[ajross]

> I can't get Gemini to not cut me off mid-sentence and reply to only half of what I wanted to say.

This kind of criticism is unpersuasive: "USENET and ftp are fine! Why do I need this monster 'mosaic' thing which needs 8MB of core just to try to see some buggy 'dynamic page' someone threw together where the server is always down anyway!"

New technology is incomplete, rushed to market, inconsistently reliable and rapidly improving. News at 11!

That doesn't mean Googlebook is going to take over the world. But no, you don't get to live in 2019 forever either.

[gumby271]

I would say we're reaching new levels of incomplete and rushed to market though. The original Google Search, Maps, Docs, etc. weren't feature rich, but what they had worked well. I'm not saying they were perfectly built, but they weren't falling apart to the degree Gemini and other Google products are. And Google gets to do this now, no one can compete or leave them, so why do anything well.

When Google first added Gemini to Sheets, I tried it and it completely hallucinated transaction data on the bank statement I was testing it on. It's worse than incomplete, it's broken and dangerous. But Google gets to slap a disclaimer on Gemini that says "results may be fucked" and they get a pass to ship it.

I have no expectation that Google can improve its new products since everything they've been doing is going in the opposite direction. You're not seeing pushback on embracing new technology, its purely about Google not being trusted to do things well.

[ajross]

> I would say we're reaching new levels of incomplete and rushed to market though

The year 2001 called, it was laughing hysterically and I couldn't quite make out what it was saying.

[gumby271]

I think I'm too stupid to understand what point you're trying to make, sorry.

[ajross]

It was that the AI boom/bubble is a new technology in its infancy. It's not comparable to the stuff you mention, which were all[1] second wave "Web 2.0" products being introduced into an established market. The proper comparison would be to the period in the mid-late 90's up until the dot com burst, when "internet" technology was sudden and gestational.

And it sucked. Nothing worked. TCP was "established" technology that needed tweak after tweak. The whole stack failed regularly and catastrophically, at every level from backbone routers to browser rendering. People were gluing together solutions to problems using tools written on 1970's teletypes because the "new" stuff from Microsoft was even worse.

It was glorious. But it sucked. And the same is true right now. Nothing you use is going to do what you want exactly. And we all get to figure out how to make it not suck together.

[1] Except Google search itself, which was a comparative latecomer (anyone remember DEC and Altavista?) but still part of the initial rush. And thus is sort of the exception that proves the rule: the one product in the infancy rush of a new technology wave that doesn't suck (or sucks least) gets to be the next tech giant. Seems like the smart money right now is on Anthropic, but we'll see.

[gumby271]

> And we all get to figure out how to make it not suck together.

Well I appreciate your optimism I guess ¯ \ _ ( ツ ) _ / ¯

[jaapz]

Youre looking at the past with rose tinted glasses.

[gumby271]

Were there features in original Google docs that would insert random garbage? I'm not saying those products didn't have bugs in 2001, but I don't recall them having features that were both barely functioning while also being touted as the most advanced amazing thing ever, and that's what modern Google is doing.

[isityettime]

I used Google Docs before Google bought them. They were created by a startup called Writely. I don't remember bugs that inserted random garbage.

[isityettime]

I never mentioned any alternative to the Googlebooks, let alone said something like "$PRECEDING_PRODUCT is fine". You're refuting a kind of categorical naysaying that I haven't actually done.

Maybe some day, someone will make a computer with integrated digital assistants that virtually everyone really wants. Maybe that someone will even be Google.

But I wasn't writing sci-fi or putting together a pitch deck on Google's behalf. I was considering the actual technology (and priorities) they have on offer.

[mrhottakes]

Your criticism is unpersuasive; you haven't laid out any actual argument, just sarcasm.

[disqard]

Your argument has merit.

I'm just going to point out that NCSA Mosaic was released in 1993, and the hand-wringing over it being an absolute memory pig is more 1999 than 2019.

But your overall point is well-taken.

[ajross]

FWIW, I wrote 2019 as a representation of "last year before AI mattered" and not when people cared about the early web. Also by 1999 Mosaic had been forgotten and everyone was running Netscape and IE4. :)

To wit: complaining about AI software quality right now is isomorphic to whining about how much the early web sucked. It did suck. But it was still the future.