OPINION:
Researchers and practitioners in the AI space were alarmed this month after discovering that GPT-4, the latest and greatest upgrade to the AI chatbot ChatGPT, is no longer open-sourced. This decision could have the unintended consequence of stifling innovation.
Despite its founding mission statement and commitments to collaboration, OpenAI has locked up ChatGPT’s secret sauce and made it impossible for third parties to vet OpenAI’s claims that the new GPT-4 has addressed the safety, accuracy, and bias concerns recently raised by users.
This includes limiting access to crucial details about the machine learning data set and to the methodology used to train the new ChatGPT. Previously, researchers and the general public could use this to validate and study how the underlying AI model works.
In an interview with The Verge, chief scientist and co-founder Ilya Sutskever confirmed OpenAI’s decision to deny access to critical information previously open-sourced, citing the growing “competitive landscape” of AI chatbots as well as hypothetical concerns over “safety” should bad actors utilize the GPT-4 model to do harm.
Many believe that closing off models in this way could stunt the long-term growth of generative AI. This is especially true for the many small, independent developers relying on open-source tools like ChatGPT to power their products.
Early internet pioneers faced similar conundrums decades ago and decided to build their networks with open protocols, standards, and a light-touch regulatory approach. As a result, the internet’s immense benefits and capabilities were made widely available to everyone without restricting access to the underlying technical architecture. Those decisions created robust market competition and widespread economic prosperity across the states without unnecessary government intervention.
We are at a similar inflection point now in the development of artificial intelligence. If private sector leaders in this space, including OpenAI, follow the proven model established by the open internet, entrepreneurs will be able to experiment with tools like ChatGPT and build on the GPT-4 model in innovative ways.
More concerning is the threat posed by lawmakers in the U.S. and Europe, who stand ready to interfere in the free market by enacting overly restrictive mandates on AI, chatbots, autonomous systems, and algorithms if given the opportunity.
In addition to President Biden’s recent executive order targeting AI, a group of federal lawmakers sponsored legislation last Congress to grant the Federal Trade Commission sweeping new authority over automated systems and their algorithms to combat alleged bias in critical sectors such as health care, housing, and education. Overseas, European regulators are editing their broad proposal for an Artificial Intelligence Act to encompass general-purpose AI like ChatGPT.
Those harsh policies would stifle America’s blossoming AI sector, punish good faith actors, and limit consumers’ access to this transformative new technology. Government officials should step out of the way, embrace the free market, and support our best-in-class technology innovators in their mission to make modern AI tools accessible and affordable to all.
- Jake Morabito is director of the Communications and Technology Task Force at the American Legislative Exchange Council (ALEC).
