GenAI platforms like OpenAI, and Vertex AI are some examples of paradigm shifts in the field of AI and almost everyone is gaga about it. Everybody is struggling to get early mover advantages and trying hard to build something using this technology, be it tech or domain use cases. All are over-enthusiastic and ever-optimistic because everyone realized that this is the true democratization of AI for the very first time. Now you can start consuming AI capabilities in your applications without learning about intrinsic details and the functioning of AI.
What makes these technologies cutting edge is their immense capabilities within their foundation models along with amazing abilities to extend themselves using shots, prompts, custom data sets, and that too without writing any line of code.
Now it is interesting to understand that even though this seems to be a promising technology breakthrough of the current era, these tools are already banned, blocked, and not allowed in schools, colleges, institutions, organizations, and so on. Why? Simply because it will impact the creative abilities of students, writers, and programmers along with the possibility of having similar content or artifacts in a competitive world. It is easy to detect plagiarism using handy tools available in the market, but GenAI has the capability to alter the original content and present it in a new form that is undetectable at least with the current set of tools and technologies. Simply, blocking a platform may not be effective since technology is available in everyone's pocket, it will hardly impact their usage. we need to find better ways to detect such content, maybe an anti-pattern kind of concept which is yet to be discovered.
Now, let me assure you that this blog of mine is not generated by any such GenAI tool. I think we are now getting into an era where a blogger and original content generator needs to certify that his/her blogs are not generated via GenAI.
Let me quickly explain what goes into training a model like ChatGPT.
Huge volumes of digitized content, human-to-human conversation samples, and reinforcement learning from human feedback are used to self-tune the model.
I believe that a model like GPT is lucky to have a huge amount of digital content primarily generated over the years by none other than humans with a cut-off date is somewhere around Sep 2021.
As GPT is publicly available for sophisticated content generation, almost everyone is bound to use this today or tomorrow. Now the downside of such democratization may lead to spur usage of synthetic content in the coming decade slowly wiping out original creative content and human-to-human interactions from the digital world. Let's assume the next GPT training will happen in the next 10 years and by then most of the content available to us will be generated by a tool similar to GPT. Yes, history, events, data, and information will be refreshed, but what about creative content and the way human interacts with each other? For GPT to learn further, it is important for humans to regularly publish and create content, but this may not be possible in the future. After all, that's what GPT is focusing on, to provide content more like humans. I presume machines may take over all such jobs in the future. What I foresee is a probable deadlock where creativity will be limited and synthetic.
Now enough of the negative sentiments on such revolutionizing, innovative, and futuristic solutions, let us talk about some probable use cases where this technology can be used to come up with industry solutions to improve productivity, cost savings, accuracy, and less error-prone operations. I was looking at a short video posted by a friend of mine on LinkedIn, it was about Google CEO giving a short demo on how Gmail can now be used to generate meaningful responses, the example he showed was asking for a refund and how to write an impactful email to get your refunds.
I am not against the use case, may be a good pitch to attract mass users, but I don’t get it, why would you do that? Maybe to save time and effort, enhance the content, better grammar, etc., but what if the recipient of this email is again an automated program? What sense this conversation will make? With all of this going around, in 5 years, no one will be able to differentiate between human and AI-generated content be it on email, songs, movies, blogs, books, etc., and there will be tons of such synthetic content everywhere.
Let us look at some low-hanging industry use cases that can be implemented using GEN AI, these are not cutting-edge examples but surely will create considerable business impact.
Smarter Virtual Assistants – Last year, I was sure that ChatBOTs are not contributing much and will get sunset soon due to their inability to form meaningful and human-like contextualized conversations. However, with the invention of GenAI, this will no longer be true. Almost all live support can now be initiated using GenAI to handle initial customer volumes with sophisticated responses using the knowledge database, historical ticket data, and integrated systems. Content filtration, sensitive data, and unwanted message exchanges need to be avoided here. I bet deflection rates will be much lower as compared to what we have to offer on current tech today.
Support Helpdesk – A flavor of the above use case, I would rate this as the prime use case for generative AI, of course, need to merge this with the self-service helpdesk. Custom-trained models developed on historical data with feedback loops will enhance productivity and turnaround time. LLM models guide helpdesk agents to take appropriate actions, Knowledge Database search, and facilitate accurate responses.
Speech Analysis, Search & Automated Response – Yes, analyzing and searching contents from speech, audio, and conversations is very much possible using GenAI, and that too using live streams. Sentiment analysis and interactive GenAI-based IVR, all are possible now.
IT Infrastructure - Deployments & Monitory – This one may prove to be a game changer for the way infrastructure is deployed, monitored, and supported. One of the best use cases for generative AI involves, the consolidation of application, server, and system logs, and then funneling it via GenAI prompts. Even prompt-based infrastructure rollouts can be possible now.
Agent (Human) Monitoring & Training – Remember the sentence, "This call may be monitored or recorded for training purposes". If you are into BPO operations, you may not need to listen to lengthy recorded customer conversations, instead desired information can be pulled out like agent’s performance, sentiments, recurring issues, and analytics using GenAI
Document Extraction – I am sure that GenAI will give document extraction players a run for their money. Now basic OCR + Prompt Engineering can easily extract desired contents from a document/image/PDF.
Transcriptions (Analysis & Summarization) – Any use case which involves lengthy domain-specific transcription analysis, be it medical transcription/understanding legal proceedings, audio blogs, and performing manual operations to get meaningful content can surely be replaced using GenAI.
Content Moderation – Data entry, cleansing, moderation, and summarizing are some of the use cases
These are just the tip of the iceberg, there are many more industry-specific domain-driven use cases that will revolutionize the way IT is done today. Please do share your comments and also uses cases which you think are relevant for such implementations.
Very well written, and good usecase examples