It's been about two years since the splashy release of ChatGPT brought chatbots to the general public's attention, (about a year after another AI tool, GitHub Co-Pilot, made a splash in the developer pool). LLMs have been derided as prone to hallucinate and, in what is relevant to scientific disciplines, bad at fact-checking, logic and math. Two years represents eons in tech, however, and news reports indicate LLM powered engines with improved math skills have been devised. Progress is possible, a matter of training and optimizing performance for particular tasks, developing an appropriate training database or making different hemispheres of the AI brain work together.
SO/SE corporate has embraced AI but the initial encounter was tumultous. A tech blog's account of SO/SEs attempts to engage with the technology:
“If you can’t beat them, join them” seems to be the reaction to Stack Overflow’s recent announcement of its partnership with OpenAI. This collaboration, which merges Stack Overflow’s extensive database of developer knowledge with OpenAI’s AI models, might look like a smart move. But is it possible that Stack Overflow is just trying to cut its losses, having realized it can't compete in the AI-driven landscape? [...] Stack Overflow has found itself in a very uncomfortable position. Since the rise of AI tools like ChatGPT and GitHub Co-Pilot, Stack Overflow has witnessed a notable decline in direct user traffic. Developers increasingly favor AI’s instantaneous solutions over community interactions. This shift threatens the very foundation of Stack Overflow—its engaged, contributive user base.
Talk of regulating AI seems so yesterday. The platform is caught between a rock and another one. This is certainly discouraging for those who enjoy the traditional model, but the choices are understandably to either find a new model or go the way of the dinos:
SO/SE corporate is being bravely enthusiastic about AI, but there is a persistent problem:
[...] we’ve moved well past the inflection point of the AI hype cycle. The customers, partners, and technologists I engage with every day grapple with the very real questions and challenges of how to implement and use AI tools to increase productivity, improve their developer experience, build better products, and drive value. So while change is constant—and AI has certainly kicked off a super cycle in technology innovation and investment—the common thread I see is that trust is the key to building thriving knowledge ecosystems: on our internal teams, for our customers, our partners, and in our broader community.
Can AI be trusted (yet)? Trust in the accuracy of answers is the most important feature of the SE model. It lies at the core of SE's purpose. But does it provide a sustainable advantage? Are there other advantages to human-driven platforms such as SE compared to chatbots:
What role do we have in the software community [emphasis mine] when users can ask a chatbot for help as easily as they can another person? How can our business adapt so that we continue to empower technologists to learn, share, and grow?
Is there room down the road for coexistence of SE with generative AI? If so, how?
The new model:
[Edit ++: I plan to revisit the questions in this post and to perform further comparisons of SE with AIs. ]
[Comment on DALL-E image] The DALL-E image was generated with a prompt requesting that labels be placed above Scylla and Charybdis. The LLM instead placed a single label in the clearing above the ship. Although possibly accidental, it seems a sly defensive revelation of ego by the bot. Rather than be labelled as monsters it chooses to be the hero!
