Have you found ways to take advantage of AI hype or areas where you needed to temper the hype with reality?
Sort By:
Oldest
Chief Technology Officer in Software6 months ago
The hype was beneficial in that it drew the attention of top management and led to investments and funding. However, the models we are building are not ready for enterprise use just yet. Many models available, especially in generative AI, can do your work up to a certain extent. But after reaching around 75% of accuracy, improving further becomes very difficult due to the infrastructure required for training and other factors.Apart from the hype and reality, there's also a limitation in terms of technology. The technology needs to mature a bit more so we can build more enterprise-grade solutions. Many companies have started doing POCs and pilots, but they're not able to release them to their internal stakeholders or external customers because of these reasons. The models are not that mature yet, so this is another aspect to consider.
Chief Data Officer in Travel and Hospitality6 months ago
GenAI has captured the imagination of boards and SLT. This has made AI a mandatory topic this year. As a result, we have been able to raise awareness of the existing immaturity of core data, BI, and analytics capabilities.
There's a debate about whether traditional machine learning or deep learning is required, or if the same problem can be solved using traditional methods. Maybe, you don't have to invest so much in a compute-heavy deep learning models. It all depends on the return on investment and how the solution fits into the existing workflows.
There's also a focus on how responsible the solution is and how it aligns with regulations. As regulations transform, the initial hype is now resulting in a focused approach towards aligning with data pipelines, ROIs, and forthcoming regulations.
Misinformation and cybersecurity are also coming to the forefront. Leaders and practitioners are more informed and are not just going for the hype. They want to see the practicality. This will ensure that AI doesn't become something akin to what happened with crypto, but something that adds value to everyone.
People are going for low hanging fruit to try out things and remove hallucinations. Techniques like retrieval augmented generation are being tried out to enhance fine tuning. This is a time to double click on those implementations and data: see how they work. If we can get over this phase, it will result in clear solutions that have been proven for the data pipeline, compute, ecological and other impacts.