If you’re a programmer, then you probably already know the answer to the question, is ChatGPT open source? The long and short is that no, ChatGPT itself is not open source at this point. But similar LLMs are popping up as open source projects, and ChatGPT may go Open Source in the future.
Read on to find out where you can start dabbling in the hopes of creating something amazing.
What Is Open Source Software, Exactly?
Open source software is basically code that is made available freely to anyone in the public who wants to modify and redistribute it. So, you don’t necessarily have to be a programmer to use or even play around with open source code.
The Linux operating system is a classic example of open source software.
The open source model allows for collaboration and software that works better for many users. Plus, being a decentralized model that relies on community, it’s very often free or much cheaper than dedicated software that does the same job.
In addition, open source software is far more flexible, with communities coming together to solve any problems in different ways.
The open source model allows for collaboration and software that works better for many users. Plus, being a decentralized model that relies on community, it’s very often free or much cheaper than dedicated software that does the same job.
The majority of open source projects are hosted on sites like GitHub, where multiple open source “versions” of ChatGPT has been making waves lately.
Keeping up with the AI agenda, there’s also Hugging Face, which started as a company that built a chat app for teens, but has now transformed into an open source AI community hub that has lets users discover, build, train, collaborate, and deploy machine learning natural language programming models. Hugging Face has published projects like Stable Diffusion, which we have talked about in depth.
- Stable Diffusion Negative Prompts — What Are They?
- Stable Diffusion Upscale – How To Improve Image Quality
Is ChatGPT Open Source?
The basic answer is, no, ChatGPT itself isn’t open source, at least not right now. This is because its parent company, OpenAI, which programmed the large language models that runs the AI chatbot — GPT-3.5 and GPT-4 — published them as closed source.
OpenAI started as an open source project — hence the “open” in the name. But due to a variety of factors, it switched from an open source project to a closed source project. Here’s the story:
OpenAI: From Open to Closed Source
OpenAI, initially an organization with a strong emphasis on open-source principles (founded in part by Elon Musk due to concerns over proprietary AI models), had a transformative journey with the development and release of GPT models, particularly ChatGPT.
This shift from open-source to closed-source was a strategic move, triggered by a multitude of factors.
Phase 1: Open Source
The initial versions of the GPT models, GPT-1 and GPT-2, were released as open-source projects. OpenAI’s philosophy was grounded in the belief that AI should be accessible and beneficial for all of humanity, and open-sourcing their models was a way to democratize access to artificial intelligence. With the release of GPT-1 in June 2018, OpenAI shared the full model, code, and dataset, enabling researchers and developers worldwide to build upon and innovate with this technology.
Phase 2: Closed Source
The release of GPT-2 in 2019 marked a turning point. Despite its substantial improvements over GPT-1, OpenAI initially chose to not release the full GPT-2 model due to concerns about potential misuse. The power of the GPT-2 model to generate high-quality, realistic text raised worries about the creation of synthetic misinformation or “deepfake” text. This shift showed OpenAI’s increasing focus on the ethical implications and societal impacts of AI technology.
Over the following months, OpenAI staged the release of increasingly larger versions of GPT-2, while monitoring the impacts and potential misuse. It wasn’t until November 2019 that they fully released the GPT-2 model, after finding no strong evidence of misuse.
However, this event set a precedent for the subsequent handling of GPT-3 and ChatGPT.
With the release of GPT-3 and ChatGPT, OpenAI made a more pronounced shift towards closed-source. GPT-3, an even more powerful model, was not released in the same way as its predecessors due to the same concerns around misuse. Instead, OpenAI chose to provide access to GPT-3 and ChatGPT through an API, which allowed them to maintain more control over how the model was used.
This decision was met with some controversy in the AI research community, with critics arguing that the move away from open-source principles was contrary to OpenAI’s mission of ensuring that artificial general intelligence benefits all of humanity (including from funder Elon Musk). In fact, Musk stated that OpenAI was “not what I intended at all,” calling it a “closed source, maximum-profit company effectively controlled by Microsoft.”
However, OpenAI maintained that this was a necessary step to ensure the responsible use of increasingly powerful AI models.
Despite these changes, OpenAI has continued to contribute to the AI research field in substantial ways. They regularly publish research papers and smaller models, and they have launched initiatives to explore shared public policy, safety, and standards research.
While the shift to a more closed-source model for GPT-3 and ChatGPT represents a significant change in approach, it reflects OpenAI’s evolving understanding of the balance between openness, safety, and responsibility in the age of advanced AI.
Is There An Open Source Equivalent To ChatGPT?
If you’re looking for open source alternatives to ChatGPT to play with, then there are a couple new kids on the block when it comes to matching its large language programming model.
However, with the first one we’ll take a look at, you’ll be really lucky if you’re able to use it. Plus, you will probably need a computer with loads of RAM if you’re able to run it at all — a laptop just won’t cut the mustard.
PaLM + RLHF
It’s called PaLM + RLHF, and has been developed as an AI text generator that kind of works like ChatGPT. It brings together Google’s PaLM large language model, and a technique that is known as Reinforcement Learning with Human Feedback, or RLHF.
The two work together to do things like generate computer code, draft reports and emails, and other text-based tasks, just like ChatGPT does. The big difference is that PaLM + RLHF hasn’t been programmed and trained on old website data like ChatGPT has, although it does essentially ‘predict’ words to generate.
ColossalChat
Another open source alternative to ChatGPT is ColossalChat It’s an AI chatbot that can have conversations like ChatGPT, as well as respond to requests in prompts, and write code.
ColossalChat is based on Coati, which is an acronym for ColossalAI Talking Intelligence, the large language model based on Meta’s open source LLaMa. RLHF is also a feature of ColossalChat, just like it is in ChatGPT, so it learns from human feedback just like a dog does when learning how to do a new trick — and the AI ‘gets rewarded’ for good behavior.
Luckily, ColossalChat is much easier to use than PaLM + RLHF, and it’s free. Plus, you don’t even have to create an account to use it, which makes it a great open source alternative to ChatGPT.
Future Versions of ChatGPT?
And finally, ChatGPT’s parent company OpenAI recently announced that it was going to release a new AI language model that is open source.
There hasn’t been a release date announced yet though, but it’s thought that the growing competition in the booming open source AI development race has spurred on OpenAI’s decision to do so.
Conclusion
So, is ChatGPT open source? The simple answer is no, but you now know of a couple options available for you to have a look at so that you can create your own open source AI chatbot project that just might be the one that becomes more popular than ChatGPT.
Plus, it probably won’t be long before OpenAI releases its open source language model, considering the recent news reported on Reuters.
