Battle of the Chat Bots
By michaelskelley
Battle of the Chat Bots
ChatGPT is having its superstar moment, an overnight media sensation that (like many other media superstars) is actually built on decades of plodding, iterative development. In the initial tidal wave of press and speculation, it often takes the form of the 21st century’s transformative moment — Christensen’s disruptive technology for AI, entering at first as a better search engine with no cost to the consumer, but with the potential to create and redefine every on-line intellectual transaction.
ChatGPT is more like us — literally like us — than many other technologies, deriving its strength from a neural network underpinning, pursuing the penchant for pattern recognition that is the obsession of the human brain. But unlike our brains, it has no limit to what it can ingest, combing through an internet estimated by Google’s Eric Schmidt this past Feburary at 5 trillion megabytes. To make sense of what it finds (or rather to find the sense most likely to please the user it is interacting with) GPT-3 used 175 billion parameters to adjust what it was finding to make sense.
What makes sense to the user can, of course, be deeply flawed, potentially creating an echo chamber of authoritative, convincing and self-validating argumentation of epic proportions. OpenAI acknowledgement, sheepish and mildly remiscent of the “What have we wrought?” moment of Manhattan project participants, that “Internet-traind models have internet-scale biases” is an understatement. In a time of fiercly divided political extremes, the fallacy of alternative facts based on internet postings is concerning. In the ancient teleogoical debate of whether we can divine the intention of the clock-maker from the clock, the rise of AI front-ended by ChatBot technology comes down clearly on the side of “we have met the enemy and he is us.”
But the potenial for education and research to benefit is enormous. Therapeutic support, recursive and friendly training modules, evaluative mechanisms to support cognitive capacity to ensure safety are promising, entry-level applications. The potential for AI and chat bots to lead to pattern recognition in the light of DNA sequencing knowledge and curative therapies are profound. As OpenAI and Google start the competition to engage and promulgate this burgeoning technology for commerical benefit, it’s time to start talking about some guardrails for how this technology could potentially be implemented. Chat bot applications will quickly become less than benign; developing common-sense standards for their use through professional committees that combine the need to allow creators to explore and profit from their creativy while also protecting the privacy and rights of individuals in a free and open society is better served as an ongoing, iterative process than in reaction to crisis.
For a look at how ChatGPT came from, check out this article from the MIT Technology Review.
