The open-source vs. closed-source AI debate is in the spotlight.
The divide is playing out among big tech companies vying to gain an edge in the great AI race.
Meta launched its open-source large language model (LLM), Llama 2, earlier this year, while Google and OpenAI have taken closed approaches so far.
The private LLM developer market is also split.
Proponents of open-source AI — where source code is available for public use — highlight advantages such as:
- Increased model flexibility & customizability
- Accelerated innovation that comes with a large developer community
- Transparency into AI models, data, and code vulnerabilities
But opponents of open-source AI fear it will be misused to fuel cyberattacks, AI-generated hate speech, and more — and highlight closed-source LLMs’ performance edge, additional development resources from deep-pocketed backers, and greater support from the companies that developed them.
While developers debate the philosophy of open vs. closed, early LLM adopters appear to be more interested in the day-to-day practicalities of what companies are offering.
For this buyer we spoke with, OpenAI’s ease of use stands out. But open-source models are attractive for their potential cost-savings.
OpenAI
OpenAI’s developer API and the developer experience is definitely the best. It’s really managed, super clean APIs, well documented, and has the most integrations. There was a big push toward using open-source models and then fine tuning them with your own data. That’s still a big thing. We’re actually evaluating that for some of the public data set stuff because it’s a lot cheaper, especially if you add in more huge amounts of data versus a giant OpenAI model. — Partner, Early-stage VC firm
Read the full transcript here.
This OpenAI customer, who is also considering Hugging Face and Meta’s Llama 2, indicates that open-source may have better ROI, even if performance is weaker.
OpenAI
Meta’s invention… it may be cheaper than the OpenAI [model] because it’s open source. Then we believe the performance of the Hugging Face and also the Llama 2 is also comparable to the OpenAI [model]. Maybe just a little bit weaker than that, but maybe the overall…ROI is quite a good deal. — Cloud, Data & AI Lead at Fortune 500 company
Read the full transcript here.
The open-source AI development market map
Beyond open-source foundation models, a number of vendors have emerged over the years to offer open-source tools for different parts of the AI development process.
For example, in LLM application development, Rasa provides an open-source conversational AI framework.
Rasa Technologies
We considered several other vendors in this space before choosing Rasa Technologies, such as IBM and Watson, LivePerson, Eva from Amplify, and Sprinklr’s conversational bot solution…We leveraged Rasa Technologies as the natural language understanding (NLU) on the back end…The pros for us of using Rasa Technologies were the price model and our ability to grow with the product, the quality we got for it, and the support. We were able to own our own data and our own environment and set things up how we wanted it. — Senior Manager, Fortune Global 500 company
Read the full transcript here.
Dig in with the full map.
Our AI research director recently dove into the landscape in our Future of Large Language Models briefing — download the recording for free here.
If you aren’t already a client, sign up for a free trial to learn more about our platform.