In a recent earnings call, Apple’s CEO Tim Cook provided a cautious insight into the company’s interest in generative artificial intelligence (AI). While refraining from revealing Apple’s plans, Cook stated that he sees AI as important and holds vast potential. While the company has not yet ventured into the generative AI space, its research work indicates some inclination towards related models.
One such research paper, scheduled for publication at the Interaction Design and Children conference in June, discusses a system aimed at reducing bias in the development of machine learning datasets. AI’s tendency to make unfair or erroneous predictions based on incorrect data is a major concern for the ethical development of generative AI models. The new Apple research suggests an innovative method to address this issue.
The paper proposes incorporating multiple users for developing an AI system’s dataset to ensure equal input. In contrast to the existing generative AI development that includes human feedback at later stages, Apple’s research incorporates user feedback at the inception of model development, democratizing the data selection process. Although designed for educational purposes and intended to generate novice interest in machine learning development, the research offers an alternative approach to combating bias.
However, scaling these techniques for use in large-language models (LLMs) like ChatGPT and Google Bard could be challenging. Successfully developing an LLM without bias stands to significantly impact the technology sector, specifically fintech, cryptocurrency trading, and blockchain. Unbiased stock and crypto trading bots with human-level reasoning might revolutionize the global financial market by making high-level trading knowledge accessible to all.
Furthermore, proving an unbiased LLM could help address government concerns about safety and ethics in the generative AI industry. For Apple, this development would be particularly advantageous, as any generative AI product it creates or supports could leverage the iPhone’s built-in AI chipset and the 1.5-billion user base.
In conclusion, Apple’s research into addressing bias in generative AI models offers an interesting and collaborative approach to dataset development. Although the scalability of the proposed methods remains to be tested, the potential impact on the global tech landscape, finance, and other industries cannot be underestimated. Keeping an eye on Apple’s developments in this space could provide valuable insights into the future of AI and its applications.