Introduction
Generative AI - AI that enables users to quickly generate new content based on a variety of inputs. Inputs and outputs to these models can include text, images, sounds, animation, 3D models, or other types of data.
- Gen AI - 2014, LLMs - 2018
- Scaling Challenges in Generative AI
- Just inference is MBs vs GBs w/ normal ML and DL
- Training is alot more computative!
- Incorporating Multi-Tenancy in Requirements
- Need a system to be EXTREMELY scalable
AI @ Scale w/ AWS
- AWS supports Gen AI in all layers of the Stack
- AWS offers 3 levels of services: ML Frameworks & Infra, AMAZON Sagemaker, AI services
- Some new services:
- Amazon CodeWhisprer - Kind of like github Co-pilot, but from AWS
- AWS Bedrock - Easiest way to build and scale generative AI applications w/ FMs
- AWS in Canada:
- Building more regions in Canada (Quebec, Toronto, Vancouver, etc), for more services and low-latency available to public
Blackberry - Threat Zero (Cybersecurity)
- Blackberry stopped 1,578,733 malware-based cyberattacks - 12 attacks every minute
- 200k of those attacks were NEW attacks, NEVER seen before!
- Traditional cybersecurity rules and old solutions are not fully effective anymore
- Most Targeted Industries: (40%) Financial Institutions, (15%) Healthcare & Pharma, (12%) Food and Staples, … and Other
- The attacks have increased by 50% in the last 2 quarters! (Hypothesis: Generative AI is helping them)
- Solution: Blackberry Cylance Security
-
Ensemble Hybrid Model Inference
-
Threat Zero Cloud Platform running on AWS serving BILLIONS of DAILY inference
-
Diverse models trained on Peta Bytes of data
-
Classifier of Malicious Behavior and scoring Service
- Solution Efficiency:
- Endpoint and Cloud Hybrid models enable deeper analysis
- Operates at scale w/ near zero False positive tolerance !
- virustotal/gui/stats - Cylance has zero false positive @ scale
- Key Takeaways:
- Scalable cloud AI platform is a MUST to defend
- Distributed ML Models that understand behaviour and predict outcomes
Ada Limited - Uses Gen AI in Customer Service Automation
- Interact across every channel: Web, Mobile, Voice, SMS, Social Media