My assumption is that I found it here but I'm not sure.
Does anyone know what I'm talking about? I want to find it again.
And it'll happen again. Investors will figure out that what currently exists has limits, and 90+% of the value is in cheating in high school essay writing, spam, catfishing, de-careering non-celebrity artists, or minor chatbots and small text generating utilities that are mostly decorative features for existing large software moats/monopolies.
This time is different from in the past tho, and it's not quite clear how different. AI winter is starting, and it could be the shortest, or the last one.
For now, companies are still losing money by running this. Take github copilot as an example, licenses are at 10 dollars a month (plus your data ofc), but microsoft spends an average of 40 per user by running the machines.
But... two more papers down the line, more optimized algorithms and cheaper hardware...
I don't know the article you mean, but there's a famous event in the history of computer science call the AI Winter: https://towardsdatascience.com/history-of-the-first-ai-winter-6f8c2186f80b?gi=1b737f4eb886
A lot of people think that we'll be in a new AI Winter once the flaws of LLMs are more widely known.
honestly the Wikipedia has a better summary https://en.m.wikipedia.org/wiki/AI_winter
At this point to me it seems like there will be more jobs lost due to companies pivoting into AI than due to AI automating things.
