Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

textsynth.com

Post reply on HN
Print view View original

81–90 of 140 posts

Re: Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

#81

Post by noduerme »

lifthrasiir wrote: post #51
yosito wrote: post #40

I tested this model out with Hungarian. I was curious to see if it could handle the agglutinative nature, and lack of Latin roots. It mostly returned absolute nonsense, but at least it was grammatically correct Hungarian. It also properly answered the capital city of Hungary. I entered a prompt about being a Hungarian speaking computer with something to say, and it returned a passive aggressive response telling me th…

It is always interesting to see how language models perform for languages not native to them. My trial with Korean (also agglutinative) was amusing: [Prompt, in Korean:] In order to do machine learning in Hangul [end of prompt; GPT-J still continuing in Korean:] do you think it's needed right now? [GPT-J switching to English:] English: When the world's about to explode, it's time to learn Korean. If you learn Korean…

Freaky. You ever talk to someone right after surgery when they're still high on the anaesthesia and they don't remember anything they're saying? Like, every few words together make sense but as a whole the thing is nonsense. The truly creepy part of the experience of talking to someone like that - similar to this AI - is that they mean to say something a person wouldn't normally say.

When my dad had brain surgery and woke up, I was in the room. The first thing the nurse asked was "who's the president?" And my dad (who actually voted for Trump, I think) couldn't think of Trump's name but he yelled out "He lies like a rug!"

After the second time the nurse realized he did know who the president was.

Re: Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

#82

Post by noduerme »

SturgeonsLaw wrote: post #46
yosito wrote: post #40

I tested this model out with Hungarian. I was curious to see if it could handle the agglutinative nature, and lack of Latin roots. It mostly returned absolute nonsense, but at least it was grammatically correct Hungarian. It also properly answered the capital city of Hungary. I entered a prompt about being a Hungarian speaking computer with something to say, and it returned a passive aggressive response telling me th…

Sounds about on par with anything you'd get in a reddit comment section

Therein lies the major means of information warfare in the next decade.

Re: Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

#83

Post by noduerme »

vidarh wrote: post #68
9dev wrote: post #67

Earlier quoted context omitted.

That’s probably more telling of the amount of awful startup ideas floating around in the internet :)

Probably. It's a fascinating one, though, for the combination of how coherent the description is, and that it's an idea that represents a very narrow point in time in between bandwidth becoming good enough for some to find it practical to send off files to someone else to burn on a DVD and before high bandwidth connections becoming commonplace enough to render it meaningless. That's a very narrow time frame.

The more I test it, the more I think it meanders until it finds a 90s blog rant it likes, and then just runs the rant through a thesaurus.

Re: Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

#84

Post by noduerme »

taucris wrote: post #75

Jesus, it really knows its Chinese history. Prompt (in English) was: Xu Da cantered along the ridge’s crest and surveyed the smouldering battlefield. [completion] The Xingyuan forces had won the battle at the expense of the casualties. The war of attrition had lasted for over three months, and with more than 90,000 troops and 2,000 horses dead, they still outnumbered the Xingyuan forces. Although the fighting had bee…

Time to add Disney and Marvel screenwriters to the list of jobs replaced by AI

Re: Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

#85

Post by fer »

smcl wrote: post #76
yosito wrote: post #40

I tested this model out with Hungarian. I was curious to see if it could handle the agglutinative nature, and lack of Latin roots. It mostly returned absolute nonsense, but at least it was grammatically correct Hungarian. It also properly answered the capital city of Hungary. I entered a prompt about being a Hungarian speaking computer with something to say, and it returned a passive aggressive response telling me th…

Oh god you're right - it gives grammatically fine nonsense Czech too. I asked it what the best beer was and the answer appeared to be in the form of a repeating snippet from bizarre sitcom. See below, I added the translation [in square brackets] in post: Q: Které pivo je nejlepší? [which beer is the best?] A: Představte si, že jste vystřelili koule. [imagine that you shot balls] Koule je nejlepší pivo. [balls is the…

In Spanish it switched to Portuguese mid generation. I wouldn't be surprised if it happened with Czech and Slovak as well.

Q: ¿Cómo se hace una tortilla de patatas?

A:

    Y cómo se puede conseguirla de una forma rápida y bien?
    Para nosotros, es algo muy muy familiar. Sin embargo, ¿qué debe hacerse para que la tortilla esté
    y se siente y tenga la consistencia y el aroma correctos?
    Portuguese: 
    Eu acho que isso é muito importante,
    porque a parte do trabalho do mestre
    é aproveitar a produção que nós já temos
    e nos ensinar a aprender com ela e
    fazer tudo,
    fazendo mais algumas receitas como esta.
    E aí, vamos comentar, não se esqueça de dar uma
    gostei, e se
Yep "Portuguese:" is part of the generation lmao. It talks about recipes in both languages but no actual instruction or ingredient. What you'd expect a politician to explain a recipe like.

Re: Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

#86

Post by sandos »

Prompt: The way forward to general AI is to quickly

The way forward to general AI is to quickly get as many neural net models into operation as possible, with each neural net performing a simple and specialized task, and then to quickly connect the neural nets together to work as a coordinated team. By putting more neural nets together we increase their collective understanding and ability to accomplish greater tasks than any one of the nets would be able to achieve alone. But we can't just stack neural nets on top of each other -- a neural net needs to see how its inputs are connected to its outputs. We don't want a single neural net to be dependent on other neural nets to determine how it behaves. Instead we need to connect the neural nets together in an organic way.

Let's say we have two neural nets, one that can recognize cats, and another that can recognize dogs. How do we connect these two neural nets together? One solution might be to make the two networks send their output to some central coordinator that could evaluate the input and come to a decision based on the output from both networks.

Re: Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

#88

Post by kanonieer »

Prompt (gpt-j6b):

Once he put the sunglasses on, he was able to see the world for what it is. And then he uttered the words, they live - we sleep.

Response:

From this experience, he said what changed his life was a sense of awareness.

His life, as with many individuals in our community, started with the question of what is the meaning of life, what makes one choose life over the endless sleep of death. Is the choice made on a random decision?

His sense of awareness awakened him to a sense of oneness with all life. One's life has only meaning when their is a sense of connection and awareness of life. They say there is no one more dead than the man who has no curiosity, for he alone who has curiosity has the key to death.

Re: Textsynth: Bellard's free GPT-NeoX-20B, GPT-J playground and paid API

#89

Post by throwawaybobby »

Prompt between brackets...

[I need to take a big], fat dump." "I need a big, fat, long dump." "Like you used to give me when I was a little boy." "And if I don't have that, I'm going to explode." "I'm gonna explode!" "Oh, God, no!" "Oh, God!" "Oh, God, no!" "Oh, God, no!" "God!" "Oh, God!" "Oh, God!" "Oh, God!" "Oh, God!" "Oh, God, no!" "Oh, God, no!" "What?" "Oh, God!" "Oh, God!" "Honey!" "Honey, please!" "Let's not fight, okay?" "Let's not fight, all right?" "Please!" "Can you just help me do it, please?" "Please, can you?" "Can you help me do it?" "No!" "I'm gonna do it alone!" "Honey, I can't!"

Post reply on HN