Free Databricks-Generative-AI-Engineer-Associate Actual Exam Questions - Question 9 Discussion
A Generative Al Engineer is tasked with developing a RAG application that will help a small internal group of experts at their company answer specific questions, augmented by an internal knowledge base. They want the best possible quality in the answers, and neither latency nor throughput is a
huge concern given that the user group is small and they’re willing to wait for the best answer. The topics are sensitive in nature and the data is highly confidential and so, due to regulatory requirements, none of the information is allowed to be transmitted to third parties. Which model meets all the Generative Al Engineer’s needs in this situation?
A, because it’s lightweight enough for secure on-prem use without third-party data sharing.
A imo, Dolly 1.5B is smaller and easier to run fully on-prem without sending data out, which matches the confidentiality need. The question doesn’t guarantee they have big hardware for Llama2-70B.
Option D, since only it can be fully on-prem to meet the strict confidentiality rules.
Maybe D is the best fit since it can be fully hosted on-prem, ensuring no data leaves the company, which is crucial for confidentiality. The size also suggests higher quality compared to Dolly or BGE-large.
Makes sense to back D here as well because the key is keeping data fully on-prem and getting high-quality output. GPT-4 (B) is out since it involves sending data offsite, which breaks the confidentiality rules. Dolly 1.5B (A) might be safe privacy-wise but it’s too small to deliver the best quality answers. BGE-large (C) is also less capable on quality compared to Llama2-70B, so D really fits best for this use case.
Makes sense to go with D here since Llama2-70B can be deployed fully on-premise, so no data ever leaves the company, which ticks the privacy box. Also, it’s one of the better models for quality answers without worrying about speed. B and C are out because of third-party hosting or lower quality, and A just isn’t strong enough for top-tier results. So D fits best overall.
Actually, I’d rule out C here because BGE-large isn’t known for top-tier quality compared to Llama2-70B, especially when you want the best answers. Since the user group is small and latency isn’t an issue, running a bigger model like Llama2-70B on-prem makes the most sense for quality and privacy. Dolly 1.5B is definitely safe privacy-wise but just can’t match the performance you’d get from Llama2-70B for complex, sensitive queries. So D really checks all the boxes here without sacrificing quality or compliance.
I’m thinking the key here is the combination of top quality and strict privacy. So that pretty much rules out GPT-4 (B) since it’s a third-party service and sends data offsite. Dolly 1.5B (A) is good privacy-wise but probably not powerful enough for the best answers. Between C and D, Llama2-70B (D) is much larger and likely to produce better results on complex queries than BGE-large (C). Does anyone think BGE-large could be strong enough, or is it just safe to go big and local with Llama2?
D imo, it’s really about balancing quality with privacy here. Dolly 1.5B (A) is open source and good on privacy but lacks the power to deliver top-notch answers compared to Llama2-70B. GPT-4 (B) is solid but sends data to OpenAI, which breaks the no third-party data rule. BGE-large (C) isn’t as well-known for top-tier performance or on-prem options. So, Llama2-70B ticks both boxes: strong model for high quality and fully on-prem for strict confidentiality.
Option A could work since Dolly 1.5B is open source and can be run entirely on-prem, ensuring no data leaves the company. It might not be as powerful as Llama2-70B but meets the privacy need well.
D (Llama2-70B fits best, no third-party data sharing)