Went to Unite 2023 in Amsterdam.
It was okay, but just okay. In the future, I’ll choose to go to GDC or devcom every time over Unite.
In the past, the talks (that I’ve seen on YouTube) have useful for knowing how to use their tools. This Unite was more about selling us on their new tools. In one interview, the Unity person felt he had to preface every question with comments about how great Unity was.
Suggestion for future Unite (if Unity people read this): Almost every conference in the universe has a visual schedule with all the room presentations and their times, so it’s really clear what talks overlap with what. It allows attendees to plan. And even let’s you see that you’ve scheduled multiple talks on the same track at the same time. (e.g. At the end of the day, two Monetisation talks were at the same time, while programmingProgramming talks had dried up).
#Some thoughts on Unite 2023
1 messages · Page 1 of 1 (latest)
I WANT F# SUPPORT SO I DONT HAVE TO USE A DDL
but yes, theyre really leaning on AI tools just for the sake of leaning into buzzwords and thats it
no new actual useful tools got announced en masse
damn using a ddl you a savage
I would check Muse out before just writing it off as a "buzzword and thats it"...
Muse is more expensive than any other tool on the market, they ask for $30/month, which gives you credits that you can go through really quick. Then you must pay for additional credits or what ever.
And the output is of lower quality than you get anywhere else, even free local LLMs.
The chat is just fine tuned ChatGPT 3.5
which is free to use
and any image generator outputs better results than Muse texture or sprite gen
difference being valve is allowing muse content
due to unity only using training data they have the rights to
Unity are saying that but we don't know where the data is sourced from, their own blog post mentioned the usage of Stable Diffusion at least in some capacity.
But that's all moot now, Valve are allowing all use of AI now as long as you disclose what it was used for:
Back in June, we shared that while our goal continues to be shipping as many games as possible on Steam, we needed some time to learn about the fast-moving and legally murky space of AI technology, especially given Steam's worldwide reach. Today, after spending the last few months learning more about this space and talking with game developers, ...
Yeah, muse is just them trying to cash in on the AI train and essentially rebrand and sell everyone else's free open source work. In my humble opinion of course.
I figure, given their recent track record, they will begin removing and rejecting AI tools from their asset store to try and force people to muse.
Or making it a mandatory purchase for Industry users, kind of like their Mars and other projects that flopped and needed user numbers to justify keeping around.
Pretty sure they trained it in an "Ethical way". I read once that if you ever have issue on content that has been generated with muse they would help you defend your case.
It is certainly possible, but training an image generator core from ground up takes a lot of hardware, time, money, and art resources. Ignoring all the first ones, it is hard to see where Unity would have access to so much ethically sourced art.
It is far to doubt their statement, that being said they are still ready to help you (legally) if you ever get in trouble because of their AI tool which in my opinion means that they do have a considerable amount of confidence. I am unable to find the statement though.
forget muse for now, its code complete isn't polished. but Sentis is actually good
It's more expensive because you're paying for an LLM that is (according to Unity's word) not trained on anything that Unity doesn't own the rights to. Does that make it lower quality than the unethical options that Steam WILL remove for usage of? Of course. I don't use it, but I completely understand why it costs at least $10 more per month than the stuff that generates content Steam would remove for.
You are right that we have to take their word for it, but I mean, it's a $12.5B company that would probably never recover from being caught lying here. I for one don't have any worry that they would be lying.
more people using it and providing feedback would speed up this polishing, but Sentis seems great yes
they lack direction, no graph, no back propagation unless you hack your model
and their forum is hidden away from the main one
many obstacles for users
Unity Muse chat is just a fine tuned ChatGPT 3.5, which is trained on data Unity don't own. Their own blog post mentioned using Stable Diffussion at least in some capacity. The story about their tools not being trained on anything that they don't own is all smoke and mirrors, what it actually means is that they fine tuned existing models on data they own but the base models are the regular LLM faire. Unity don't have half a billion and a library of millions of images to train an LLM from scratch, they're bleeding money as it is.
You could get ChatGPT 4 and use Stable Diffusion locally for better and cheaper results and basically same foundational training data.
They do mention Stable Diffusion, yes, but there's more to it:
"Recently, Stable Diffusion has been the subject of ethical concerns because the model was originally trained on data scraped from the internet. We limited our reliance on pretrained models as we built Muse’s Texture and Sprite capabilities by training a latent diffusion model architecture from scratch, on original datasets that Unity owns and has responsibly curated. By using the Stable Diffusion model minimally as part of our data augmentation techniques, we were able to safely leverage this model to expand our original library of Unity-owned assets into a robust and diverse repository of outputs that are unique, original, and do not contain any copyrighted artistic styles. We also applied additional mitigations on top of this that we’ll describe below. Our training datasets for the latent diffusion models underpinning Muse’s Texture and Sprite capabilities do not comprise any data scraped from the internet...
...After augmenting our existing data, there were still gaps in a range of subjects that we needed to fill. To do this, we trained Stable Diffusion on our own content until its behavior was significantly changed. Using these derivative models, we created entirely new synthetic data using a prefiltered list of subjects. The list of subjects went through both human review and additional automated filtering using a large language model (LLM) to ensure we did not attempt to create any synthetic images that would violate our guiding principles and go against what we were trying to achieve: a dataset completely devoid of recognizable artistic styles, copyrighted materials, and potentially harmful content."
https://blog.unity.com/engine-platform/responsible-ai-and-enhanced-model-training
Using GPT 4 and Stable Diffusion is doable but not advisable for a game you want to put on steam, or make $ off of
It's just PR. And Steam now treats all AI tools the same. Doesn't matter if it's Muse chat or ChatGPT4: https://steamcommunity.com/groups/steamworks/announcements/detail/3862463747997849619
So there is not actual benefit if using Muse over anything else.
Back in June, we shared that while our goal continues to be shipping as many games as possible on Steam, we needed some time to learn about the fast-moving and legally murky space of AI technology, especially given Steam's worldwide reach. Today, after spending the last few months learning more about this space and talking with game developers, ...
i.e. ownership of the dataset does not matter to Steam anymore. And Muse didn't align with the previous terms, anyway, since you, the game's developer don't own right to the Muse dataset, which is a black box.
Unity's claims to defend anyone in court is just PR. No one's helping some random indie. They made that claim knowing that individuals are unlikely getting sued, everyone's going after the LLM companies because that's where the money is.
"Live-Generated: Any kind of content created with the help of AI tools while the game is running. In addition to following the same rules as Pre-Generated AI content, this comes with an additional requirement: in the Content Survey, you'll need to tell us what kind of guardrails you're putting on your AI to ensure it's not generating illegal content."
Edit: apparently I forgot to add the first half:
Pre-Generated: Any kind of content (art/code/sound/etc) created with the help of AI tools during development. Under the Steam Distribution Agreement, you promise Valve that your game will not include illegal or infringing content, and that your game will be consistent with your marketing materials. In our pre-release review, we will evaluate the output of AI generated content in your game the same way we evaluate all non-AI content - including a check that your game meets those promises.
Muse = the guardrails
People have literally produced Mario with Muse.
it's hardly guarded.
and you don't live generate textures or sprites in 99% of use cases.
"how can you guarantee your Ai generated art is ethical and legal"
"I used Muse"
"Understandable, have a nice day"
This is the basic concept of it. Steam is not treating all AI tools the same, they've simply announced that they will accept art from the ethical tools (such as Muse)
as for Mario, I mean, I am sure you could use it to generate mario easily, yeah
but nintendo would definitely be on your ass if you tried to use that
Trained ethically doesn't mean it can't be used to create something that purposefully resembles existing IP
That won't fly with Valve, doesn't matter how it was trained if it outputs copyrighted material.
So Muse itself is not the guardrail... Valve also won't greenlight some LLMs over others, it's impossible to validate that on their end.
don't look at it as Muse being greenlit, look at it as all the others not being allowed
Muse doesn't get special treatment, it just follows the requirements
or at least, Unity is claiming as such
Muse is not special just because Unity wrote some PR blog post where they promise to use Stable Difffusion minimally but don't actually prove it.
You are definitely correct that if Unity is lying here, it will be bad for them
It's all on their servers, no one can prove it either way, which is my point. Muse is not special just because Unity says so.
Muse isn't special just because Unity says so, it's special because Steam also agrees
They have not revealed their dataset, its sources or in what capacity their work is based on existing models.
And muse chat is literally just ChatGPT 3.5
with a bit of Unity data on top
Definitely, I totally understand being skeptical. I was merely explaining how Muse relates to Steam's guidelines in comparison to unethical AI, in the event that Unity is 100% truthful here.