open AI | TheaterFire

[-]

Admirable-Star7088@reddit

I have no problem with companies wanting to close their software, they decide for themselves what they want to do with their own products. But what bothers me is the very misleading and poor choice of name. They are everything but **Open**AI. Like, wtf?

Reply

[-]

SpiteCompetitive7452@reddit

It's even worse that they exploited nonprofit status to raise capital and create the product that they now profit from. They conned donors by creating a for profit subsidiary that benefits from the product built off generosity. Those donors should be entitled to stake in the corporation that clearly fleeced them out of investor status.

Reply

[-]

AbdulMejidII@reddit

Craziest part is that I just realised those donors didn't receive anything or even a stock from "Open" AI, straight up cash cow

Reply

[-]

shutterfly2011@reddit

It’s by design to start with. Sam Altman wants all along to have this “humanity” persona while deep in his core he is just a capitalist. I have no problem him being a capitalist, what really irks me is he is just, a whore pretending to be a virgin (I don’t mean to demeaning woman or sex industry)

Reply

[-]

Admirable-Star7088@reddit

The problem is that he does not want to let the rest of us be capitalists in the LLM world. Personally, I'm a capitalist and believe strongly in private property, this is why I love open source and open LLM weights, I don't like being dependent on someone else's service (in this case, openAI), I want to own and have control over my software and LLM.

Reply

[-]

x3gxu@reddit

I barely know anything about economic systems, but isn't something "open" closer to socialism/communism and "private" to capitalism? Like you want other people's stuff to be open source for you to use privately?

Reply

[-]

ThisGonBHard@reddit

Capitalism is about free trade. Sharing stuff for free is capitalism if you are doing it voluntary. This shit is why I hate the shareholder capitalism system. It FORCES maximum greed under legal liability, in the interest of a minority of shareholders, even if 99.9% are contempt to make a boatload of money instead of ALL the money. Combine that with the governments holding up corporations that should fail, and the system starts looking less like capitalism, and more like feudalism to me.

Reply

[-]

mariofan366@reddit

Capitalism is about private property. In socialism you can have the workers own the means of production, own personal property, and freely trade their personal property.

Reply

[-]

djm07231@reddit

I still have no idea why they are not releasing GPT-3 models (the original GPT-3 with 175 billion parameters not even the 3.5 version). A lot of papers were written based on that and releasing it would help greatly in terms of reproducing results and allowing us to better compare previous baselines. It has absolutely no commercial value so why not release it as a gesture of good will? There are a lot of things, low hanging fruits, that “Open”AI could do to help open source research without hurting them financially and it greatly annoys me that they are not even bothering with a token gesture of good faith.

Reply

[-]

Admirable-Star7088@reddit

LLMs is a very new and unoptimized technology, some people take advantage of this opportunity and make loads of money out of it (like OpenAI). I think when LLMs are being more common and optimized in parallel with better hardware, it will be standard to use LLMs locally, like any other desktop software today. I think even OpenAI will (if they still exist), sooner or later, release open models.

Reply

[-]

TellJust680@reddit

probs laptops with ai capability

Reply

[-]

Innocent__Rain@reddit

Trends are going in the opposite direction, everything is moving "to the cloud". A Device like a Notebook in a modern workplace is just a tool to access your apps online. I believe it will more or less stay like this, open source models you can run locally and bigger closed source tools with subscription models online.

Reply

[-]

Admirable-Star7088@reddit

Perhaps you're right, who knows? No one can be certain about what the future holds. There have been reports about Microsoft aiming to start offloading their Copilot on consumer hardware in the near future. If this is true, then there still appears to be some degree of interest in deploying LLMs on consumer hardware.

Reply

[-]

KillingItOnReddit@reddit

The way new laptops are marketed with AI chips and the way Apple is optimizing their chips to do the same I can see it catch on for most products that use AI like that

Reply

[-]

Spiritual-Formal-744@reddit

Thanks Sam. Embarassing.

Reply

[-]

New-Drawing7845@reddit

XD

Reply

[-]

catgirl_liker@reddit

Remember when they've said gpt-2 is too dangerous to release?

Reply

[-]

Poronoun@reddit

Sam Hypeman

Reply

[-]

Zediatech@reddit

If GPT-2 is dangerous, then Wikipedia must be a WMD!

Reply

[-]

Admirable-Star7088@reddit

And the Internet would be the destruction of universe itself.

Reply

[-]

Melancholius__@reddit

apolycapse indeed

Reply

[-]

Dauntless-4ever@reddit

https://preview.redd.it/gxvhqdf11t1d1.jpeg?width=610&format=pjpg&auto=webp&s=4b6481b590c93732bf73ea21b4b7440c05efdce7

Reply

[-]

Dauntless-4ever@reddit

They are going to terk er jerbs!!!

Reply

[-]

Melancholius__@reddit

which jobs? do you look forward to being replaced? may be rote

Reply

[-]

Dauntless-4ever@reddit

All the jerbs!!! They going to take all our jerbs!!! Lol, don't mind me... Just being goofy.

Reply

[-]

Independent-Egg821@reddit

Windows ai

Reply

[-]

highheat44@reddit

Interesting

Reply

[-]

Hopeful-Site1162@reddit

Even if OpenAI stuff was the absolute best possible it wouldn’t be able to compete with the sea of open source locally available models there are. I’m really curious to see how this company will survive in the next years.

Reply

[-]

Slimxshadyx@reddit

It is absolutely competing with all the open source models out there lmao. I know this is a local and open source model subreddit but literally everyone else uses OpenAI.

Reply

[-]

deleted_by_reddit@reddit

In all honesty, llama 3 (8b) really feels pretty close to GPT-3-3.5. I am not sure about the larger model because I can't run it locally (examined it only a little bit). In fact, for my task llama 3 is superior to GPT-3.5, I know it because GPT-3.5 is actually incapable of performing it and llama 3 is. GPT-4 of course does it a bit better but it's super expensive. I don't think they will be able to hold their superiority for much longer. I talk about the instruct model.

Reply

[-]

svearige@reddit

Not expensive compred to GPUs to run the latest open source models, that won’t even be able to run them fast enough in a few new variations..

Reply

[-]

Sushrit_Lawliet@reddit

(C)ope n AI

Reply

[-]

-take@reddit

Thats actually funny

Reply

[-]

Clean-Description-23@reddit

Everyday I’m sad the board was fired and Sam Altman was not

Reply

[-]

sarveshgupta89@reddit

Why did they take it offline then?

Reply

[-]

Felipesssku@reddit

The worst thing is that the Microsoft is in it.

Reply

[-]

PaleTomorrow8446@reddit

Microsoft has done more for open source than most bigger tech companies.

Reply

[-]

tovefrakommunen@reddit

Oh and a big thanks to the underpaid africans btw.: https://www.theguardian.com/technology/2024/apr/16/techscape-ai-gadgest-humane-ai-pin-chatgpt

Reply

[-]

MysteriousPayment536@reddit

Meta and others do the same

Reply

[-]

PaleTomorrow8446@reddit

doesn't mean it is right. they are all horrible

Reply

[-]

sigiel@reddit

while activly lobbing congress to ban open source !

Reply

[-]

PaleTomorrow8446@reddit

how would they do this anyway? it is absolutely mind boggling how people in power still do not have any clue whatsoever how the digital world works.

Reply

[-]

PaleTomorrow8446@reddit

what do you expect from a genocide apologist?

Reply

[-]

I_will_delete_myself@reddit

https://preview.redd.it/ggw1a5pzhaxc1.png?width=494&format=png&auto=webp&s=989fdd10156b0a9ac854203695595a8fc2d2db29

Reply

[-]

Capt-Kowalski@reddit

This is partially incorrect. Pretraining is done using low quality internet content, but it the easy part as after pretraining network is of little use. Their power comes from taming, or fine tuning as they call it, and that is a process that requires a lot of manual work to put together a specialised training dataset and tune the network using it. Without it the network, for example, would not be able to operate in an assistant mode, or do anything remotely useful.

Reply

[-]

Entire-Plane2795@reddit

The key to pre training is diversity and scale. Which came from (most likely) a lot of copyrighted material/material that was never intended or authorised for such use. True, there is added value in the curation of fine-tuning data (also questionable in origin in the case of OpenAI). But I mean it's like comparing the materials that buildings are built ok to the land on which they're built. And arguably, training a foundation model on copyrighted material is a bit like building on other people's land...

Reply

[-]

sebramirez4@reddit

What I don't like is how much he talks about the importance of AI governance, how much he talks about the importance of doing things in the open and the open source community, and then hasn't even open-sourced just normal bad GPT-3, it bothers me so much how much his talk doesn't match his actions if a company called closedAI made all of these decisions and sam altman didn't talk about those things so much I'd actually really like the company.

Reply

Reply to Post

47 Comments