AI in Publishing: What to Expect and What Not
With AI gradually taking over automated tasks in almost every industry, the writing/publishing domain is no exception. Most would say, AI applications will improve speed, efficiency and quality. True that. Publishers are working with technology firms in their bid to improve overall work flow and, consequently, customer experience.
AI, it is believed, will do away with unwanted interference with language, as is more often the case with manual editing. Machines, it is thought, will keep changes to the point, restricting to only where deemed necessary. Second, it is expected to result in significant time savings, as machines are definitely faster.
AI could be most beneficial in analysis of information and making recommendations based on that. In other words, this constitutes ‘machine learning’. Machines will, therefore, help in understanding changing customer requirements and catering to them accordingly. In fact, algorithms would be self-generated to meet new specifications. So far so good.
A few things need to be kept in mind, though.
No algorithm is error-free. It requires constant upgrades. Also, today, no one AI solution can fix all problems or do all tasks. Different systems will be required to perform different functions even within the domain of editing/publishing. Moreover, remember the apprehension when IBM was launched? It was believed it would replace manual calculations and eat into jobs. What happened thereafter was phenomenal. Yes, computerized calculations did become the norm but as humans gathered mastery over computer languages, it only paved the way for computers to enter almost every aspect of life, laying the foundation for something much bigger that we are seeing today.
Coincidentally, we are at the same crossroad once again. It is this, therefore, that raises doubts about how valid our fears are. For instance, in the literary field, creativity is the hallmark of any writer. Is AI capable of being creative, not replicate it? Deep learning basically entails machines working based on fixed templates – they replicate the ideation pattern of an individual and produce work accordingly. While this may as of now appear as one of the biggest achievements of technology, probably the biggest limitation lies right here. It is like living in a world divided into groups of people that function along one style of thinking.
Also, language requires a pragmatic approach be applied to it. In other words, responses should be situational – a certain construction of sentence may work in one context but could differ in another. Can machines be embedded with that intuitive sense?
Then there are the security concerns. With so much data being fed and massive input of information, it will be an uphill task to ensure privacy. Think it this way: if machines are intelligent or more than that to replace humans, they could (and would) get smarter to hack the very algorithms devised to protect data!
Honestly, attempting to answer the question raised above is like throwing darts in the dark. Pressed by the need to generate more in lesser time, organizations would be compelled to take the plunge. The scale of the impact when the ‘asteroid’ finally hits, we can only guess. But if we look deeper, there seems to be some time, especially as there is no one answer to all problems, and given the relativity of how soon machines can substitute humans in the respective domains. There’s something no one can deny: humans created machines, not vice versa. The key, therefore, lies in figuring out the key where we can outsmart machines.