There is probably a threshold effect above which the technology begins to be very useful for production (other than faking school assignments, one-off-scripts, spam, language translation, and political propaganda), but I guess we're not there yet. I'm not counting out the possibility of researchers finding a way to add long term memory or stronger reasoning abilities, which would change the game in a very disorienting way, but that would likely mean a change of architecture or a very capable hybrid tool.

▲

DaedalusII 9 hours ago | parent [-]

the greatest step change will be when mainstream business realise they can use AI to accurately fill in PDF documents with information in any format

filling in pdf documents is effectively the job of millions of people around the world

▲

oblio 6 hours ago | parent [-]

That would require accurate validation of said documents, which is extremely hard now. Pointing 1 million PDF LLM machine guns at current validation pipelines will not end well, especially since LLMs are inherently unreliable.

	▲	coffeefirst 3 hours ago \| parent [-]
		This is lost on people. A 98% accurate automation is useful if you can programmatically identify the 2% of cases that need human review. If you can’t, and it matters, then every case needs human review. So you lose a lot of benefits to the time sync, but since people tend to have their eye glaze over when the correction rate is low, you may still miss the 2% anyway. This is going to put a stop to a lot of ideas that sound reasonable on paper.