A Microsoft executive emphasized the importance of AI learning to seek assistance from humans, as it could still save “tons of money”while receiving some much-needed help.
Currently, the tech industry is saturated with generative AI tools. Prominent companies such as OpenAI, Google, and Anthropic are actively working towards improving the accuracy of these tools by minimizing occurrences of AI “hallucinations,”in which incorrect information is generated. However, they have not yet reached their desired level of accuracy.
Vik Singh, a vice president at Microsoft, stated that many problems with existing AI models can be resolved by acknowledging when they are unsure of what to do.
During an interview with AFP, Singh expressed that the main issue today is that models do not speak up and admit when they are unsure and in need of assistance.
Singh believes that an AI model with a more modest approach could still hold significant value. Despite needing to rely on human assistance 50% of the time, it would still result in substantial cost savings.
According to AFP, Singh joined Microsoft in January and assumed the role of leading the teams responsible for developing Copilot in the summer. Copilot is an AI assistant created by Microsoft, focusing on sales, accounting, and online services.
Despite being a commonly used term this year, AI has not yet reached the promised revolutionary phase. According to The Economist, despite significant investments from tech companies in AI development and related hardware, the technology has yet to make a significant economic impact.
The Information reported in July that OpenAI, the current leading AI developer, faces the possibility of declaring bankruptcy within a year and potentially losing up to $5 billion by 2024.
To remain viable, the startup requires additional funding, which could potentially come from major companies such as Apple, Microsoft, Nvidia, and Tiger Global. This could result in a valuation exceeding 100 billion USD.
Leave a Reply