A.I. is on a collision course with white-collar, high-paid jobs — and with unknown impact::Technology has disrupted many workplaces. Artificial intelligence like ChatGPT may have an outsized impact on higher-paid office jobs, experts said.
A.I. is on a collision course with white-collar, high-paid jobs — and with unknown impact::Technology has disrupted many workplaces. Artificial intelligence like ChatGPT may have an outsized impact on higher-paid office jobs, experts said.
This is one of the best takes I’ve seen for a while. LLMs seem like they are the “new Google” in that just searching for information super-charged productivity. Now, instead of using some google fu and having to wade through some links and read, someone can just ask a direct question to an LLM and get a reasonably good answer that may or may not need some work to fix up.
In fact, I’ve started using GPT to summarize large reports/emails and generate the base code for projects so that I just have to tweak it. It has made work that would take hours or days into an hour or two. Honestly, GPT and Llama have made me a much more productive person. Understanding how to use LLMs to one’s advantage is going to be a skill going forward just like how effectively using a search engine is a skill now. It’s not a skill that will likely be appreciated (much like how effectively googling isn’t), but it will set workers apart.
My favorite has been exploring models and obscure software that is challenging to get working. When I get the inevitable failure, I just paste the entire error message into a WizardLM 30B model and I usually get quite helpful insights. I’ve gotten much further into compilation and finished installing projects I never would have managed otherwise. It has expanded my bashrc, and commands knowledge substantially. Sed, awk, and regex are easy now. I can practically get an AI to exit vim for me.
I just doubled the sysRAM in my machine 30min ago and am a quarter of the way into a 70B Llama 2 instruct GGLM. If the jump from 13B to 30B is any indication, this should be around 95%-98% accurate even with obscure technical questions.
Yeah! This is a great use! For me, I needed to get a prototype webapp with integrated prediction model up and working using various libraries, and I was way behind. Instead of just trudging through the documentation to learn how everything worked, I just had GPT-3 spit out a starting point, and then I just debugged from there and added features as needed. The only downside is that it’s made me impatient with developing other software with private libraries.