Wake me up when AI can... | Hacker News Reader

Wake me up when AI can...

stoppels.ch

29 pointssigna1112 comments

> An LLM multiplies any two unseen numbers with an arbitrary number of digits by following the rules of multiplication

by the way, has anyone actually tested/measured this rigorously?

I know modern systems had calculators bolted on almost immediately, but how well do modern LLMs deal with multiplication just by talking it out?

Hmm, but these are all programming tasks. AI hasn't made a meaningful positive difference in most people's lives, and it is making a negative difference in many areas.

That's not really what the post is about.

Tell that to all the non-programmer Gemini/ChatGPT users.

My non-engineer friends are have enveloped it into their lives, and I don't see them going back. Some of it is negative, a lot of it is positive.

They just wanted a new digital landlord. I doubt their productivity and quality of life has risen.

Is this about goalpost moving?

Essentially. Although it’s hard to move the goalposts that quickly since when we want from Opus 4.1 being the state of the art this time last year to Fable 5.2 now. Whatever goalpost you might set gets knocked down pretty quickly.

So we’re seeing a lot less of goalpost moving now. The above post highlights this.

it's more about history of goalposts and whether they were reached

In the spirit of manifesting a bright future: wake up me when AI can figure out how to live in peace with humans and not wipe them out economically.

we will never get peace, only a truce in which we are useful or necessary, as batteries or other support.

iRobot would like a word. The probable solution will probably not to our (humans) liking…

"When an AI is capable of creating new explanatory theories that are GOOD (not world salad)" and statements like it miss an important point. Here, let me explain via snark:

"When a person is capable of creating new explanatory theories that are GOOD (not world salad)"

Let's not flatter ourselves, we were never very good at a lot of things to begin with. Then we hold AI at the unattainable standard of zero mistakes permitted.

Reply on news.ycombinator.com