehicles have caused accidents, sometimes with lethal consequences, which shows that even with the latest technology and the best intentions such projects are perhaps too ambitious.

Why do AI projects fail?

It is estimated that 85% of all AI projects fail.There are several reasons why such projects fail and I will mention the most prominent ones.

• Sometimes a problem is far too broad and/or broad to be solved simply by AI. Yes, there have been many advancements in new technology and yet there remain a lot of difficulties, when one tries to complete automate a process
• Unrealistic expectations of the projects- people expect to have a fully functional AI that could pretty much do anything it is asked to do with minimal human intervention. That couldn’t be further from the truth. Yes, there are many advanced AI platforms and tools that are used in a myriad of industries, however they still rely on programmers and researchers to optimize, tinker and improve their capabilities. If a code moves away from its’ designated function it is up to humans to change that direction.
• Insufficient support and understanding from senior people- as with any new technology it requires a lot of vision, drive and support not only from the department that is implementing the said technology but also from senior stakeholders that could easily resolve any road-blocks along the way. Nobody wants to be stuck months on end for a simple sign-off from one department as this pushes the project behind schedule and makes things more costly in the end.
• Infrastructure- developing and using advanced AI system relies on the development and implementation of the Data Infrastructure. If there are severe errors in the data’s quality that could compromise the decision-making process that is being in place and wreak havoc on the final results. Nobody wants to have a SkyNet responsible for all military decisions that would turn on humans (a reference to the Terminator).  
• The list is long but it signals that in order to create viable and efficient AI projects clear guidance have to be put in place and having a robust project management structure is the key for success.

In conclusion

In this post I have shown that even though technology has advanced tremendously over the last hundreds of years we still need to be mindful of some its’ implementation and uses. AI is a tool that could be used for both good and bad and it is up to humanity as a whole to decide how it would be used- to save lives and improve them following strict guidelines or allowing it to bring to fruition ruin and misery that could have been avoided. Nobody wants to limit the spread of technology; however there should always be clear checks and balances not only from a regulatory perspective but also from an ethical one.

Sources:

https://www.dailymail.co.uk/sciencetech/article-6001141/IBMs-Watson-suggested-inaccurate-unsafe-treatment-recommendations-cancer-patients.html

https://www.independent.co.uk/life-style/gadgets-and-tech/news/facebook-artificial-intelligence-ai-chatbot-new-language-research-openai-google-a7869706.html

https://www.cnet.com/news/what-happens-when-ai-bots-invent-their-own-language/

https://www.theverge.com/2016/3/24/11297050/tay-microsoft-chatbot-racist

https://www.theguardian.com/technology/2016/mar/24/tay-microsofts-ai-chatbot-gets-a-crash-course-in-racism-from-twitter

https://www.washingtonpost.com/news/on-small-business/wp/2018/01/26/fabio-the-robot-gets-fired-after-only-a-week/

https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G

https://www.theguardian.com/technology/2018/oct/10/amazon-hiring-ai-gender-bias-recruiting-engine

https://www.independent.co.uk/life-style/gadgets-and-tech/amazon-ai-sexist-recruitment-tool-algorithm-a8579161.html

https://www.businessinsider.com/world-cup-predictions-goldman-sachs-and-banks-failed-why-2018-7?r=US&IR=T

https://economictimes.indiatimes.com/news/sports/in-10-years-dhoni-will-be-permanent-boss-of-csk-team-ceo-kasi-viswanathan/articleshow/76836528.cms

https://news.sky.com/story/amazon-echo-orders-dollhouses-after-hearing-tv-presenter-talking-10722985

https://www.theguardian.com/money/2017/nov/19/alexa-sent-mistake-amazon-order

https://nypost.com/2017/07/10/alexa-calls-cops-on-man-allegedly-beating-his-girlfriend/

https://www.telegraph.co.uk/news/2017/11/08/alexa-nein-police-break-german-mans-house-music-device-held/

https://www.techrepublic.com/article/ai-misses-repeat-in-2017-kentucky-derby-but-heres-what-we-learned/

https://www.bbc.com/future/article/20170410-how-to-fool-artificial-intelligence

https://www.radiologybusiness.com/topics/artificial-intelligence/ai-fails-beat-radiologists-mammography

https://www.techrepublic.com/article/why-85-of-ai-projects-fail/

https://www.nytimes.com/2009/08/27/technology/27compute.html

https://www.economist.com/technology-quarterly/2020/06/11/an-understanding-of-ais-limitations-is-starting-to-sink-in

https://observer.com/2020/06/how-coronavirus-and-protests-broke-artificial-intelligence-and-why-its-a-good-thing/

https://www.globalbankingandfinance.com/the-impact-and-potential-of-ai-in-supply-chain-management/

https://www.newscientist.com/article/2221840-deepminds-starcraft-playing-ai-beats-99-8-per-cent-of-human-gamers/#:~:text=DeepMind’s%20StarCraft%2Dplaying%20AI%20beats%2099.8%20per%20cent%20of%20human%20gamers,-Technology%2030%20October&text=An%20artificial%20intelligence%20can%20now,per%20cent%20of%20human%20players.

https://www.nature.com/articles/d41586-019-03298-6

https://syncedreview.com/2019/12/19/2019-in-review-10-ai-failures/

https://www.forbes.com/sites/ronshevlin/2020/07/06/bankings-delusions-of-digital-transformation/#3eba22834e12

https://hbr.org/2020/04/the-case-for-ai-insurance

https://blogs.scientificamerican.com/observations/why-machine-learning-is-critical-for-disaster-response/

https://eandt.theiet.org/content/articles/2018/10/japanese-researchers-develop-machine-learning-technique-for-natural-disaster-detection/

https://www.forbes.com/sites/cognitiveworld/2019/03/15/how-ai-can-and-will-predict-disasters/#1c5e8db95be2

http://news.mit.edu/2020/aleksander-madry-machine-learning-0308

https://www.theverge.com/interface/2020/3/18/21183549/coronavirus-content-moderators-facebook-google-twitter

http://news.mit.edu/2020/bringing-deep-learning-to-life-0224

https://japantoday.com/category/tech/nec-and-jamstec-develop-ai-based-system-for-measuring-microplastics-in-the-ocean

https://www.channelnewsasia.com/news/video-on-demand/coded-world/choice-11870150

https://www.businessinsider.com/self-driving-cars-fully-autonomous-vehicles-future-prediction-timeline-2019-8?r=US&IR=T

https://time.com/5502192/smart-traffic-lights-ai/

https://www.technologyreview.com/2016/10/07/244656/algorithms-probably-caused-a-flash-crash-of-the-british-pound/

https://www.bloomberg.com/opinion/articles/2019-01-03/the-computers-are-sorry-about-the-flash-crashes