Amazon Quick exhibits a significant flaw in session management: new chat sessions are not properly isolated from previous conversations. Information discussed in earlier sessions persists and influences responses in supposedly fresh sessions, leading to:
- Hallucinated responses based on context from unrelated prior conversations
- Inability to start clean sessions despite creating new chats
- Contaminated outputs that reference information never provided in the current session
This fundamentally makes it impossible to have independent, context-free discussions. Also curious why Amazon Quick demonstrates notably inferior reasoning capabilities compared to other leading large language models. Why does a platform marketed as an enterprise AI solution perform substantially worse than comparable LLMs in fundamental reasoning tasks?
Hi @Jonathan_Lowe ,
WThank you for taking the time to share such detailed feedback about your experience with Amazon Quick Suite. I understand how frustrating it can be when session persistence issues affect your workflow/use-case, and we genuinely appreciate you bringing these concerns to our attention.
We’ll share your feedback with our product team to help us continue improving the platform.
This memory feature was recently launched “ memory feature for chat agents “[1] that allows Quick Suite to remember your preferences across conversations, eliminating the need to repeat context about response formats, dashboards, and integrations in each session.
Alternatively, Users also control what Quick remembers about them – all the memories are viewable and removable by users, and users have the choice to start chat in Private Mode in which conversations are not used to infer memories or delete previous conversations manually(not ideal)
Hope this helps and thanks again for sharing your feedback.
[1] Amazon Quick Suite now supports memory for chat agents - AWS
Quick is still hallucinated from conversations months ago. I will ask it to revise a message or document and it pulls in completely irrelevant and wrong context from past chats. I have tried creating new chat agents to prevent this but nothing works. Quick is unusable at this point.