Bhavesh Upadhyaya on AI’s Ethical Future in Streaming

Moving beyond the hype surrounding artificial intelligence, the streaming industry is now focusing on its tactical and ethical applications to solve real-world problems. In a recent discussion, Bhavesh Upadhyaya, a Live Operations Expert with the Streaming Video Technology Association, emphasized the shift from viewing AI as a vague future concept to implementing it practically within streaming workflows. He advocates for a thoughtful approach that addresses both operational efficiency and the broader implications for businesses and creators.

Upadhyaya brings nearly two decades of experience to the table, having worked on major projects like the Beijing Olympics and held key roles at companies such as iStreamPlanet, Verizon Digital Media, and Warner Bros. Discovery. Today, he advises organizations on integrating AI and optimizing their operational strategies.

When discussing the tactical use of AI, Upadhyaya compares the challenge to “eating an elephant”, it’s too large to tackle all at once. He suggests breaking it down by identifying specific problems within streaming workflows and applying machine learning or generative AI to address them directly. This method moves past abstract industry predictions and delivers measurable results, such as automating repetitive tasks or enhancing data analytics.

The conversation also touched on past technological trends, like the 3D boom, which promised widespread transformation but failed to gain lasting traction. In contrast, today’s AI advancements feel more grounded, with practical applications already making an impact. However, Upadhyaya cautions against over-reliance on generative AI tools for critical tasks, noting their non-deterministic nature. He stresses the importance of understanding confidence scores and knowing when human oversight is necessary to verify and complete AI-generated outputs.

Versioning AI models emerged as another key topic, highlighting the need to treat AI systems like products with defined lifecycles and updates. This is especially relevant for backend processes that run continuously, where consistency and reliability are paramount. By versioning models, organizations can maintain control over AI performance and ensure it aligns with their operational standards.

Ethical considerations around AI implementation cannot be overlooked. Upadhyaya points to the delicate balance required in setting guardrails, knowing when to block or allow content, and building trust with consumers. Intellectual property concerns, potential job displacement, and the need for legal protections also demand attention. He believes that while businesses will inevitably evolve with AI, individuals must proactively educate themselves and adapt to these changes.

Reflecting on historical parallels, Upadhyaya notes that AI’s current centralized phase resembles the early days of mainframe computing, which eventually gave way to personal computers and mobile devices. He predicts a similar decentralization for AI, with powerful models eventually running on everyday devices. Learning from past technological shifts will be crucial for navigating this new era responsibly.

Ultimately, the discussion underscores that AI’s true value lies not in replacing human ingenuity but in augmenting it. By focusing on practical applications, maintaining ethical standards, and fostering continuous learning, the streaming industry can harness AI’s potential while mitigating its risks.

(Source: Streaming Media)