Can Artificial Intelligence Replace Humans?
Introduction
Multiple reports in the past decade claim how a major part of the human workforce will be replaced by artificial intelligence. The role of a human being behind the machine gradually becomes obsolete as more and more research is poured into this field which indicts a job crisis in the years to come. Major names like Stephen Hawking already warns the world that the “development of robots and intelligent machines” could foster a creation that takes over tasks that are currently being performed by humans (Sulleyman, 2017). However, many ponder if our creativity can truly be automated by a machine both accurately and transparently. It is known that an automated translator can output at a much faster and efficient rate than what humans are capable of but there have been cases that indicate inaccurate results produced by the machine. These cases mainly derive where cultural analytics are integrated into the information process which machines are unable to read the input correctly. Thus, in this paper, I will do a side by side comparison of artificial intelligence and human intelligence and argue why humanity will not be taken over by automated machines because of a lack of creativity and connotative abilities. This argument will be rationalized by analyzing the common concerns pertaining to artificial intelligence which is not only specific to cultural associations in the literary and connotative meaning and the definition of beauty in artwork.
Literature Review
Artificial intelligence has been around for a very long time with critics and scientists questioning and theorizing how we can architect such a creation. Many scientists and articles today bring up how artificial intelligence fails to replicate natural language processing (Pedro, 2016). The machine has major difficultly understanding whenever there is any hint of a foreign language is involved. Although the integration is successful in many parts of the world, when we pit the software against a human translator, it results in many “grammatically awkward” sentence structures (Pedro, 2016). This faulty trait of artificial intelligence speaks volumes as to how far away we are from becoming dependent on such a machine. Likewise, when the introduction of Siri and Cortana came up, we all thought it would be a revolutionary piece of work but in the end, it’s only capable of doing simple tasks which leads many to disappointment (Pedro, 2016). This review leads us to believe that automated machines still have a long way to come in terms of replacing the human factor in the equation.
In other circumstances, we ponder how scalable it’ll towards other departments such as beauty and cultural associations. One shocking example is the first ever international beauty contest solely judged by an algorithm. Much to its surprise, the algorithm’s result only produced 44 winners out of 6000 applicants which among them only had one dark skin candidate. This left the world to debate whether the algorithm is inherently biases (Levin, 2016). Here, the integration into its deep machine learning is mainly composed of feeding it thousands and thousands of pictures that depict beauty. However, when we approach the definition of beauty from a human perspective, it has such an abstract meaning compared to a machine learning it. The literal meaning of beauty is a combination of qualities, such as shape, colour, or form, that pleases the aesthetic senses. This not only illustrates how ambiguous beauty is because it could range differently from individual to individual. Which leaves us to ponder how exactly can we accurately code the definition of beauty so that an automated machine can flawlessly carry out predictions without fail. Furthermore, when we place artificial intelligence amidst a foreign culture, without the western background, how accurately are they able to interpret and translate anything? Many cultures are disruptive in the sense that their cultural meanings have no direct translation to the western culture. This transitional trait results in an automated machine that by default lacks the human creativity will produce awkward grammar structures (Chung Hyun-Chae, 2017).
Interpretation
Language comprehension is a cognitive process that is currently only apparent in humans but has been an ongoing development for robotics. In these past decades, artificial intelligence does the brute work of content analysis better such as spell checking, grammar, and translating. It however, still lacks the ability to think and relate meanings of the text. As an automated machine, these simple measures can be easily coded into a database because of its simplistic and literal translations between the two transactions.
As an example, I’ve inserted a figure caption to demonstrate what I mean by sentence interpretation. As a human being, we read the following sentences as “I see a brown bear, I see a blue bird, I see the red crab”. As an automated machine, the following is read in the exact same order. Yet the two approaches to this figure differ heavily when we begin to regurgitate the results and ask questions that relate to the example. The question of “What subjects were discovered in the writing” would pose difficulty to that of a machine because of its inability to differentiate and find relevance within the word. Whereas if the question were to be answered by that of a human, then the following response would flow with ease. Considering this subtle flaw, the rest of the paper will carefully dissect how the machine approaches situations like these in depth and illustrate the importance of a human behind content analysis.
Understanding and Relation
Understanding like interpretation is another cognitive process that of which automated machines have yet to learn. It is a daunting task but nevertheless not impossible to develop. In the previous example, we explored how a machine can approach a figure of writing by just translating its direct meaning without any sort of relational context. Here, I want to place a specific emphasis and highlight how us a human bring forth a connection when reading any sort of article.
With our initial approach in reading the following figure, we relate the subject to its activity by a verb linkage. As we read the first sentence of “the dog runs fast”, we build a mental picture and keep in mind of the fact that the dog can run fast. This digestive trend of relating its subject with action may seem tedious but the cognitive process is quite intuitive. That of which again a machine fails to do because of its sole ability to translate words through parsing or coding. How a machine interprets the said example is majorly through its ability to parse each word one by one. The database begins to fill up with singular terms in a fashion like “the/dog/runs/fast” but never does the database creatively connects the words together. This limited function demonstrated by a machine hinders its ability to qualitatively perform in a comprehensive atmosphere (Cuilenburg, 1988).
This notable limitation affects a vast part of content analysis as it defeats the overall purpose of this certain methodology. If a machine is unable to intuitively connect and find relevancy within their data to answer questions, then what difference would that make in the real world. The ongoing debate of whether automated machines are effective at producing qualitative results be debatable solely because of this limited functionality. Likewise, it is within the best interest of any avid reader to seek comprehensive data from a reputable facilitator to best maximize their outcome. In the situation that when a computer provides a false understanding, it can greatly misguide certain scenarios and thus hinder certain actions ranging from contract agreements, deals, interpretative analysis, or critics. The goal of content analysis is to seek and sort out of what the article is talking about and differentiate how the facts are laid out. With the lack of understanding demonstrated by an artificial intelligence may prove to be a nuisance because of volatility of true or false results. This suggests that humans are better suited for this methodology approach.
Semantic and Syntactic
Considering the understanding and interpretation of automated thinking, we can depict that artificial intelligence is a work in progress. The ongoing development in the automated industry promotes cognitive psychology because of its sole potential to enable the machine to think like a human. Although the task is difficult, it is not impossible because we can code the connective measures that we mentioned previously so that a machine can understand “the dog runs fast”. This practical yet tedious approach means architecting a database comprised solely of verbs that acts as a connection between two parsed terms. Whenever a key term appears, the machine will then be able to link and comprehend that relation between the two subjects (Cuilenburg, 1988). This however contains a major syntactic flaw. When we alter the sentence structure to “The fastest animal is a dog” compared to “The dog runs fast”, it begins to form confusion among automated thinking. Semantically speaking, the dog still runs fast as mentioned by the statement but if the output were to be generated by a machine then the connection would be false (Cuilenburg, 1988). This exclusive example not only illustrates how language complexity can alter automated results but brings forth the consequences from the lack of cognitive thinking.
After mentioning how syntactically altering the variables in a single sentence can manipulate artificial intelligence outputs, we begin to analyze how alteration in a whole contextual environment can also affect it. An example of this would be reading a novel and the references of characters in conversations could stipulate different endings. A sentence such as “Miss Taylor talks to each student” could translate like to “She likes each student” in another iteration. Here it is obvious that if these two sentences come right after another than the term “She” references Miss Taylor. Likewise, if the novel only had one female character then we can also say the same for above. The reason for this specific example to be mentioned is because authours often don’t reference the subject multiple times. Instead they referred to them in the third person to simplify conversational environment. There are multiple solutions that one might argue how these simplistic gender categorizations can be coded into an artificial intelligence. However, what happens if there are multitudes of character genders and personalities. It is then only could a human cognitively process the context of the situation.
Situational Grasp
Aside from sentence structure and context environment, automated thinking lacks in another area of reading comprehension and that is situational grasp. What these two words define is its ability to analyze any given situation critically with multitudes of characters and can differentiate who is who. Take into consideration the following example:
| “Jimmy is desperate for the teacher’s knowledge, when he walks out, he’ll try to grab his attention.”
This short sentence emphasizes the third person multiple times to stimulate how an automated machine will approach the statement. Here we see that the use of terms such as “he”, “his” are used coherently without pause. By a human level analysis, it is easy to differentiate in this situation that “he” refers to Jimmy and “his” refers to teacher. Whereas if artificial intelligence were to interpret said sentence, it may have mistakenly categorize both male terms as the subject Jimmy. Through the different uses of pronouns, it is hard to dictate which situational context path will automated thinking guide you to (Cuilenburg, 1988). Another approach to this sort of context would be situational knowledge where the reader can interpret the ongoing environment and see what relevancy the details like entail.
To explore this a bit more, I want to emphasize how automated machines do not have the capacity to obtain prior knowledge. In the instance that articles are produced by the same authour, it may contain a similarly writing style or relevant references that automated machines fail to grasp. An example of this could be a novel that has multiple sequels. The connections that characters form in the first of the series will be dully noted by artificial intelligence but there would be no relevant connection to the second or third series. This in term affects the overall output as each new series beginning will create a clean analytic state for the machine. When we do this side by side comparison to that of a human, we begin to realize that a human’s cognitive capacity greatly outperforms that of a machine in terms of qualitative analysis. Reason being is that we’re able to perform something called short term memory to withhold crucial information that allows us to fully comprehend the full contextual environment. Likewise, an automated machine may have better quantitative states, it fails to find relevancy within its own data. These small yet critical flaws are what outweighs a human compared to a machine.
Homographs
Finally, within our final stage of critically analyzing artificial intelligence, we begin to explore the English language of homographs which are words that are spelled the same but could contain different sets of meanings. These different interpretative meanings could change depending on whether it derives from the context or expression. Given in an optimistic manner that our current artificial intelligence can acquire the abilities to complete all the former mentioned before then this will prove to be the next obstacle. Intuitively speaking, it is difficult as it is to grasp what some homographs mean in certain situations. Now as an example:
There could be different interpretations of the word bow, row, lie, fair, or tear. All these tricky homographs could lead an artificial intelligence to a wrong output. At the same time, it may cause confusion among certain connections. Imagine subject A was connected to subject B by a means of a homograph. From a technical standpoint, the automated machine may be intuitively right being this key term linking the two subjects. However, the ending result and the overall interpretation could be completely subpar compared to that of a human’s understanding. The idea of the natural language being understand by a computer is furiously sought out in today’s time. It is an ever-growing process as the science of cognitive psychology becomes more renowned, the more problems arise for the side of artificial intelligence.
Opinion
The expansion of media news serves as a difficult platform for automated machines to interpret considering the multitudes of slangs and verbs that can be used to describe the ‘news’. From a human backed content analysis, it is easy to differential what certain expressions lead to (Nichols, 2017). Likewise, differentiating between fake news and real news can affect outputs greatly due to a lack of intuitive thinking.
In the figure, above, it depicts how two characters are simultaneously browsing the internet when one poses the question of how to differentiate fake news. This situational context may vary from different media articles thus proposing how automated machines may not be able to qualitatively collect data due to the lack of opinion (Nichols, 2017). This reasoning is justified by the sole fact that behind the automation, there is no cognitive process and that is clear from the sections we’ve dissected above (Cassimatis, 2012). Not only does this new approach of news media can cause detrimental effects to automated content analysis but it serves as a mandatory preceding before machines can be qualified to replace humans behind the screen.
Conclusion
After critically analyzing how artificial intelligence takes its approach to information sifting, it can be said that machines lack the high-level cognitive abilities. In the study of content analysis, it can branch out to other areas such as complex-reasoning, plausible inferences, natural language, and complex problems. All of which require the sole necessity of a cognitive brain which there evidently is a lack of in today’s current technology. These factors come together to form a cohesive, qualitative, comprehension that derives from a human brain from which an automated machine does not stand in comparison currently.
It is evidently clear that artificial intelligence lacks to some extent the basic cognitive functions of which a human mind possesses. From semantic differentiation, cognitive psychology, to interpreting the English language, a simple parsing and coding program couldn’t possibly replace a human yet. Situational context and opinion knowledge plays hand in hand in content analysis which gives a human the ability to full interpret and comprehend media to its fullest extent. These interpretations not only require more than simple parsing and database translation but it preaches from a cognitive psychology standpoint and justifies why the ability to think is a must. The overall concern is whether a computer can outperform human labour because it obviously can but if it can output the same qualitative work generated by one. The dangers of precoding a program with prior knowledge may persuade the machine to think in a manner and that is not what the purpose of artificial promotes. Instead, the more humans learn to understand the human brain and how we think, then we might able be able to intuitively develop a like-minded program to perform similar actions. With the current state of technology, it is not doubt that artificial intelligence might one day reach the pinnacle where it is effective enough to replace humans behind the screen but as of right now, the ever-lasting of ambiguity especially with opinionated media can prove to be a detrimental consequence for the machines. To conclude, artificial intelligence is not where it needs to be therefore it is not potentially threatening to the human position.
Bibliography
Artificial Intelligence and Fake News. (2017, November 15th). Retrieved from http://s3.amazonaws.com/libapps/accounts/88450/images/Signe_Cartoon_12_02_16.jpg
Bredenkamp, A. (2016, May 2). Artificial Intelligence Meets Content: What the Rise of the Content Robots Means for You. Retrieved from Acrolinx: https://www.acrolinx.com/blog/artificial-intelligence-meets-content-what-the-rise-of-the-content-robots-means-for-you/
Cassimatis, N. L. (2012). Artificial Intelligence and Cognitive Modelling. Theoretical Foundations of Artificial General Intelligence, 22.
Chung Hyun-Chae. (2017, February 22). Koreantimes. Retrieved from Human Touch Key Factor In Translation, Interpretation: http://www.koreatimes.co.kr/www/nation/2017/02/181_224481.html
Cuilenburg, J. J. (1988, January). Artificial Intelligence and Content Analysis. Quality and Quantity, 66.
Daroczy, G. (2010, January). Artificial Intelligence and Cognitive Psychology. Applied Informatics .
Homograph Image. (2017, November 15th). Retrieved from https://i.pinimg.com/originals/19/f0/2d/19f02dc45f8750cb68a7d090311d51d3.jpg
Levin, S. (2016, September 8). TheGuardian. Retrieved from A Beaty Contest Was Judged By AI and The Robots Didn’t Like Dark Skin: https://www.theguardian.com/technology/2016/sep/08/artificial-intelligence-beauty-contest-doesnt-like-black-people
Luinge, H. J., Kleywegt, G., Van’t Klooster, H., & der Maas, J. (1987, August). Artificial intelligence used for the interpretation of combined spectral data.
Nichols, M. R. (2017, October 16th). Could AI Help Stop Fake News In The Near Future? Retrieved from Big Data Made Simple: http://bigdata-madesimple.com/could-ai-help-stop-fake-news-in-the-near-future/
Nunez-Mir, G. C., Iannone III, B., Pijanowski, B., Kong, N., & Fei, S. (2016, July 12). Automated content analysis: addressing the big literature challenge in ecology and evolution. Methods in Ecology and Evolution, 7(11).
Pedro, V. (2016, March 12). Techcrunch. Retrieved from Artificial Intelligence and Language: https://techcrunch.com/2016/03/12/artificial-intelligence-and-language/
Sentence Structures. (2017, November 15th). Retrieved from https://funwithsintext.files.wordpress.com/2013/05/ioesgg.jpeg
Sulleyman, A. (2017, November 2). Independent. Retrieved from Stephen Hawking Warns Artificial Intelligence ‘May Replace Humans Altogether’: http://www.independent.co.uk/life-style/gadgets-and-tech/news/stephen-hawking-artificial-intelligence-fears-ai-will-replace-humans-virus-life-a8034341.html