These things, as they are right now, are essentially at the performance level of an intern or recent graduate in approximately all academic topics (but not necessarily practical topics), that can run on high-end consumer hardware. The learning curves suggest to me limited opportunities for further quality improvements within the foreseeable future… though "foreseeable future" here means "18 months".
I definitely agree it's a bubble. Many of these companies are priced with the assumption that they get most of the market; they obviously can't all get most of the market, and because these models are accessible to the upper end of consumer hardware, there's a reasonable chance none of them will be able to capture any of the market because open models will be zero cost and the inference hardware is something you had anyway so it's all running locally.
Other than that, to the extent that I agree with you that:
> GenAI in its current form will never be reliable enough to do so-called "Agentic" tasks in everyday lives
I do so only in that not everyone wants (or would even benefit from) a book-smart-no-practical-experience intern, and not all economic tasks are such that book-smarts count for much anyway. This set of AI advancements didn't suddenly cause all cars manufacturers to suddenly agree that this was the one weird trick holding back level 5 self driving, for example.
But for those of us who can make use of them, these models are already useful (and, like all power tools, dangerous when used incautiously) beyond merely being coding assistants.
These things, as they are right now, are essentially at the performance level of an intern or recent graduate in approximately all academic topics (but not necessarily practical topics), that can run on high-end consumer hardware. The learning curves suggest to me limited opportunities for further quality improvements within the foreseeable future… though "foreseeable future" here means "18 months".
I definitely agree it's a bubble. Many of these companies are priced with the assumption that they get most of the market; they obviously can't all get most of the market, and because these models are accessible to the upper end of consumer hardware, there's a reasonable chance none of them will be able to capture any of the market because open models will be zero cost and the inference hardware is something you had anyway so it's all running locally.
Other than that, to the extent that I agree with you that:
> GenAI in its current form will never be reliable enough to do so-called "Agentic" tasks in everyday lives
I do so only in that not everyone wants (or would even benefit from) a book-smart-no-practical-experience intern, and not all economic tasks are such that book-smarts count for much anyway. This set of AI advancements didn't suddenly cause all cars manufacturers to suddenly agree that this was the one weird trick holding back level 5 self driving, for example.
But for those of us who can make use of them, these models are already useful (and, like all power tools, dangerous when used incautiously) beyond merely being coding assistants.