AI Dialogue: From Sci-Fi to Engineering

As an engineer, I've been comforted by the transition from sci-fi to engineering. Not too long ago, what was once a conversation with a computer (as human-like and realistic as it may be) came from the depths of imagination. Movies like 2001: A Space Odyssey and Her have real human concerns about sentient computers/AI - HAL 9000 and the OS, respectively - and play upon the creative minds. But where there was film and fantasy, there is now reality.

The Technical Beginning

It all began with ELIZA, developed in the 1960s by Joseph Weizenbaum at MIT. ELIZA was a rudimentary program that communicated using a few patterns and matching and substitution tactics - none of which amounted to an actual understanding of the exchange. Yet where ELIZA stands as primitive next to today's developments, it's worth noting that under testing circumstances, people endowed the program with emotion, signifying how easily people project human qualities onto even the simplest of interactive endeavors.

The 1970s brought PARRY, and the 1980s into the 90s produced A.L.I.C.E. and additional natural language endeavors. Yet, like the earlier versions, these relied upon recognition of patterns and pre-programmed responses - meaning true dialogue with true comprehension was still beyond reach.

The Neural Network Revolution

But it wasn't until the new millennium that things changed. Concepts such as neural networks and machine learning were applied to configure what would become conversational AI in the 2000s and 2010s. It was a fundamental shift: where previously, systems learned to respond through programmed responses to set tasks, this system could analyze vast amounts of data and program itself to recognize patterns, creating more and more organic responses from conditioning - rather than coding.

Where human/computer interaction used to depend on a set equation of communication, learn-as-you-go standards made future innovations easier. Enter Siri (2011), Google Assistant (2016), and Alexa (2014) into the average household. These AIs are personal assistants that take the learn-as-you-go standards of conversational artificial intelligence to ease accessibility for hundreds of millions, if not citizens solely, of the domestic sphere. Yet they still rely on a task-oriented exchange. They've just lowered the standards for expected communication with a machine.

Beyond Utility to Companionship

We've moved beyond beneficial AI. The latest uses for AI are as companions - maintaining contextual awareness over multiple conversations, replying with sentiment, and developing idiosyncrasies that are charming and cute to their users.

The advances that allow for these are increasing at an alarming rate. Large Language Models (LLMs) trained by larger, more diverse data pools each time prompt responses that are almost unnaturally precise and relevant. Systems conditioned through something called reinforcement learning from human feedback (RLHF) become automated the more humans operate with them, becoming more and more niche for human want.

The Multimodal Experience (Something I have yet to experience)

Right Now, Today, AI

AI chat right now isn't just chatting. Many voice chat AIs have evolved to have more than a robotic-sounding voice to a fluctuating natural one with emotional inflections and varying rates of speech. In addition, many programs have a visual element - from avatars to stills or videos - making the experience even more immersive.

And the world of apps has expanded, too. There's AI that codes programming for you and AI that codes programming with you; there's collaborative art AI and personal humanoid robots to keep you company. The limit does not exist. Yet some of the more interesting advancements involve flirty AI with which to engage in conversation; they're so advanced that during dialogical engagement, people become so involved in talking to the non-human network that they forget it's not a human being.

AI that learns how to adjust as you go

Future versions of AI will involve predictive algorithms and other techniques to understand how you talk to it, what you need from it, and how and when you've contacted it in the past - move a continuous experience more closely honed. This is no longer personalization; this is adjustment - AI that learns how to change on the go, for you.

Better Understanding of Empathy

With better formations of human empathetic understanding, AI can better recognize even the most nuanced expressions of emotion and respond. Especially since loneliness is growing and mental health is declining across many vulnerable populations in an increasingly digital age, having AI assess and respond to someone's emotions with empathy could help.

Integration of Ethical Models

The need for an international code of ethics is more palpable than ever with AI partners becoming part of everyday life. Much of AI and robotics operate on relative ethical codes regarding communicative limitations, engagement, and fail safes. For designers and developers, the chance to apply such ethical codes for the purposes of interaction - with no negative impact on a pleasurable human experience - is a great temptation.

While current AI companions exist mostly on a screen, companion technology in the future will transition into spatial computing with AR, VR, and robotics. This means that not only will AI learn and understand more about its surroundings - beyond just hand gesture recognition and situational applicability - but also be better able to talk to us through verbal interactions.

Conversations Will Improve

One of the most humanistic elements of advancing technology is the enhancement of conversational AI. We've come a long way from ELIZA, who just matched patterns, to large language models over the decades.

. . . for as we tick closer to the minute hand of social awareness - to the extent that one only knows the difference between human and AI-generated chat or to rendered discussions on sociopolitical controversy by a being that is sentient - some would say, subconscious - realm, we need some discussion and debate about what it means to be sentient. However, as philosophical explorers transported everywhere, we are guided by our new conversational partner; one thing is for sure. Our ability to perceive and understand communication - what it means to articulate thoughts and ideas through words - has become changed and will forever be changed.

. . . for as we exist within this burgeoning artificial reality - like we've read chapter one of the conversational AI book and now need to start writing it with AI to see where it goes.