I’m getting a lot of inquiries lately about ChatGPT-3, an AI natural language model from Open.ai that has learned to churn out believable writing in different styles – including poetry – based on a simple text prompt.
It’s no wonder that interest in this project is peaking right now, besides the news of a $10B Microsoft investment in an OpenAI partnership. We are having a collective ‘uncanny valley’ moment where we are starting to realize that communication with AI models might someday resemble human interaction closely enough to be indistinguishable.
Perhaps we are witnessing a real step-change in AI that will influence our personal lives. But as an industry analyst, I’m still seeing more business utility in ‘applied AI’ or ‘augmented intelligence’ approaches that help humans become more effective at the same time.
We’re still quite a ways off from the singularity. While fascinating, ChatGPT is just a waypoint on a continuum of AI and machine learning tools that are forging new capabilities for intelligent process automation and low-code development.
AI goes to school
ChatGPT is making headlines in large part due to its propensity for fulfilling academic writing assignments. We’ve got ChatGPT passing admissions tests and written exams in a minute with researched prose.
Who doesn’t dream of what high school and college would have been like if we could have auto-generated that book report or term paper?
And just imagine being an instructure, trying to figure out whether or not the text was written by an AI, when the model was tuned on real student papers. A thorough and careful student writer may come off sounding like a bot.
The resulting backlash of universities banning ChatGPT altogether hit quickly, although there might be no way to police the addresses students can visit on their own computers. Maybe schools will re-commit to bringing back test proctors or even oral exams.
Will bots replace knowledge workers?
The largely remote tech workforce that expanded significantly during the pandemic years is also looking at the advance of AI development with trepidation, especially given current geopolitical and economic conditions.
If AI can code, and talk to other systems and humans, why would companies need to hire developers anymore? If AI can create legal contracts, examine X-rays, and process insurance claims in seconds, what happens to the knowledge worker?
I believe most of these fears are unfounded. Yes, there will be certain niches where AI would win in a zero-sum fashion. If anything, the combination of AI and human knowledge workers will reliably create more powerful results, automating away the most difficult and labor-intensive aspects of data capture, research, and correlation.
Knowledge workers are brought in only for unique decision points or to tune the model based on interpreting a need for course correction, technically increasing their productivity by 10x or 100x or more with better predictive accuracy.
And, as for developers, it’s important to remember that while a conversational AI like ChatGPT can generate code just as well as dialogue, its reinforced goal is to be believable to the reader, rather than accurate. AI can review a developer’s code, and a developer is still needed to review the AI’s code.
Between conversation and intelligent process automation
AI is already blurring the lines between many of the technology spaces we cover, as it moves through markets and bends existing definitions.
ChatGPT is only one kind of conversational AI that is geared toward a set of individuals – it’s really an outlier in the space. OpenAI exposed its natural language understanding functionality as a loss leader to attract interest and investment, sort of like YouTube in its early days. (It’s rather expensive to host on-demand compute and data storage, so the free version is often unavailable to users).
There are several other vendors out there producing conversational AI models that are purpose-built for business and attracting actual revenue. Customer service concierges, intelligent agents, and researchers for data collection and curation are available for hire to assist humans.
But before we get too hung up on conversational AI, it has several AI counterparts that are being used or embedded within other technology spaces. Conversely, you could say any automation software that is improved by human and system interaction and algorithmic-based optimization starts to cross over into the machine learning realm to feed AI-like conversations with people and systems.
- Digital process automation solutions are using AI to self-discover, monitor and connect multiple functions, with that conversation happening through APIs and system-to-system communications.
- If an RPA bot becomes sophisticated enough through process mining and complex decision algorithms to handle complex unsupervised work, it starts to have a machine learning basis for layering on AI calculation and decisions.
- AI Ops solutions are making observability and security tools function far better by sifting vast data streams, reducing alert fatigue and resolution times for DevOps, SRE and SecOps teams, allowing them to focus on solving business-critical issues.
- Still other data movement and processing AIs are doing the essential work of image recognition and language understanding and interpretation, converting unstructured visuals and audio signals into meaningful data.
There’s a lot more than I could cover in one column without leaving out and offending many great projects, as usual, but ChatGPT is just one leading indicator of a much larger wave. A big analyst firm calls this space ‘hyperautomation’ – kind of a big hairball where all of the threads of so many AI-like features, intelligent processes and tools get tangled together.
The Intellyx Take
Another question I get asked sometimes: Why don’t I just have ChatGPT write about my solution instead of an analyst?
Am I worried about a chatbot replacing Intellyx? Nah. Bring it on!
I’m not saying we’re such great thinkers and writers that we could never be replicated by an AI. However, we do bring our own unmistakable style and identity to every piece. Our content is based on irreplaceable life experience rather than data synthesis.
To denote this, you might notice that we’re now putting a “No AI Used Here” disclaimer at the bottom of every one of our posts and whitepapers, and even this Cortex column – as tempting as it would have been to get a bot to write it for me!
©2023 Intellyx LLC. Intellyx retains editorial control over the content of this column. No AI was used in the creation of this text. Image source credits: Franck Vervial, Kim Love, flickr open source composite, CC2.0 license.
