The Rush to AI: Now the Hard Part
The introduction of ChatGPT by OpenAI in March 2023 marked a significant milestone in the development of large language models (LLMs). The technology’s potential to transform industries and revolutionize the way we interact with machines sparked a frenzy of interest among enterprises worldwide. Senior executives were eager to capitalize on the trend, and IT teams were tasked with identifying, prioritizing, and developing AI-powered projects that would showcase their company’s commitment to innovation.
However, as the initial excitement begins to wear off, IT teams are now facing the daunting task of developing useful, accurate, and attention-grabbing AI applications. Recent roundtable discussions sponsored by CIO have highlighted the challenges and issues that arise when building and delivering AI solutions.
One of the primary concerns is the issue of hallucinations, where AI systems provide incorrect or poor responses. To mitigate this risk, many companies are restricting AI systems’ ability to interact directly with customers, instead using them to assist human staff who can serve as a firewall against providing bad advice.
Data problems are another significant hurdle. IT executives knew that the data used to develop AI apps was often not ready for the task. Data formats varied, and much of it was stored in legacy systems, requiring extraction and integration. Moreover, AI demands real-time or timely data feeds, which can be challenging to deliver. The integration of disparate datasets with different timing can create a single reality, and issues related to compliance, governance, and privacy are particularly thorny.
Infrastructure is another area of concern. The development of AI solutions requires significantly more costly computing resources, causing sticker shock for many enterprises.
In conclusion, IT executives are waking up to the realities of developing AI-enabled tools, but executive management is still putting pressure on them to deliver enterprise AI. The companies that can successfully navigate these challenges will have a significant head start in the AI race. However, for this to happen, IT vendors will need to address these sticking points to maintain momentum.
The hard part of AI development
Data integration challenges in AI development
The high cost of computing resources for AI development
