OpenAI's White Flag. How Open-source Models are Disrupting Big Tech.
When ChatGPT launched in November of 2022, it took the tech world by storm. Within five days of launching, the chatbot became a social media sensation and gained over one million users. Users were wowed by AI capabilities ranging from essay writing and coding to image generation. Companies excitedly implemented the technology to improve their processes and workflow. On the back end, OpenAI, Google, Meta, and other major players poured millions into building and scaling models. And smaller players licensed those models rather than getting squeezed out, knowing they would never have the resources necessary to venture out on their own AI frontier.
The AI Revolution
Then, in a stunning announcement, Sam Altman verbalized what many developers already suspected would eventually happen: “I think we’re at the end of the era where it’s going to be these, like, giant, giant models.”
Big tech’s glory days are over. Gone are the days of scaling AI in a silo. There are too many limitations on the number of data centers a company can build in a limited time frame, said OpenAI. And, even if it were feasible to continue feeding the monster, it turns out that size isn’t all that. In the race to grow bigger, data quality fell by the wayside. Funny, nobody seems to see the value in a model that produces limited or inaccurate results.
Fast, Flexible, and Cost Effective
So, what does the future of AI look like? It seems that the answer to scaling was there all along. While the big guys were taking the long road, developers worked on nimble open-source models behind the scenes. After all, they’re faster, more flexible, and affordable. And you know what? When using both models, most users can’t differentiate between an open-source model or ChatGPT. In one study, a significant percentage preferred using the open-source model.
Built For Collaboration and Innovation
Open-source software was developed in the late ‘80s by programmers as an alternative to expensive licensed software. The free tools were designed for collaboration. Developers were not only able to contribute and improve on existing programs, but they were encouraged to do so. Talk about continuous improvement. Better software and more efficient processes led to new ideas and solutions that would not have been possible.
Democratizing Access to AI Solutions
It’s no wonder the big guys are conceding. Open-source AI has democratized access to innovative AI solutions. Built on the fundamental principles of open-source software, new open-source AI models are pre-trained, pre-built systems that developers and researchers can use and improve without any proprietary restrictions or licensing fees. Private databases are estimated to hold 1,000 times the amount of data controlled by the five big tech players. Open-source AI models can unleash that data quickly and accurately.
Open-source Models Bring It
With open-source, innovation rules, allowing for natural language processing, image recognition, and cutting-edge solutions for vertical markets, including healthcare, finance, and transportation. A $100 open-source model with 13 billion parameters can easily take on a $10 million dollar high-end Google model with 540 billion parameters. Plus, it can learn and be developed in a fraction of the time.
Data Integrity
Let’s talk bias. It’s no secret that ChatGPT has produced some bloopers. Closed and proprietary systems give internal developers complete control and responsibility over their models’ behavior. Input and perspective is limited to a finite group controlled by a single organization. Open-source systems, on the other hand, build on the work of a global community of contributors. Inputs flow freely, and developers pool their knowledge base. Having more irons in the fire allows everyone to see what everyone else is doing. When something doesn’t look right, someone is sure to pipe up or fix it.
Everyone Wins In an Open Market
Here’s an example. A car dealership uses AI to qualify customers for loans. A private school uses AI to award scholarships. The two are not competitors, however, they use similar AI tools. Both organizations can license the same existing AI open-source model. Both organizations can also communicate back to the model to improve data quality for all future users. Everyone benefits: the car dealership, the school, and any other organization who taps the model in the future.
Big tech has been holding onto control for dear life, and now it’s coming back to bite them. Developers had no choice but to seek another option, and they found it in open-source AI models. Now who’s in the driver’s seat?