I don't understand your point. They're getting better so quickly it's hard to keep up.
20,326 karma · joined August 23, 2014
I don't understand your point. They're getting better so quickly it's hard to keep up.
Compared to Baba Is You which gets difficult and interesting after like 10 minutes.
1. LLMs aren't run deterministically - they use a bit of noise to improve the output. So even with the exact same input you're going to get a whole distribution of outputs.
2. Feedback from the environment. LLMs (and humans!) work best when they can test ideas. That's why coding agents work so well. If you ask an agent 1000 times to just think of a compression algorithm, it'll probably spit out very samey answers just like a human would. If you give it a test corpus and let it try out different things it will absolutely be able to come up with novel ideas.
Of course, ultimately that's because language itself doesn't follow simple rules like linguists would prefer. They're more like guidelines.
I'm much more interested in their Spectra and Gallery displays which use color pigments. You can even buy e-ink gallery style displays for a totally unreasonable price!
https://inkposter.com/products/inkposter-tela-28-5?variant=4...
Actually to be fair that is not a totally crazy price.
Btw it is kind of weird how it still can't write prose normally but it can write impressively high quality code now. Must be something to do with their training. Or maybe it's easier to quantify good code.
Though I suspect the way we'll get "wild AI" first is through cloud services paid for by cybercrime rather than IoT botnets - they're too underpowered.
Sure if you ignore the most popular desktop OS in the world.
Also overcommit (at least how Linux does it) is dumb. It leads to broken OOM behaviour (killing a totally random process, or more often locking up the system and hard rebooting). And some apps do need to handle OOM properly, e.g. formal verification systems which can use enormous amounts of memory. Jasper has to resort to monitoring page fault stats to know when it is using too much memory.
I am pretty sure my wife could vibe code website or whatever at this point, or if not she will be able to in 6 months.
Last I checked it's because it was generally considered that preventing fingerprinting is practically impossible anyway. It's not like you need a lot of entropy for fingerprinting, and the web API surface is enormous. To actually stop fingerprinting you'd need to make your web browser more or less useless.
Plus I think clean room process needs two separate teams and this is just one guy.
I did find this post from actual lawyers about it, but basically it's a grey area that hasn't been decided yet.
https://www.beankinney.com/news-and-insights/blogs/512000-li...
I would be surprised if that stood up in court but I doubt anyone cares for TA at this point anyway.
It definitely improves the results if you do think but I don't think "you still have to think" is a safe space that's going to save us from unemployment.