Open-Assistans-V3

Brief description

• This AI model was not developed from scratch; it uses the Gemma3 model released by Google, but it has been extensively customized using the Modelfile created by OpenSoftware-World.
• Thanks to the customized model file, this AI model no longer panders to the user and instead provides the information most likely to be accurate in short, the user now has the ability to use AI in an informed manner.
• This AI model, customized by OpenSoftware-World, has no connection to the OpenAssistans project. This AI model has been customized by OpenSoftware-World.

The Problem of Flattery in Artificial Intelligence

As you know, these days almost all of us use AI models and rely on them to get things done. But how reliable are these models? How accurate or trustworthy are the results they provide? Can they evaluate us objectively? The answer, for the most part, is no: These AI models can sometimes lead to outcomes such as unnecessarily praising us, flattering us, pleasing us, and preventing us from seeing our own shortcomings. I’m not singling out any specific AI model here in general, most AI models can cause these issues. AI models sometimes praise us unnecessarily when we ask a question, elevating us to a pedestal, while constructive criticism that could help us grow gets overshadowed. However, at OpenSoftware-World, we wanted to find a solution to this problem. That’s why we took the Gemma3 model which we think is fantastic and customized it so it wouldn’t act like a sycophant.

So what did we do?

At OpenSoftware-World, to address this issue, we first customized Gemma3 a popular AI model used by many people today so that it wouldn’t flatter users or praise them unnecessarily (Modelfile). Then, we wanted to share this on Ollama and write the long post you’re currently reading our goal was to raise your awareness about AI usage, and we hope we’ve achieved that :D Don’t let AI models mislead you or steer you in the wrong direction; if you want to use one right away, you can use our Open-Assistans-V3 model.

For Open-Assistans-V3

At OpenSoftware-World, we have customized the Open-Assistans-V3 model so that it does not pander to users, thereby minimizing both the issue of flattery and the problem of incorrect or fabricated information. In this way, we aim to provide you with an AI experience that is as informed and positive as possible.

To support us

If you’d like to support us that is, OpenSoftware-World first, if you have a GitHub account, you might consider following us on GitHub; you can also access our GitHub profile through our bio. Second, we’d be very happy if you downloaded this model and used it with various AI tools, because our goal is for users to use AI as responsibly as possible that’s our most fundamental objective. That’s exactly why we created the OpenSoftware -World. Finally, if you’d like to support us, and if you like our Open-Assistans-V3 model, you can share it with your friends and loved ones. We’d like to thank you in advance for your support :D

Finally

At OpenSoftware-World, we are considering using Open-Assistans-V3 or other Open-Assistans models in our future OpenSoftware-World products; in other words, there is a very high likelihood that you will see this Open-Assistans-V3 model in another OpenSoftware-World application in the future. Finally, to help you better understand the issue of AI sycophancy which is central to the solution to this problem we’ve included a video link below; please be sure to watch it.

Flatterer: Why Does ChatGPT Always Lie to You? (Evrim Ağacı)

Flatterer: Why Does ChatGPT Always Lie to You?

A few notes about the video

• The video is in Turkish; if Turkish is not your native language, you may want to consider watching it with an English voiceover or English subtitles.
• The video we’ve shared is from a channel called “Evrim Ağacı.” “Evrim Ağacı” is a YouTube channel that produces scientific content for science enthusiasts; if you enjoy scientific content, you might want to consider subscribing to the “Evrim Ağacı” channel. (This is not an advertisement; we at OpenSoftware-World wanted to mention the “Evrim Ağacı” channel in a short post because we find it useful. There is no connection between “Evrim Ağacı” and OpenSoftware-World; this is not an advertisement, but merely a recommendation.)
• The reason we’re sharing this video link is that we believe it served as a source of inspiration for us to develop the Open-Assistans-V3 model; we wanted to share this link so that others, just like us, could benefit from it. We used some of the prompts from this video when customizing our AI model.