I expect the markets are reflecting that soon there will be more competition.
It'll take time, and as LLMs improve, it'll take even less time.
That's not the situation we're talking about though. It's someone saying "hmm, I need to edit this picture. Can I get ChatGPT to do it?" where 3 years ago they would have had to buy Photoshop and learn how to use it.
Similarly, if they need a tool to batch-convert a thousand images, they're getting an LLM to construct the specific tool they need in a couple of hours and then running that, rather than buying a software product that can do it.
You don't need a whole dev team to build a one-off tool for a specific job, which is probably 90% of the demand for those software products. LLMs are becoming the general-purpose tool for a lot of use cases.
Yup, same reason you can't throw manpower at a software project and expect a proportional outcome (Brooks's Law). AI amplifies what's already there; it doesn't conjure taste or product vision out of thin air.
This is something I really hope takes off for the common person. ChatGPT is perfect for bespoke little programs that do one thing and can be discarded after use.
> It'll take time, and as LLMs improve, it'll take even less time.
People have written great software in ed(1). We have tools like uxn[0] written on potato computers and billions and years later, we still have to hope for AI output.
One technique I’ve used for cleaning up AI-generated images, was a Python script driving ImageMagick-and an LLM helped write the Python script (although it took a few iterations, because the LLM’s first attempt didn’t actually work)
That's my best-case scenario as well: LLMs are scripting languages for a broader audience. They just barely automate busywork, but are not a reliable foundation.
>You don't need a whole dev team to build a one-off tool for a specific job, which is probably 90% of the demand for those software products. LLMs are becoming the general-purpose tool for a lot of use cases.
No, all of these tools have 90+% revenue coming from B2B sales, consumers dont buy software products anyway. All of the software purchases are tax deductible so corporations buy even if they use very little of it.
Again, for 90% of use cases it will be good enough. For the 10% of use cases where it's not, yes, they'll still need to buy and learn Photoshop.
> No, all of these tools have 90+% revenue coming from B2B sales, consumers dont buy software products anyway.
I don't know the exact market breakdown of Photoshop, but I suspect it's not 90%+ B2B corporate. And my point was that even then, most corporate users are not going to need the entire Photoshop feature set, they're using it for one or two tasks that could be done better by a bespoke tool.
> All of the software purchases are tax deductible so corporations buy even if they use very little of it.
This is true now, but will change once procurement and accounting departments realise that LLMs can replace most of it.