Hacker News Reader

Comment by johnxie | Hacker News Reader

On the topic of AI, you've brought up some valid points, and I understand the skepticism, especially given the AI hype cycle. However, there's also plenty to get excited about beyond the marketing gimmicks.

Consider the recent Google demo. It was recreated and, while still fun and impressive, it's definitely not ready for commercial use. You can see it at https://sagittarius.greg.technology. This demo is a preview of what's to come, and I think it's important for us to stay open-minded.

I believe that in 2024, we're going to witness a significant expansion in what's possible with AI, while costs decrease. Increased contributions in the open-source community will likely fuel another wave of startups and new applications.

View on HN

My current reference class with AI today is 3D games in the 90s.

Doom has just turned 30; when it was introduced, it blew people's minds with the quality, and fuelled a moral panic about it being a "murder simulator". Every year or so after that, a new game would come out, be hailed by the press as "photorealistic", and then forgotten as the next in the cycle replaced it… but even then, they're called "Doom clones", and the original doesn't disappear.

I even remember one of the magazines bemoaning that Riven was pre-rendered rather than real-time "given Quake proves the PowerPC chip can handle polygon-based gaming" (or something close to that quote).

I think we're in a similar phase with AI: ChatGPT-3.5 blew people's minds with the quality, and fuelled a moral panic about it being a fully automated plagiarism/cheating engine, we keep getting news stories about new models or ways to get much more out of any model, which are then forgotten as the next hot thing repeats the cycle… but even then, they're called "ChatGPT clones", and the original doesn't disappear.

It's only an analogy; I don't know how long we have to wait before AI becomes an almost unnoticeable enhancement of everything in the kind of way CGI did.

Check out this on AWS's Q from Corey Quinn .

https://www.lastweekinaws.com/blog/aws-degenerative-ai-blund...

Generative AI is awesome but also crazily overhyped.

Remember when Siri / Alexa / Cortana / the Google thing were the big thing?

Now presumably there are more Alexas in cupboards than anywhere else.

That Corey Quinn article is great, thanks for sharing. It's a more thorough mirror of my own experience with Q. After seeing the obnoxious popups all over the AWS docs I tested it out with a few simple questions about AWS services.

My main takeaways were 1. It refuses to answer a lot of simple questions. 2. It provides a lot of factually incorrect answers.

I didn't even try any complicated or subtle questions; it failed at simple factual asks.

So many companies are rolling out obviously half-baked and useles Gen AI features that I would be embarrassed to release. All of these terrible user experiences must be damaging some company reputations. The only justification I can see is that many large corporations care more about being able to tell investors and journalists that they have checked the box on having the latest shiny trend than they do about building actually useful tools and user experiences.

Reply on news.ycombinator.com