i think the concept of "professional software", big GUI apps which provide many powerful features for expert users with specific and niche workflows, which the industry was largely built on, is just going to totally fucking die. open source people don't care, they just like to cobble together a thousand ever-changing scripts. and big corps don't see any money in it. their temptation to replace "you can learn these tools and become adept" with "AI will do it for you :)" will become irresistable
0 Boosts
0 Likes
9 months later
@jk I've long been concerned about overspecialization, information silos, and tunnel vision. I've been advocating for generalists to connect the dots and prevent isolated experts from losing the big picture. I still think this is a problem, generations of specialized climate scientists didn't avert climate change, generations of expert democratic politicians couldn't head off the rise of global fascism.
But now I have a new fear, the death of expertise, which would weaken generalists too.
@skyfaller 'trad' information tech like the web (cf Memex from 1945) really should help generalists connect with specialists (and bring specialists in different fields together), as well as surface buried or obscure ideas. LLMs, through latent-space relationships in their corpus, as well as endless chain-of-thought reasoning filtered through 'formal' evaluation, might imitate this. possibly. but they may also erode the skills and social structures that make those connections valuable or useful
9 months earlier
i think this is one of the reasons for productivity stagnation in 'advanced world economies'. to become genuinely exceptional at anything, or even to be reliably good enough that you could teach others, requires you to do it for like, 15 years. an increasingly large share of people never do any single thing for that long anymore. for technological reasons, and economic ones, everybody's forced to repeatedly be an impostor-syndrome beginner at everything now. hence all the grifting everywhere
9 months later
@jk People seem to forget that even if someone is already an expert in whatever programming language a company uses, that company still has a bunch of internal software and processes that take a *long* time to learn and master. Retaining those experts is vital, and yet companies act like we're all interchangeable cogs.
@rbender learning has always been a cultural thing, based on mimicry of practices and rituals, motivated by fitting into social structures formed by individuals' aptitudes and relationships. 'mentorship' etc arises from those situations. the desire to learn how to do or make things, even alone, still comes from a need to connect with others and fit into a culture of peers who make similar things. disbanding teams, or dislocating junior and senior roles, destroys that emergent cultural knowledge
9 months earlier
perhaps there's something inherent to digital technology. creating objects is much harder than creating data, but so is destroying them. physical tools continue to work forever, but software, despite being free of physical constraints and costs, is slaughtered relentlessly every time apple et al deprecates something. so you get a very high turnover, but little accrued value. churn in methods, churn in tools, churn in employees. a high-tech, high-velocity world, alas measured not in mph, but rpm
in the last few decades there's been an erosion of the concept of being serious, long-term, about any specific job role, or career path, which started long before AI. before, you could invest millions into developing software for a specific professional, and reap years of profit if it became an industry standard. but everybody i know loses their job every couple years. employees no longer expect loyalty, so long-term specialisation is a poor bet to make. deep, multi-decade learning is a rarity