Has anyone tried out LLaMa 2? Would you build with it in its current state?

Innovation & Emerging Trends Machine Learning (ML)

4.9k views3 Comments

Sort by:

Senior Vice President, Engineering in Software2 years ago

Yes, we have recently tried this in a POC for one of our clients. It's quite useful if you are conscious of OpenAI API costs, this can be self-hosted and is able to cater to most of the GenAI kind of applications like Chatbots, Finding similar objects, summarising texts etc. Another benefit is that LLaMa2 can be easily trained using general libraries available for Python and you can get stuff done that OpenAI models would usually refrain from doing.

2 Replies

no title2 years ago

Llama 2 is a Meta product, what OpenAI API costs are you referring to?

no title2 years ago

What I meant was using OpenAI models via API costs a fortune if you use those at a huge scale. If you want to save this cost, it's rather easier to self-host LLaMa2 and pretty much get everything done at a lower expense.

Content you might like

Do you manage AI and traditional IT infrastructure investments as separate portfolios (budgets, governance and decision criteria) or do you combine them? What are the benefits and drawbacks of your approach?

What language has been most effective for helping non-technical leaders understand the true cost of delaying foundational IT work to instead focus on AI implementation? Can you share an example?

Have you established an AI governance team that’s different from your previous IT/data governance group?

Yes82%

No18%

No, but we're working on it

View Results

We're several quarters into the AI hype—where would you work with external AI experts to help your organization?

Getting started with our AI strategy39%

Evaluating vendors' AI capabilities21%

Building (Agentic) AI apps in-house24%

Upskilling employees on how to use AI12%

Other (See comments...)2%