# Interesting research paper: “Rethinking open source generative AI: open-washing and the EU AI Act”

**URL:** <https://discuss.opensource.org/t/interesting-research-paper-rethinking-open-source-generative-ai-open-washing-and-the-eu-ai-act/429>\
**Category:** Open Source AI\
**Created:** [June 28, 2024, 12:21pm UTC](https://discuss.opensource.org/t/interesting-research-paper-rethinking-open-source-generative-ai-open-washing-and-the-eu-ai-act/429 "2024-06-28T12:21:47Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![hook](https://yyz1.discourse-cdn.com/flex011/user_avatar/discuss.opensource.org/hook/32/872_2.png) [@hook](https://discuss.opensource.org/u/hook)\
**Post date:** [June 28, 2024, 12:21pm UTC](https://discuss.opensource.org/t/interesting-research-paper-rethinking-open-source-generative-ai-open-washing-and-the-eu-ai-act/429/1 "2024-06-28T12:21:47Z")

</div>

I was recently forwarded this LinkedIn post, which I found interesting:

> **[Etienne Grass on LinkedIn: #opensource #genai #aiact #llm | 27 comments](https://www.linkedin.com/feed/update/urn:li:activity:7211620451620392960/)**
>
> A large number of people support #OpenSource in #GenAI. But rare are the one really able to define it.  
> 
> The recent EU #AIAct 🇪🇺 entails a nice provision… | 27 comments on LinkedIn

It summarises the paper _“Rethinking open source generative AI:  
open-washing and the EU AI Act”_ (which, is clearly, even more interesting):  
[https://dl.acm.org/doi/pdf/10.1145/3630106.3659005](https://dl.acm.org/doi/pdf/10.1145/3630106.3659005)

> **Abstract**
> 
> The past year has seen a steep rise in generative AI systems that  
> claim to be open. But how open are they really? The question of  
> what counts as open source in generative AI is poised to take on  
> particular importance in light of the upcoming EU AI Act that reg-  
> ulates open source systems differently, creating an urgent need  
> for practical openness assessment. Here we use an evidence-based  
> framework that distinguishes 14 dimensions of openness, from  
> training datasets to scientific and technical documentation and  
> from licensing to access methods. Surveying over 45 generative AI  
> systems (both text and text-to-image), we find that while the term  
> open source is widely used, many models are ‘open weight’ at best  
> and many providers seek to evade scientific, legal and regulatory  
> scrutiny by withholding information on training and fine-tuning  
> data. We argue that openness in generative AI is necessarily com-  
> posite (consisting of multiple elements) and gradient (coming in  
> degrees), and point out the risk of relying on single features like  
> access or licensing to declare models open or not. Evidence-based  
> openness assessment can help foster a generative AI landscape in  
> which models can be effectively regulated, model providers can be  
> held accountable, scientists can scrutinise generative AI, and end  
> users can make informed decisions

It does mention OSI-AI as well:

> **1.2 The moving target of open source AI**
> 
> Until recently, classifications of software as open source could sim-  
> ply rely on the availability of source code under appropriate li-  
> censing: if some software is released under a licence approved by  
> the Open Source Initiative (OSI), it means that it is fully open and  
> minimally restrictive [45]. For software that is relatively portable  
> and user-deployable, this was long sufficient, and it afforded users  
> the rights to make copies, to tinker and, and to make improvements.  
> However, the rise of large language models and text-to-image gen-  
> erators means that a different approach is needed [47].
> 
> There are various efforts underway to update and tailor the  
> definition of open source to current generative AI systems. One is  
> a public consultation dubbed the “Open Source AI Deep Dive” that  
> the OSI board may draw on in their efforts to update their definition  
> of open source in the age of generative AI. Another is the “Joint  
> Statement on AI Safety and Openness” by parties including Creative  
> Commons, Mozilla, LAION and Open Future. The challenge that  
> these efforts face is to adopt the notion of open source, which used  
> to be fairly unambiguous, to the increasingly complex world of  
> generative AI systems [34].

And the interesting pictures …

 ![Figure 1: Comparison of BloomZ and Llama 2 on 14 dimensions of openness, illustrating the framework.](https://canada1.discourse-cdn.com/flex011/uploads/opensource2/original/1X/494cf85bdf4952b288ef3e66096f55bfbd968e70.png)

 ![Figure 2: Openness of 40 text generators described as open, with OpenAI’s ChatGPT (bottom) as closed reference point. Every cell records a three-level openness judgement (✓ open, ∼ partial or ✗ closed). The table is sorted by cumulative openness, where ✓ is 1, ∼ is 0.5 and ✗ is 0 points. RL may refer to RLHF or other forms of fine-tuning aimed at fostering instruction-following behaviour. For the latest updates see: https://opening-up-chatgpt.github.io](https://canada1.discourse-cdn.com/flex011/uploads/opensource2/original/1X/bae3486547cd398e3f9a1665ee5ce33c217028bb.jpeg)

 ![Figure 3: Overview of 6 text-to-image systems described as open, with OpenAI’s DALL-E as a reference point. Every cell records a three-level openness judgement (✓ open, ∼ partial or ✗ closed). The table is sorted by cumulative openness, where ✓ is 1, ∼ is 0.5 and ✗ is 0 points.](https://canada1.discourse-cdn.com/flex011/uploads/opensource2/original/1X/5b16d86aa8c05f1eb9bbcfdd93ae2447f5d5657d.png)

---

<div class="post-metadata">

**Author:** ![zack](https://avatars.discourse-cdn.com/v4/letter/z/f6c823/32.png) [@zack](https://discuss.opensource.org/u/zack)\
**Post date:** [July 1, 2024, 12:52pm UTC](https://discuss.opensource.org/t/interesting-research-paper-rethinking-open-source-generative-ai-open-washing-and-the-eu-ai-act/429/2 "2024-07-01T12:52:09Z")

</div>

Hello @hook, we have discussed this paper with the authors in a separate thread already, starting at [Open Source AI needs to require data to be viable - #13 by Mark](https://discuss.opensource.org/t/open-source-ai-needs-to-require-data-to-be-viable/351/13)

---

<div class="post-metadata">

**Author:** ![hook](https://yyz1.discourse-cdn.com/flex011/user_avatar/discuss.opensource.org/hook/32/872_2.png) [@hook](https://discuss.opensource.org/u/hook)\
**Post date:** [July 1, 2024, 1:11pm UTC](https://discuss.opensource.org/t/interesting-research-paper-rethinking-open-source-generative-ai-open-washing-and-the-eu-ai-act/429/3 "2024-07-01T13:11:54Z")

</div>

Whoops, missed that one, I apologise.

---

<div class="post-metadata">

**Author:** ![nick](https://yyz1.discourse-cdn.com/flex011/user_avatar/discuss.opensource.org/nick/32/542_2.png) [@nick](https://discuss.opensource.org/u/nick)\
**Post date:** [July 1, 2024, 3:21pm UTC](https://discuss.opensource.org/t/interesting-research-paper-rethinking-open-source-generative-ai-open-washing-and-the-eu-ai-act/429/4 "2024-07-01T15:21:51Z")

</div>

No worries, @hook. Thank you for sharing the images.
