I myself was very surprised by chatgpt; I studied AI in the 90s, AI winter, and, although progress was interesting since 2012, I really didn’t think this would happen anything soon. In my companies, I have used ML for some things, but it was never anything really groundbreaking; just quite boring (as I work in payments).
With chatgpt and gpt3.5 thing really changed for me. So did stable diffusion (esp midjourney), but that is more because I cannot draw even a stick figure if pressured to, so it is enjoyable make images that mean something to me. But chatgpt I see something more in for the future, programming and business; things that interest me.
If anything, it rekindled my interest in ML and I started playing with models on huggingface and colab, which is fun. Made some quite open source things for me to make it easy to try chatgpt on work and personal projects quickly to see if I can do anything with it (even though some commercially viable; if davinci 003 is any sign, it will be way too expensive for most to make a profit on reliably longer term even if it does bring value).
But now everyone indeed is launch projects in high tempo to make money early on, launching projects on producthunt and Twitter, getting a lot of attention and a bunch of them seem to make money even though it doesn’t actually work.
Most are no more than a saas template on the gpt3 api (or chatgpt unofficial as ‘prelaunch’) or one of the stable diffusion api’s, with a bit of lazy prompt priming showing a few cherry picked demos.
Again, I had good results with chatgpt for real things; I used php parser, chatgpt and a lot of prompting (I refuse to call it prompt engineering) to automatically analyse a 300k LoC php project and provide me with md files (and so a pdf of those files) of the reviewed code with explanations. It also generates an executive overview without code.
But if I would have to pay for tokens, because of how you have to (automatically) tune and refeed depending on the output (which I could read, pull through the PHP parser to check if it was complete), I would have a huge loss. With gpt3 davinci pricing, I would be out well over 25k.
You can see that people are now hiring people to quickly create ‘MVPs’ with the purpose of quick pump and dumps. Just get a lot of traction, sell a few subscriptions and flip it. At least many in the nft/crypto community lost their own money by ‘hodling’. Here is seems most ‘products’ are created with the creators knowing it will never work: probably had some great results when manual prompting, made an automated product and now have something they ask (a lot) of money for which does not really work so they need to move it. It are indeed many of the dame ‘get rich quick’/‘the new world order’ crowd.
They did not understand nfts or crypto but spent 1000s (…) on it anyway. They built crappy web3 stuff without value and flipped those fast. So I guess what I am seeing is more akin to web3 as that has the same ‘wow the new future, much launch something fast’ to ‘hmm, this is not really working but I have some hype, let’s flip it’.
I guess it is not very known here but for ‘internet marketers’, people with a lot of Twitter followers and mailing list subscribers (millions of them), this is how it works; they monitor the hypes, create products, send out to subscribers; many of them are starry-eyed fans and will sign up and pay immediately (usually it includes some kind of free ebook; more now as most of these are now mostly generated) and then flip them. They don’t really care about whatever it is; nft/crypto/web3/Ai. They move on and don’t care if it fails or not; they are out already.
These people are actually the driver for others to join hypes; they hear from their guru/mentor etc that this is really go time this time and get invested by trying to join the train. It ruins lives as we have seen with crypto (but I have seen it many times before in the last 20+ years); these people follow the marketers, buy their books, courses, saas etc but never read, follow or use them. If they would have, they would know to get out on a high and move to the next thing. They hardly ever actually do that but keep holding on to the idea they bought into this time (often they are young and too new to know this is a cycle).
I think we will see 1000s of ‘products’ which are half baked API layers or ‘beta’ products waiting for investment by people like you describe; quick upwork hire and suddenly you have an AI company, gold for investors. Many of them will appear for sale in the coming months and people will buy them if they have any kind of traction (and let’s be honest, traction is trivial at this point still).