Hey Reader!
The most significant investment in Generative AI is at an all-time high. According to a Goldman Sachs report, AI investment will reach $200 billion globally by 2025. As per CBInsights, investment in Generative AI has risen to 4x in 2023 compared to the previous year. This all pressures organizations to get onto the bandwagon of Generative AI quickly.
Everyone is trying to find the best use case that can be implemented for the organization. Navigating the buzz is challenging and time-consuming for everyone, including me.
Over the last year, I have explored and curated different use cases published in the earlier newsletter, 100+ Generative AI Use Cases Across Industries. You can download the guide.
Today, we will discuss deploying large language models (LLMs) in the IT ecosystem, which involves choosing the right approach and following a strategic process. The most common question that comes to mind is how to decide to make or buy LLMs for your organization. Luckily, the Initiative for Applied Artificial Intelligence has provided 9 points on A Guide for Large Language Model Make-or-Buy Strategies: Business and Technical Insights to follow before deciding on this area.
- Strategic value
- Customization
- Intellectual property (IP)
- Security
- Costs
- Talent
- Legal expertise
- Data
- Trustworthiness
Knowing the various deployment strategies for LLMs will help you leverage their capabilities within your existing IT ecosystem. This will allow you to integrate and unlock your organization’s full potential seamlessly.
We have also discussed a few approaches for deploying Generative AI, starting with Consume, Embed, Extend, and build. According to Gartner, here’s a breakdown of the five significant approaches for deploying GenerativeAI in various applications:
1 – Consume: Gen AI Embedded in Applications
In this approach, generative AI is already integrated within applications. Users consume the AI capabilities as they are without needing to modify or customize the underlying AI models.
2 – Embed: Gen AI APIs in Custom Application Frameworks
Here, generative AI capabilities are accessed through APIs and embedded into custom applications. It allows for a more tailored use of AI functionalities within a specific application framework.
3 – Extend: Gen AI Models via Data Retrieval
This method extends existing generative AI models by retrieving and incorporating external data. This approach enhances the AI model’s capabilities by leveraging additional data sources.
4 – Extend: Gen AI Models via Fine-Tuning
This approach also extends generative AI models through fine-tuning rather than data retrieval. Fine-tuning adjusts the AI model’s parameters for specific tasks or data sets, providing more customized outputs.
5 – Build: Gen AI Custom Models from Scratch
The most advanced approach involves building custom generative AI models from scratch. This allows complete control over the AI model’s design and capabilities, which can be tailored to unique and specific requirements.
Deployment Steps:
- Evaluate Needs: Based on your selected use case, identify the tasks and functionalities you want the LLM to perform.
- Assess Resources: Consider the infrasrcuture resources, technical expertise, and budget.
- Choose Approach: Select the deployment option (consume, embed, extend, build) that best aligns with your needs and resources.
- Prepare Infrastructure: Depending on your chosen approach, you might need to set up cloud access, APIs, or dedicated infrastructure for training.
- Data Preparation: For embedding, extending, or building, ensure your data is clean, labeled, and relevant to the LLM’s task.
- Model Integration: Seamlessly integrate the chosen LLM or custom-built model into your IT ecosystem.
- Testing and Monitoring: Thoroughly test the integrated LLM for accuracy and performance and monitor its behavior in production.
Based on the observation, the best approach is to start small with the initial use case with a simple task and a less resource-intensive approach (consume or embed) before attempting complex implementations.
The most critical point is that Security is paramount to any approach; therefore, you must ensure your chosen LLM provider offers robust security measures, especially when dealing with sensitive data.
|
Weekly News & Updates…
This week’s AI breakthroughs mark another leap forward in the tech revolution.
- Claude 3 from Anthropic consists of Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku. It sets new industry benchmarks in reasoning, math, coding, multilingual understanding, and vision.
- GPT Researcher is an autonomous agent designed for comprehensive online research on various tasks.
- Struggling to Pick the Right AI Model? Let’s Break It Down
The Cloud: the backbone of the AI revolution
Favorite Tip Of The Week:
Here’s my favorite resource of the week.
- AI Decoded: Demystifying AI and the Hardware, Software, and Tools That Power It
Potential of AI
Things to Know
- Anthropic Cookbook provides code and guides designed to help developers build with Claude, providing copy-able code snippets that you can easily integrate into your own projects
The Opportunity…
Podcast:
- This week’s Open Tech Talks episode 129 is “Unlocking Customer Loyalty with Web3: An Inside Look with Eric Mchugh. President at SHOPX Labs
Apple | Spotify | Google Podcast
Courses to attend:
Events:
Tech and Tools…
Data Sets…
- Croissant: a metadata format for ML-ready datasets. The Croissant format doesn’t change how the data is represented (e.g., image or text file formats). It provides a standard way to describe and organize it
-
MuST-C is a multilingual speech translation corpus whose size and quality facilitate the training of end-to-end systems for speech translation from English into several languages
Other Technology News
Want to stay on the cutting edge?
Here’s what else is happening in Information Technology you should know about:
Earlier Edition of a newsletter
* { word-break: break-word; }
That’s it!
As always, thanks for reading.
Hit reply and let me know what you found most helpful this week – I’d love to hear from you!
Until next week,
Kashif Manzoor
The opinions expressed here are solely my conjecture based on experience, practice, and observation. They do not represent the thoughts, intentions, plans, or strategies of my current or previous employers or their clients/customers. The objective of this newsletter is to share and learn with the community.
|
|
