“The value of an idea lies in the using of it.”
— Thomas Edison
New, innovative, and disruptive technologies are not without their detractors. This has certainly been true of ChatGPT which I’m guessing everyone reading this post has heard about and likely played around with. Some of these critics are pioneers in Machine Learning themselves, well-known, respected, and widely followed in the community. They have been ranting, sometimes daily, in online forums that ChatGPT’s output is “nonsense”, “is garbage”, and that the whole thing is nowhere close to being “intelligent”. To prove their point, they are posting (often hilarious) screenshots showing, for example, how easily ChatGPT can be fooled into believing 2+7 = 10. Throw in some more ChatGPT trolling and a few expletives in the ensuing chat forums and … you get the idea.
They have a point. However, they are completing missing a very important point.
Where they are correct is that ChatGPT’s goal is not to seek the truth in answering questions. This is because (very roughly speaking) it’s a large language model (LLM) trained to provide plausible and confident sounding answers; this is not the same as providing correct ones. You need to look no further than OpenAI’s CEO, San Altman, who tweeted in December 2022: “ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness. It’s a mistake to be relying on it for anything important right now. It’s a preview of progress; we have lots of work to do on robustness and truthfulness.”
The point that the detractors are missing is that the release of ChatGPT is, itself, the significant moment. The fact that all of us got to see this innovative technology in an easy-to-use and accessible interface *was* the milestone. Prior to this, LLMs were largely the domain of data scientists and developers. Sure, if you knew Python and had access to a Jupyter notebook, you may have been able to play around with these models; but we’re still talking about a very limited population.
We have to prepare for a world where there are hundreds of these LLMs, interacting with us, performing specialized tasks, and, providing answers. Whatever their strengths and weaknesses, whatever ethical issues they raise, and whatever the best practices for using them — all of these discussions should and must happen in the public domain. Indeed, it’s through the exponentially fast adoption and use of ChatGPT that so many of these issues have come to light to the broader public in the first place.
Data Science is too important to be left to the data scientists.
Despite all its flaws, ChatGPT has given the world a front-row and hands-on seat into the most innovative and exciting area of AI.
I for one am glad that this beautiful, imperfect, and (yes) very helpful, platform was released to the world. I didn’t make any money off this AI IPO, but I’m smiling like I did!
This article was originally published on LinkedIn on 2023-02-12.