Live from the event
Can’t watch the event? (https://www.oracle.com/cloudworld) We’ve got you covered. Follow along to our updates below.
4:32 PDT: Try out the performance optimization of your model, yourself: Http://bit.ly/Vianai
4:29 PDT: Demonstrate 100X improvement for Schneider Electric.
4:28 PDT: Edge machines live for decades… You need to run high-quality models on these, but usually they’re too big. Shrink down an AI model, and have them run on the device. Accelerated predictive maintenance by 1,000X.
4:26 PDT: Accelerated tabular data — we accelerated inference time by more than 100X, and shrank the footprint of the model by 300X.
4:25 PDT: Our acceleration is in software, regardless of the hardware that it’s running. 10X acceleration of inference for a semiconductor system, reduced the size by 27X.
4:23 PDT: Accelerate the performance of a model on an autonomous vehicle — created a 4X improvement on inference acceleration and shrank the size to 1/5 of it.
4:21 PDT: Performance of AI — the third part of my talk. As the models have become larger. GPT had 175 billion parameters. Google released a 540 billion parameter model called PaLM.
4:19 PDT: Payment processor processing thousands of events per second — the breakthrough for us is the sheer scale. Tested on 95 billion transactions in a quarter.
4:17 PDT: Bias, fairness, make it actionable — requires monitoring. This is our platform. This is designed for massive scale.
3:55 PDT: The video just started up and we’re tuned in. Keep an eye on the blog for more updates!