Let the machine do the talking (and reading)
We like to think of ourselves as different from animals because we can speak. But, we may be in for a shock if we won’t be able to distinguish between a machine or a human writer.
Various open-source models are available and are being developed to simulate human writing, summarizing, and natural language learning. These tools may still be in their opening chapters in some cases. Still, experimental chatbots have become increasingly uncanny over the past couple of decades.
Besides just chatting to us humans and learning simple patterns, how do these language-learning algorithms do it? One example is through crawling the internet and using probabilities to predict what makes sense to say in a particular context. A parallel can be drawn to how we learn a language as kids: by repetition and immersion.
Essential functions of these algorithms include generating texts based on a given idea, following precise instructions about what text to generate, and generating summaries and dialogues.
The above immediately opened up commercial possibilities for anything that requires reading through thousands of documents. For example, DocuSign (NASDAQ: DOCU) snapped up at this chance for AI-assisted contract analysis. This sounds boring on paper, but it’s okay because you probably won’t want a robot lawyer anytime soon.
Another critical capability for machine writers is following creative processes, such as script-writing. Take, for example, OpenAI has a language model. It can take the start of a conversation and continue that conversation and create original dialogues. Mixed with the creativity of talented scriptwriters, could this be a new way forward to create stories in a totally unique way? Indeed, it would help with writer’s block.
Imagine an independent film studio that had characters created thanks to machines. After being fed with all the works of J. R. R. Tolkien, Jorge Luis Borges, or Isaac Asimov, the results could be unpredictable and intriguing.
Although it may seem like there is nothing new under the sun, reality shows us that there is still a vast amount of expansion for human creativity. Sometimes it means getting a different kind of input and output.
Whether in imaginative or legal texts, all previous creations will ultimately be an inspiration for future authors, or maybe they will be lost in the midst of time. Machines might be a tool to help preserve and reiterate upon all that content, just by using it as a learning source.
On the other hand, we should not let machines do to writing what calculators have done to our mental arithmetic. It’s easy to get complacent when the machine can do all the thinking and calculating. Thankfully, at least for now, I won’t have to end this article with this plot twist: “This was a machine doing the writing!”