So what's wrong with ▒▒▒▒▒▒▒?
- So what's wrong with cryptocurrency?
- So what's wrong with internet advertisements?
- So what's wrong with Generative AI?
My blog is insignificant, and I intend to keep it that way! But besides being a waste of bytes, it's part of the internet, and the internet is very important to a lot of people!
Let's play "So what's wrong with ▒▒▒▒▒▒▒?"
So what's wrong with cryptocurrency?
First of all, it has dubious utility except for speculation and criminal activity, and honestly sometimes those are the same thing. Some people, nice good-meaning people, don't mind that it's wiping out decades of green energy advances, for reasons unknown (lmao, I mean exchanging fake internet money for real, usable, money).
Even the purportedly green proof-of-stake cryptocurrencies are still dumb ponzi- schemes: if you came in early you're smart and rich and if you came in late you hope to become the former.
Grade: F-
Bottom line: Stay away.
So what's wrong with internet advertisements?
Either you pay for it, or you're the product. There's a lot of wisdom in that. People my age used to watch a lot of linear TV, and we saw a lot of ads! That fucking sucked, so when fast home internet met Bittorrent, broadcast TV became irrelevant. Netflix and others improved on that by making it even easier (for a fee). Too many services and amazingly, fucking ads, are probably going to make piracy come back like Houthis in Aden.
Website ads are delivering malware straight to your browser, and there's not much you can do about it, unless you install a privacy-respecting web browser and get a good ad blocker.
Grade: E
Bottom line: Start paying for IT (amazing pun) - Mastodon, Email, and much more. Reject your surveillance capitalism overlords.
So what's wrong with Generative AI?
Great for coding. Great for helping you outline a new idea. Great for creating page after page of SEO optimized bullshit to get more eyeballs on ads.
Hilariously, generative AI probably has a bright future! And that bright future looks like this:
- Replace free online resources like Wikipedia with a $20 OpenAI subscription
- Make concise texts longer for no reason other than to make them look more impressive than they are
- Make long texts short so that our dumb monkey brains don't have to read so much
Unfortunately for the users, general-purpose LLMs are extremely mid. Mid in the literal sense: they express the average of their training set. Meaning that a generated article about Roman windows from the Middle Ages starts generating text about MS DOS after a title like "What existed before windows?".
But you should enhance your LLM with a Retrieval Augmented Generation (RAG), someone inevitably thinks. The most successful applications will use techniques like that, but currently that's not what most people are interacting with. Maybe one day.
internet advertisements
Let's visit our old friend internet advertisements again! Not many people are going to pay $20 or $10 or even $5 for a Generative AI assistant. So you'd better be fucking ready for the coming boom of juicy Coca Cola descriptions in the middle of your generated text.
Even if Copilot is admittedly probably going to be very useful for developers, developers are a tiny fraction of the addressable market (every person on the internet).
Microsoft is losing $80 per copilot user per month, and the AI revolution isn't exactly environmentally friendly either. That might change though: hardware and software will keep getting better, but will we put that to good use or will we use the slack to create even bigger things?
Grade: C
Bottom line: Has a lot of utility when applied right, but we have yet to apply it to many things that are not Very Stupid.