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

The genie escapes: Stanford copies the ChatGPT AI for less than $600

newatlas.com

Post reply on HN
Print view View original

151–160 of 173 posts

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#151

Post by ComposedPattern »

doctor_eval wrote: post #133
cjohnson318 wrote: post #90

> It seems these godlike AIs are already frighteningly cheap and easy to replicate. "godlike"? Really? I'm not religious, but this seems like an overreaction for something that has no agency.

How do you know agency is not simply the output of a large language model encoded in neurons? What is the difference between neuronal and digital weights? Considering that we don’t know how the brain works so well, and we don’t understand why LLMs work so well, simply on the basis of their output I think the safest assumption is that these models do indeed have agency, or at least the capability of agency.

Gods don't have persistent bodies or brains. They're spirits.

I don't disagree that digital systems with neural architectures could have agency in principle, but agency generally is definitely not the output of a large language model. Animals without language have agency, in that that they take actions to fulfill their desires. Current LLMs may have some degree of intelligence, but they don't even appear to have any consistent wishes or desires. You can get them to talk longingly about x... until you give another prompt and suddenly x doesn't matter to them at all.

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#152

Post by brownfox »

satvikpendem wrote: post #125
f_devd wrote: post #76

Earlier quoted context omitted.

LLama hasn't been fine-tuned with RLHF, so it requires additional prompting, check out the open-assistant[0] project for an open-source ChatGPT equivalent (WIP). [0]: https://github.com/LAION-AI/Open-Assistant

Use it here: https://huggingface.co/spaces/olivierdehaene/chat-llm-stream...

It looks stupid compared to ChatGPT.

Prompts:

* Tell me about OpenAI.

* Make it shorter, within 50 words.

* Rephrase it, like a rap singer.

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#154

Post by f_devd »

brownfox wrote: post #152
satvikpendem wrote: post #125

Earlier quoted context omitted.

Use it here: https://huggingface.co/spaces/olivierdehaene/chat-llm-stream...

It looks stupid compared to ChatGPT. Prompts: * Tell me about OpenAI. * Make it shorter, within 50 words. * Rephrase it, like a rap singer.

> What does OpenAI do? OpenAI is a company founded by Elon Musk and Reid Hoffman in 2015 to develop artificial intelligence products for the general public. They are most well known for releasing a large language model, GPT-3, in 2017, which was developed using a method called deep learning.

OpenAI focuses on creating tools and technologies that allow developers to create AI applications that are accessible to the general public. They are particularly interested in developing tools that can help developers build AI applications that can perform tasks that are important to society, such as in the areas of healthcare, education, and safety.

In addition to GPT-3, OpenAI has also released other AI tools and technologies, including an open-source platform for training AI models, called OpenAI Research, and a library of pre-trained AI models for use in various applications.

Overall, OpenAI’s goal is to make AI technology more accessible and useful to the general public, and they are dedicated to creating products that can help create a more sustainable and inclusive future for society.

> How are they funded?

OpenAI is a not-for-profit organization, so it does not rely on any financial support from individuals, corporations, or governments. Instead, it is funded through a variety of sources, including venture capital firms, research institutions, and individual donations.

Some of the largest financial supporters of OpenAI include Microsoft, Amazon, and Google, who have invested millions of dollars into the organization. Additionally, OpenAI has received financial support from a number of other prominent technology companies and research institutions.

Overall, the funding sources for OpenAI reflect the organization’s goal of creating innovative AI tools and technologies that can have a positive impact on society.

> Who are OpenAI’s competitors?

There are many AI competitors to OpenAI, including companies such as Microsoft, Amazon, Google, Facebook, and Apple. These companies are all investing heavily in developing their own AI products and technologies, and they all have their own goals and objectives in the field.

Some of OpenAI’s competitors are specifically focused on creating AI tools and technologies for the general public, while others are focused on creating AI solutions for specific industries or businesses. For example, Microsoft is known for its popular Microsoft Office products, while Amazon is known for its e-commerce platform.

It’s worth noting that the development of AI technology is a rapidly evolving field, and there will likely be many new competitors and advancements in the years to come.

----

Seems pretty decent to me, sure it's not on the level of ChatGPT but it's also 1/14th the size of the original GPT-3.

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#155

Post by starik36 »

dragonwriter wrote: post #15
starik36 wrote: post #9

From the article: Pre-trained on a trillion "tokens"... Doesn't 7B indicates that it was trained on 7 billion tokens? Or am I misunderstanding the nomenclature?

> Doesn't 7B indicates that it was trained on 7 billion tokens? No, 7B means it has 7 billion parameters.

And what does a parameter mean in this context?

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#156

Post by dragonwriter »

starik36 wrote: post #155
dragonwriter wrote: post #15

Earlier quoted context omitted.

> Doesn't 7B indicates that it was trained on 7 billion tokens? No, 7B means it has 7 billion parameters.

And what does a parameter mean in this context?

A parameter is the connection weight between two neurons in the network.

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#157

Post by starik36 »

gaogao wrote: post #3

Has anyone tried this yet on the 65B version? I'm curious if it knows how to rhyme and other emergent behavior, as alpace-7B does not.

I tried because I was curious. It's not very good. First I "quantized" the files. Took about 3-4 hours. Then I ran a prompt: "create a short poem about my wife helping me get over the stuffed nose". After 3 hours of 99% CPU usage and using all 32 GB of RAM, I got the following result:

create a short poem about my wife helping me get over the stuffed nose and throat that i've had for days now. the cold was all around him, but he didnΓÇÖt notice it at first- so much to do in this new place! finally though when they were sitting by themselves on their porch swing his sweetheart leaned into her man saying ΓÇ£we could take some of these leaves from the garden and make tea.ΓÇ¥ He looked up with a quizzical look and saw that she was serious. I've been under weather since last Tuesday. Today is day 6...and still going strong! Had to cancel two nights on stage, missed my son

https://i.imgur.com/Nl3xLEg.png

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#158

Post by disgruntledphd2 »

homarp wrote: post #96
braingenious wrote: post #79

Earlier quoted context omitted.

I’m also a bit confused by the quantization thing. Why exactly is everybody running the same program on the same file? Why not just include the quantized weights? It seems like if somebody figured out the “correct” way to quantize the 7b weights it would make way more sense to just torrent the output rather than distribute a fixed program.

quantization takes lots of RAM: https://github.com/qwopqwop200/GPTQ-for-LLaMa/blob/main/READ... says llama-13B takes 42GB and 33B takes more than 64GB...

I quantised all of the models(though the 65Bn one came out weird) and I only have 64Gb of RAM, so dunno.

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#159

Post by doctor_eval »

cjohnson318 wrote: post #150
doctor_eval wrote: post #149

Earlier quoted context omitted.

I think the precautionary principle should apply to AI. Assuming and acting as if they have agency, is safer than assuming they don’t. > it should be clear, or demonstrable, that the model (1) has goals (2) takes concrete steps to achieve those goals. That definition seems arbitrary to me; many humans wouldn’t pass this test. On the other hand, LLMs certainly seem capable of acting towards specific goals (such as hel…

> That definition seems arbitrary to me; many humans wouldn’t pass this test. What part of this is arbitrary, i.e. random, whimsical, or biased? This is a fairly comprehensive working definition of "agency". "Internally generated", which is implied, is a nice touch. Nearly every human over the age of 18 months passes this test with flying colors. Toddlers have goals, and do everything in their power to achieve them.…

Got it. So you assume these systems are inherently safe.

What could go wrong?

Re: The genie escapes: Stanford copies the ChatGPT AI for less than $600

#160

Post by Izkata »

SllX wrote: post #72
Larrikin wrote: post #54

Earlier quoted context omitted.

Highly recommend Animorphs. As an adult the books can be read through pretty quickly, but the story really holds up. Deals with some heavy topics of genocide, enslavement (even exploring the justifications of the enslavers), PTSD among other topics but was still considered a kids book. It becomes obvious in the middle when some of the books were written by ghost writers, but the books are so easy to read I don't real…

> If you must you could probably get away with reading the first ten, last ten Not sure that you can. If you were to skip any, probably only 31 through 39 are completely skippable, maybe some of the late 20s but I would still read 29 and 30 at a minimum. Some of the teens and 20s might be skippable after 13 but there’s a fair amount of world-building outside the Chronicles series in the 20s; and 40 onwards is setting…

Don't forget 20, 21, and 22 are the David arc.
Post reply on HN