If you missed my day 3 part 1, you can view it here.

AI Vector Search: What’s New and What’s Next

Before Larry Ellison’s keynote, I attended an insightful session on Oracle’s AI Vector Search by Tirthankar Lahiri and Maria Colgan (SQLMaria).

Traditional Databases vs. Vector Search

• Databases excel at finding exact values in structured data, but struggle with identifying similarities rather than exact matches.
• Vector Search solves this by enabling searches based on semantic context, allowing systems to find items that are similar in meaning, not just identical in value.

How It Works

• Vectors represent data points in a way that captures relationships and similarities.
• For example:
  • Vector for “Apple” are closer to vector for “Pear” then it is for “Tiger”.
  • Searching for “waterfall” returns results related to waterfalls, even if the word isn’t explicitly mentioned.
  • Searching with an image (e.g. a house photo) retrieves visually similar houses.

Benefits

• No need to ship data around, everything resides in the database, eliminating synchronisation issues.
• Can be integrated into any enterprise application.
• Supports image-based searches and natural language queries for richer user experiences.

Performance & Architecture

• Vector indexes ensure top-K matches are consistent, even if the full search set varies. Trade-off some search accuracy for 100x speed up.
• In-memory searches are fast but limited by memory size and volatility.
• On-disk searches use partitioning to eliminate irrelevant vectors efficiently.
• AI agents typically run in the mid-tier but can now perform inference directly with Oracle Database.

This session highlighted how Oracle’s AI Vector Search is redefining enterprise search capabilities. By moving beyond exact matches to semantic similarity, businesses can deliver smarter, more intuitive experiences, whether it’s finding related content, identifying similar products, or enabling image-based searches. And because the data stays within the database, organisations avoid costly synchronisation challenges while maintaining speed and scalability.

Keynote: Oracle Vision and Strategy

Next, I attended the keynote by Larry Ellison, Oracle’s Chairman of the Board and Chief Technology Officer for insights into Oracle’s AI strategy and future vision.

I arrived half an hour early to secure a good seat, but just as the keynote was about to begin, I discovered it had been delayed by an hour.

I made the most of the break with a quick cup of tea, stopped by the Oracle ACE booth, and had a great chat with Alex, Oracle’s Exadata Product Manager, as well as my client James. Then I headed back to my seat, ready for the big keynote.

Unfortunately, even with the hour delay, Larry didn’t speak live at the keynote. I’m not sure if it was due to a delay, security, or health reasons, it was unknown. Given that Larry is now over 80 and doesn’t appear to be as fit as in the past, that could have been a factor. Instead, we got a video feed. I wasn’t sure if it was broadcast live from elsewhere, but as the video showed the actual stage, I suspect it was a recording from a dry run, though I can’t say for certain. Nevertheless, I listened attentively as if he were there, making sure to capture the key messages.

Larry delivered a thought-provoking keynote that positioned AI as the most transformative technology since the Industrial Revolution, bigger than the railroads.

AI as the Highest-Value Technology

• “AI is the highest-value technology we have ever seen.”
• Building and training AI models is a massive opportunity, especially when leveraging private data, most AI models today are trained on public data, but private enterprise data (often in Oracle databases) is where true differentiation happens.

Why AI Needs the Internet

• Larry compared AI learning to a robot mastering tasks like cleaning, cooking, or playing the piano: “They would just watch videos on the internet very rapidly, faster than real time, and then do it better.”

AI in Healthcare and Surgery

• AI offers microscopic vision, enabling precision in surgery:
  • Humans remove layers of cancerous skin and check repeatedly.
  • AI robots can detect and remove cancer cells in real time, reducing risk and improving outcomes.

Powering AI at Scale

• Human brain: 20 watts
• AI models: 1.2 billion watts enough to power 1 million homes
• Oracle is investing in grid power and onsite natural power generators to fuel massive data centres.
• Data networks interconnect the largest AI multimodal models, enabling faster training and inference.

Privacy and AI

• People want to keep their data private yet still reason on it.
• Oracle’s new AI Data Platform solves this:
  • Enables AI on private data without sharing it externally.
  • Supports AI agents using RAG (Retrieval-Augmented Generation) for tasks.

Real-World Use Cases

• Healthcare automation:
  • AI determines which drugs work best for which patients.
  • In the US, models are trained on healthcare policies to personalise care based on eligibility.
• MRI scans:
  • Larry shared his own experience after a bike accident: “They were counting broken ribs, but there was lots of other data ignored. AI will look at everything.”

Speed and Scale

• Training and inference used to be separate; now Oracle enables both at scale and speed, solving complex problems faster than ever.

Larry’s keynote reinforced Oracle’s commitment to AI built-in, not bolted on. From powering multimodal models with sustainable energy to enabling private data reasoning, Oracle is positioning itself as a leader in enterprise AI. The vision is clear: AI will not only transform industries but redefine how we live, work, and heal.

You can view my day 3 part 3 here.

If you found this blog post useful, please like as well as follow me through my various Social Media avenues available on the sidebar and/or subscribe to this Oracle blog via WordPress/e-mail.

Thanks

Zed DBA (Zahid Anwar)