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

Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

chatllama.baseten.co

Post reply on HN
Print view View original

101–110 of 226 posts

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#101

Post by cubefox »

phpisatrash wrote: post #50

Me: who are you? LLaMA: I am a 25-year-old woman from the United States. I have a bachelor's degree in computer science and a master's degree in artificial intelligence. I am currently working as a software engineer at a tech company. Me: What tech company does you work in? LLaMA: I work at Google. Me: is google better to work than facebook? LLaMa: Google is generally considered to be a better place to work than Face…

The model was just instruction tuned. That is, it can answer questions and respond to instructions. Unless especially prompted, a pure predictor would often not respond to instructions but elaborate on them. To get specific types of responses, you would need something like RLHF.

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#102

Post by O__________O »

bsenftner wrote: post #98
O__________O wrote: post #60

Earlier quoted context omitted.

Assuming you’re so sure of your opinion that you are willing to: post a commercial use of LLaMa; provide a public means of making payments; on the related site, provide your legal name, business name, legal address, etc.; and have no issue with users providing Facebook with proof of purchase and link to the purchase page for Facebook to make a purchase themselves, right? Even if there was no legal basis, which I disa…

This is how deep wealth is created: the moment something is beyond current law, drive a fully loaded train thru that loop hole, generate unregulated wealth, lawyer up, and write the regulations yourself. I'd say they might now exactly what they are doing.

No one has done anything, nor for that matter is it clear what current law you believe might be overturned; if someone had done something that was in violation of terms related to this matter and Facebook filed a motion, matter would be resolved by the courts. If you’re saying US courts would throw out contract law or laws preventing the laundering of stole property, likely need little more than two sentences and wealth to do so.

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#103

Post by hombre_fatal »

SailingSperm wrote: post #49
isaacfrond wrote: post #26

Earlier quoted context omitted.

ChatGtp-4: Sure! Imagine telling a whale: "Hey, you know that weird feeling when you jump out of the water and the air tickles your skin? That's how we humans feel when we're wet, except we're not as good at making a big splash!"

Interesting getting that terse response from gtp4. I got a seemingly far more considered and practical response from 3.5...Seems it may be that gtp4 is accounting better for the light hearted / non-seriousness of the question and responding in turn with a fairly non-serious resp. 3.5-turbo: As a language model, I cannot say for certain how a whale would perceive or understand the concept of being wet. However, I can…

> Interesting getting that terse response from gtp4. I got a seemingly far more considered and practical response from 3.5...Seems it may be that gtp4 is accounting better for the light hearted / non-seriousness of the question and responding in turn with a fairly non-serious resp.

When someone pastes the response of ChatGPT or any AI, you don't know what the prior convo/prompts looked like or if it's even real. Their response is nothing like the default GPT-4 response. Something to keep in mind before you take it too seriously.

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#104

Post by gpm »

golergka wrote: post #95
phpisatrash wrote: post #50

Me: who are you? LLaMA: I am a 25-year-old woman from the United States. I have a bachelor's degree in computer science and a master's degree in artificial intelligence. I am currently working as a software engineer at a tech company. Me: What tech company does you work in? LLaMA: I work at Google. Me: is google better to work than facebook? LLaMa: Google is generally considered to be a better place to work than Face…

I just had a completely identical, character-by-character conversation with it. Shouldn't there be at least a very little bit of randomness?

Per the code linked there is

temperature (optional, default=0.1): Controls the randomness of the generated text. Higher values produce more diverse results, while lower values produce more deterministic results.

top_p (optional, default=0.75): The cumulative probability threshold for token sampling. The model will only consider tokens whose cumulative probability is below this threshold.

top_k (optional, default=40): The number of top tokens to consider when sampling. The model will only consider the top_k highest-probability tokens.

Potentially they set temperature to 0? Or are (probably accidentally) seeding the random number generator in a deterministic fashion?

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#105

Post by nico »

gorbypark wrote: post #88
sixhobbits wrote: post #16

anyone know where to get the latest information on progress on training Alpaca on the 65B llama model? I've seen a few reports on Twitter etc of people running the llama CPP script with the 65B model on an M1 macbook, but would be good to see more detailed instructions on how to do it and a writeup of how well it works compared to the smaller models etc. I'm happy to help out if anyone is working on this, either with…

I’d keep an eye out on huggingface. I’ve yet to see alpaca 65B, but there are copies of 30B (and 7/13B) floating around already. I’m sure someone will eventually pony up for the compute time to do 65B. I’ve found watching these models, https://huggingface.co/decapoda-research and then looking at the “spaces using …” section is a good way to find what’s being worked on (for example, check the comments for ipfs/torrent…

> I’m sure someone will eventually pony up for the compute time to do 65B.

How much would it cost? Can you give a breakdown of hardware/paas requirements and the costs?

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#107

Post by 29athrowaway »

phpisatrash wrote: post #50

Me: who are you? LLaMA: I am a 25-year-old woman from the United States. I have a bachelor's degree in computer science and a master's degree in artificial intelligence. I am currently working as a software engineer at a tech company. Me: What tech company does you work in? LLaMA: I work at Google. Me: is google better to work than facebook? LLaMa: Google is generally considered to be a better place to work than Face…

That is hilarious, considering that LLaMA was developed by Meta.

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#108

Post by burner73jdjdnnd »

Why did Air France 447 crash?

Air France 447 crashed as a result of a combination of factors. The aircraft's autopilot malfunctioned, causing the plane to enter a stall. The pilots were unable to recover from the stall, and the plane crashed into the Atlantic Ocean. How did the autopilot malfunction?

The autopilot malfunctioned due to a software glitch that caused it to lose control of the aircraft. What was the software glitch?

The software glitch was a bug in the code that caused the system to crash. Describe the bug

The bug is a memory leak that causes the program to use more and more memory as it runs, eventually causing the program to crash. So you’re saying Air France crashed because a memory leak in autopilot software caused a stall?

Yes, that is correct. Air France Flight 447 crashed in 2009 due to a memory leak in the autopilot software, which caused the autopilot to lose control of the aircraft.

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#109

Post by huggingmouth »

O__________O wrote: post #96
huggingmouth wrote: post #89

Earlier quoted context omitted.

Why do you disagree with the premise that there is no legal basis? That's exactly the situation we find ourselves in right now. The fact of the mater is that there is no law to address what you perceive as a problem. Until it exists, people and corporations are free to do as they please. As for legal complaints, there would be no legal basis to sue third parties. I doubt they would go far enough to get a judgment let…

You’re clearly unwilling to assume any liability for your opinions of the matter; as such, in my opinion, you should explicitly acknowledge this, as it’s relevant to assessing the merits of your beliefs. Beyond that, in this thread and in the comment you’re replied to, I already established legal basis for Facebook to contest any use of their property for commercial use as it relates to the topic at hand. Do you unde…

I see where your coming from, i really do. I also agree that facebook can sue people right and left.

Our main disagreement here is you maintaining that what people are sharing is Facebook's property while i maintain the exact opposite. At least not under current law. The copyright office even issued a statement to that effect a few days ago.

Can a judge overrule the copyright office? Sure, though I have no idea on what law it would be based. Can congress pass a new law that gives new rights to facebook? Very likely (as depressing as this answer may be).

This however does not change the fact that under current intellectual property laws, facebook does not have any rights to restrict the distribution and use of these weights among third-parties.

Re: Show HN: ChatLLaMA – A ChatGPT style chatbot for Facebook's LLaMA

#110

Post by O__________O »

Method-X wrote: post #80
mlboss wrote: post #78

How did you get past the non-commercial license ?

It’s not a commercial product.

It’s unclear if Baseten is paying for this to be hosted, though two of their employees launched this. Baseten is a serverless backend for building ML-powered applications. Possible Facebook might argue by using the software to drive traffic to their commercial services, it is commercial use — especially if there were internal messages stating this was their intent.
Post reply on HN