Can Opensource AI Steal a March on Open AI and Google?

Can Opensource AI Steal a March on Open AI and Google to give an easy consumer access to the code of AI generated chat

DQC Bureau
Updated On
New Update
User Experiences

The general belief is that generative Opensource AI models will significantly improve adaptability by enabling open source communities to innovate by effectively using models in domain-specific scenarios. With Opensource AI, organisations can have more control over the model's execution, data input and output. Another critical aspect is that the models can also be deployed on local infrastructure and have transparent and comprehensible components.


ChatGPT and other GPT versions from Open AI are closed-source models, which means that its code is not available to the public. This makes it more difficult to modify or customise, but it also means that it is generally more stable and reliable. ChatGPT (& its future versions) is also trained on a larger dataset of text and code, which gives it a wider range of knowledge and abilities.

Key question that needs to be answered is, whether the opensource and research community can steal a march over the billion dollar budgets and top talent driven closed source commercial technology from Open AI/MSFT, Google and others in the fray.

It is truly early days and the space will evolve but it is more than likely that business applications leveraging open-source generative AI/LLM will have a significant market share and hence enterprises need to have an open ended and flexible technology selection strategy. If current trends continue, there will be a multitude of use cases where open source will steal a march over closed source.


Some of the variables that will determine which way the technology adoption will go include regulatory impact, ethics and bias, interpretative, interoperability, IP protection, cost vs value, to list a few.

The smart contributions of the research community in the past and the general activity we are seeing in this space, including the recent Meta decision to open source its Generative AI/LLM Llama 2 project (for commercial use) indicates that opensource technology options will dominate in the vertical AI space, where enterprises would like to build their own IP, leverage their own company and industry data and build out vertical domain use cases that emphasizes context over generality.

For example, legal document and contract analysis is more effective on models trained on the industry segments and company domains, as opposed to the larger general-purpose AI approach. This is due to the fact that this is a specialised area that is built on professional domain knowledge. Similarly financial statement analysis, network optimisation in supply chain, industry research reports, tax regulations analysis and Q&A, all represent use cases where enterprises need to actively invest in building alternative open-source technology options as opposed to solely relying on closed source technology.   


To sum up, for everyone who has simplified Generative AI applications to prompt engineering and UI build outs, the recommendation is to look at the progress and the leader boards with opensource AI tech that will require additional investments and focus. What is also worthwhile is to track the domain specific GPTs that are getting launched in Finance, Healthcare and Legal as options. While composing poetry and writing essays, will still be the dominion of closed source tech, enterprise apps is a different ballgame.

--By Pani Baruri, MD & CEO, Algoleap Technologies

Read more IT news here

Read products news here