@LeonDerczynski Consider Bing and Googles LLMs with retrieval also. They give their sources sure, but also they also pose a real threat of reducing click through and ad revenue for sites which produced the content. Plus all the closed models were trained on other peoples data too.
@LeonDerczynski LAION itself is also importantly non profit, Stability, Google, OpenAI etc. could be said to be making money off other peoples data. But that practise is certainly not new. At this point I think any legal action in this area would have an impact on a huge number of orgs.
AI won't take your job. But it will transform it and create new ones.
This NYT article has quotes from economists who specialize in the effect of technology on labor markets, such David Autor, Daron Acemoglu, and @erikbryn :
"Everybody I talk to, supersmart people, doctors, lawyers, C.E.O.s, other economists, your brain just first goes to, ‘Oh, how can generative A.I. replace this thing that humans are doing?’” said Erik Brynjolfsson, a professor at the Stanford Institute for Human-Centered AI.
But that’s not the only option, he said. “The other thing that I wish people would do more of is think about what new things could be done now that was never done before. Obviously that’s a much harder question.” It is also, he added, “where most of the value is."
@davidautor : "60 percent of current occupational specialties did not exist 80 years ago."
"The Optimist’s Guide to Artificial Intelligence and Work" https://t.co/kxgHPqWyA7
@andriy_mulyar@sleepinyourhat@srush_nlp@chrmanning@mdredze@ChrisGPotts Join @StabilityAI, work on Multimodality and long context transformers beyond attention. Instead of scaling up performance on benchmarks keep the performance but scale down model size and inference cost to allow people to finetune their own powerful models much easier.
@FrankElavsky @quarto_pub .. ..and IPFS might be a cool way to decentralise sharing. Definitely ways to go to improve accessability and encourage new thinking! And I appreciated reading your micro-paper. Trying to find a related paper incase you haven't seen it, but struggling to find it frustratingly
@FrankElavsky @quarto_pub Ah quarto_pubs a new one! Twitter or Mastodon make a lot of sense too, all the academics are there anyway. Once I finally figure out a frontend I might look at adding a writing and publishing feature to my site with editor.js..
After 10 rewrites since being stuck on the first 400mb and a bunch of ChatGPT back and forth, I'm still not conceding on a laptop, colab cpu instance and pandas being a viable way to tackle proccessing 7Tb of postgreSQL dumps into an accessible format 😅