Elon Musk Sues OpenAI??

Recorded live on twitch, GET IN

Become a backend engineer. Its my favorite site

This is also the best way to support me is to support yourself becoming a better backend engineer.

MY MAIN YT CHANNEL: Has well edited engineering videos

Discord

Hey I am sponsored by Turso, an edge database. I think they are pretty neet. Give them a try for free and if you want you can get a decent amount off (the free tier is the best (better than planetscale or any other))

[SimGunther]

In other news, sun is hot and companies bragging about being "open" do not open source. Who else saw this coming?

[kuhluhOG]

Here are two things some people may not know:
- Elon Musk was in the first few years by far its biggest donor.
- OpenAI Inc is a 501(c)(3) Non-Profit, meaning a Public Charity; this also means that the founding principles are actually legally binding.

[mjsagxy]

Can suing change anything? Yes - one it can slow development to avoid the appearance of anti-trust. 2. It can drag things into the sunlight via the discovery process. 3. on occasion and many years later, it can lead to money payments.

[sporkionsuz-beero]

Hot take: If AI wants to train on artists to imitate their art, then artists themselves should be given a stake in the company by default so that if they cannot make money for their art due to ai generated imitation, they can be compensated by company's valuation.

[Iswimandrun]

The AI if your a Junior Dev takes you right up to the cliff of I don't know where I am anymore.

[Tsurael]

13:30 A couple of unspoken assumptions for that argument to hold:
1. AI (LLMs or whatever the next paradigm is) will NEVER be able to do better than the derivative art they have now. I personally wouldn't bet on this premise holding for even the next 5 years
2. That these companies legally acquired and had rights to use the works used in their training data. OpenAI and Midjourney have both been shown to have used paywalled content and even reproducing images with the Getty watermark.
3. That a huge chunk the market for art (especially in the graphic design space for businesses) won't just race to the middle/bottom in quality to cut out human artists and save on cost.
4. For the "real artists", they can't just fine tune a model to their style. In almost all cases it never needs to be perfect, *just significantly more cost effective* to put them out of the job.

Also there's plenty of precedent in law for treating the same action by humans and non-human differently. The US copyright office and several judges have already ruled that AI art cannot be copyrighted.

[yuriandrigani]

The thing is, nobody start with "really good at what they do".
What LLM does to them will stunted the whole lot of beginers. Causing even fewer people become experts in the future.

[Kane]

New dev title mentioned let’s go Senior Plus Plus

[abyzzwalker]

Future Developers requirements:
- not dependant on a LLM is a plus.

[deluxe_]

You sound like you could do a killer spongebob impression.

[sbrjmiy]

This reminds me of this quote from Gavin Belson: “I don't know about you people, but I don't want to live in a world where someone else makes the world a better place better than we do.”

[pcrizz]

@9:49 Chat "Jippity"? Well hot dog! Imma wrastle up some Juniper berries and match that with a Dippity Douglas on the 2-niner.

Sorry I just had a stroke, what did I miss?

[sam_mitschiro]

Ai for junior devs is copy paste on steroids.

[BugMug]

On linkdin i saw a manager in openai writing creating agi for human

[keyboard_g]

This case is for these headlines and thats about it.

[dunebuggy]

Oh no, not the corporation are people argument. AI is not a person. AI must pay; AI is not inspired; AI must pay!

[rrraewr]

12:10 The fundamental problem with AI art is, that those generative AIs are trained on protected artworks without explicit consent and are basically just pirate software on steroids.

[StevenOfWheel]

14:43 "What's different between [a human which was 'inspired' by or actively plagiarizing someone else's artwork] and an LLM?"
Well, the difference is that an LLM can spew several orders of magnitude more "art" per unit time than any human has or will ever be able to, hence multiplying the risk of plagiarism in "art" by, well, several orders of magnitude. That is definitely not a minor difference, which people tend to forget about. In the time it would take me to "get inspired by" or plagiarize any famous artist's work and cut even the slightest share of that specific art sector for myself, an LLM might have already created a vast collection of art pieces at the disposal of a bad actor.

While I do agree that there is no inherent philosophical difference between people or LLMs "getting inspired by" or plagiarizing other people's artwork, I do believe there is a huge practical difference, as it's a matter of scale and perspective as I've tried to say so far.

[lppoqql]

If only a few control AI, eventually AI will control the few and destroy all.

[Exilum]

13:30 Something that's ignored here is that the limit for AI isn't the average of its training data. You can give it negative training data as well, especially when it comes to diffusion models, and this significantly improves its floor and ceiling in terms of quality.
This is why there is so much negative prompting among people who do good with these models. When they negative prompt "amateur", they are essentially relying upon the model having learnt what makes an image look amateurish, and going the opposite way. So yes, if you make a random prompt, diffusion models aren't there, but quality prompts can be surprisingly good nowadays. You think you can tell what's AI-generated or not because you remember what you saw that was clearly AI-generated.
The issue is twofold: a sampling issue and survivorship bias. I don't need to expand on survivorship bias, so I'll talk about sampling.
You do not spend time in spaces you could encounter AI-generated images in. Someone with the level of skill and attention to detail to make a very good image would be bragging in those spaces, not in random unrelated spaces. The kind of people who would try to pass something off as AI-generated aren't the kind of people who would have the patience to spend several hours refining a prompt. It's not dissimilar to how a non-programmer could refer to code copilot generated out of a single comment in an empty file as idiomatic copilot code any real developer would consider usable.