Finding the balance of Generative AI adoption
Generative AI has the potential to impact every area over time —especially non-technical roles
Most generative AI efforts are taking place in companies as disconnected test pilots. Taking a careful, intentional yet flexible approach can strike the balance between caution and innovation.
Conservative & cautious
Organizations opting for a conservative approach prioritize stability and risk mitigation when integrating generative AI. They proceed cautiously, conducting thorough assessments and implementing strict regulations to avoid potential pitfalls. This approach is often observed in industries with strict compliance requirements, such as healthcare, finance, and legal.
For instance, in the healthcare industry, organizations have to comply with stringent privacy regulations and prioritize patient safety. They approach generative AI adoption conservatively to ensure compliance with regulations and maintain the highest ethical standards.
Similarly, legal firms handling sensitive client data should also adopt a conservative approach to ensure the privacy and confidentiality of information. But confidentiality is only one concern attorneys must be aware of - generative AI is confidently wrong often and tends to make things up.
Recently, Manhattan lawyer Steven A. Schwartz, came before Judge P. Kevin Castel, to answer for creating a legal brief filled with fake judicial opinions and legal citations generated by ChatGPT.
Schwartz expressed remorse and claimed he thought ChatGPT was a more robust search engine. "I did not comprehend that ChatGPT could fabricate cases," said Schwartz. The exacerbated judge questioned why Schwartz didn't conduct further research.
"The vast majority of people who are playing with them and using them don’t really understand what they are and how they work, and in particular what their limitations are," said Irina Raicu, director of the internet ethics program at Santa Clara University.
The incident has attracted attention in the tech world, legal profession, and wider public, highlighting the limitations of artificial intelligence.
It has also sparked a discussion on the responsible use of chatbots in the practice of law. "Paradoxically, this event has an unintended silver lining in the form of deterrence," said Stephen Gillers, an ethics professor at New York University School of Law.
Risk-Tolerant & innovative
In contrast, some organizations have taken a more experimental and fast-paced approach to generative AI adoption. They embrace innovation, seeking opportunities to leverage AI technologies to gain a competitive edge. These organizations are more willing to explore uncharted territory and accept potential risks in exchange for rapid progress and disruptive breakthroughs.
Tech startups are often at the forefront of embracing a fast and loose approach to generative AI adoption. They leverage technology to disrupt traditional industries and create new markets.
Furthermore, companies operating in rapidly evolving sectors, such as e-commerce and social media, are more inclined to adopt generative AI quickly with fewer internal barriers to getting started. These organizations recognize the need to stay ahead of the curve and continually innovate to meet changing consumer demands.
Stitch Fix is using AI and ML to enhance client experiences, particularly in creating engaging advertisement headlines and high-quality product descriptions.
To generate ad headlines, Stitch Fix combines latent style understanding, word embeddings, and few-shot learning. They map outfits and style keywords to a latent style space, select style keywords closest to the outfit, and then use GPT-3 to generate headlines based on the selected style keywords.
Human experts (copywriters) review and edit the AI-generated headlines to ensure they capture the style and align with the brand's tone.
"Our fine-tuned algo solution offers unbeatable time savings as well as excellent scalability without sacrificing quality of descriptions,” said Tianlin Duan, former data scientist, acquisition algorithms at Stitch Fix.
The expert-in-the-loop approach has been successfully used to generate ad headlines for Facebook and Instagram campaigns, improving efficiency without compromising quality. Stitch Fix then applies the same approach to generate product descriptions, which play a crucial role in e-commerce. They use fine-tuning to retrain a pre-trained base model on a task-specific dataset of expert-written product descriptions. This customized solution provides accurate and engaging descriptions tailored to clients' needs and written in the Stitch Fix brand voice.
"The expert-in-the-loop approach thus creates a positive feedback loop where human experts and algorithms work together to continually improve the quality of the generated content," said Duan.
Lessons learned
The varying degrees of success associated with generative AI adoption reflect the diverse approaches organizations take. While conservative strategies focus on stability and risk mitigation, they may limit the speed of innovation and potential breakthroughs. Conversely, risk-tolerant approaches can lead to rapid progress, but they also carry a higher risk of ethical concerns and unintended consequences.
Organizations that have embraced a conservative approach often prioritize compliance and ethical considerations, ensuring that they adhere to industry regulations and maintain public trust. However, they may face challenges in adapting to rapid technological advancements and missed opportunities for innovation.
On the other hand, organizations adopting a risk-tolerant approach may experience accelerated innovation and breakthroughs. However, they must remain vigilant about the potential ethical implications and unintended biases arising from unchecked experimentation. The key lies in striking a balance between innovation and ethical responsibility.
“The success of generative AI adoption lies in finding a middle ground. We need to leverage the caution and stability of the conservative approach while embracing the agility and innovation of the more risk-tolerant approach. By doing so, we can harness the full potential of generative AI while maintaining ethical standards and responsible practices.”
Hybrid Approaches
Looking ahead, the future of generative AI adoption will likely witness a convergence of conservative and innovative approaches. Organizations will develop frameworks and guidelines that enable controlled experimentation while ensuring compliance and ethical practices. This convergence will be driven by the growing awareness of the importance of responsible AI adoption and the need to balance innovation with ethical considerations.
Moreover, advancements in the explainability and interpretability of AI systems will contribute to a more comprehensive understanding of generative AI algorithms, enabling organizations to assess and mitigate risks effectively. This, coupled with robust governance frameworks and industry-wide collaboration, will facilitate the widespread and responsible adoption of generative AI.
The future of generative AI adoption will likely witness a convergence of conservative and less risk-averse approaches, with organizations developing frameworks that enable controlled experimentation while maintaining ethical standards. By striking this balance, we can maximize the potential of generative AI while ensuring responsible, ethical, and impactful adoption across industries.
AI Content Disclaimer: The human author who wrote this augmented the final post with Generative AI tools for ideation (ChatGPT), research (Perplexity AI) and image creation (Midjourney).
