3 ms·

Is probably more correct to say - my 2.5 year laptop can RETELL space invaders. Pretty sure it cannot write a game it has never seen, so you can even say - my o

by larodi 1y ago

Is probably more correct to say - my 2.5 year laptop can RETELL space invaders. Pretty sure it cannot write a game it has never seen, so you can even say - my old laptop can now do this fancy extraction of data from a smart probabilistic blob, where the original things are retold in new colours and forms :)

simonw 1y ago
I know these models can build games and apps they've never seen before because I've already observed them doing exactly that time and time again. If you haven't seen that yourself yet I suggest firing up the free, no registration required GLM-4.5 Air on https://chat.z.ai/ https://chat.z.ai/ and seeing if you can prove yourself wrong.
- th0ma5 1y ago
  [flagged]
  - dang 1y ago
    Please see https://news.ycombinator.com/item?id=44726957 https://news.ycombinator.com/item?id=44726957 and please stop posting these.
- larodi 1y ago
  I’m using all major models it a daily driver and none of these can create anything I have not spent excessive amount of time explaining. It works for me on the architectural level, but does not change the fact that your expounding on prior information and not a new one. Not sure though why I’m getting downvoted.
  - simonw 1y ago
    You said: "Pretty sure it cannot write a game it has never seen" Then you said: "none of these can create anything I have not spent excessive amount of time explaining" Presumably then they CAN create a game they have never seen if you spend an "excessive amount of time" explaining that game? So are you talking about game design here? It can implement anything, but it can't design entirely new games without human input?
oceanplexian 1y ago
So you're saying it works exactly the same way as humans, who copied Space Invaders from Breakout which came out in 1976.
uludag 1y ago
It's unfortunate that the ideas of things to test first are exactly the things more likely to be contained in training data. Hence why the pelican on a bicycle was such a good test, until it became viral.
MattRix 1y ago
No, that would be incorrect, nobody uses “retell” like that. The impressive thing about these models is their ability to write working code, not their ability to come up with unique ideas. These LLMs actually can come up with unique ideas as well, though I think it’s more exciting that they can help people execute human ideas instead.

by larodi 1y ago

simonw 1y ago

I know these models can build games and apps they've never seen before because I've already observed them doing exactly that time and time again. If you haven't seen that yourself yet I suggest firing up the free, no registration required GLM-4.5 Air on https://chat.z.ai/ https://chat.z.ai/ and seeing if you can prove yourself wrong.

th0ma5 1y ago
[flagged]
- dang 1y ago
  Please see https://news.ycombinator.com/item?id=44726957 https://news.ycombinator.com/item?id=44726957 and please stop posting these.
larodi 1y ago
I’m using all major models it a daily driver and none of these can create anything I have not spent excessive amount of time explaining. It works for me on the architectural level, but does not change the fact that your expounding on prior information and not a new one. Not sure though why I’m getting downvoted.
- simonw 1y ago
  You said: "Pretty sure it cannot write a game it has never seen" Then you said: "none of these can create anything I have not spent excessive amount of time explaining" Presumably then they CAN create a game they have never seen if you spend an "excessive amount of time" explaining that game? So are you talking about game design here? It can implement anything, but it can't design entirely new games without human input?

oceanplexian 1y ago

So you're saying it works exactly the same way as humans, who copied Space Invaders from Breakout which came out in 1976.

uludag 1y ago

It's unfortunate that the ideas of things to test first are exactly the things more likely to be contained in training data. Hence why the pelican on a bicycle was such a good test, until it became viral.

MattRix 1y ago

No, that would be incorrect, nobody uses “retell” like that. The impressive thing about these models is their ability to write working code, not their ability to come up with unique ideas. These LLMs actually can come up with unique ideas as well, though I think it’s more exciting that they can help people execute human ideas instead.