TLDR: To get the most benefits from generative AI, it is crucial to focus on ‘forks and knives’ – the tools that optimize the value of AI outputs, as well as ‘picks and shovels’ – the tools that construct AI solutions. Though the importance of ‘picks and shovels’ is acknowledged, it’s the ‘forks and knives’ that need to be given their due attention.
For any general-purpose technology, we tend to emphasize the ‘picks and shovels’ – the tools that enable the technology to function or make it ready. This analogy is borrowed from the California Gold Rush, where those who sold such tools reaped the profits. In generative AI, the ‘picks and shovels’ undoubtedly play a significant role. However, we need tools and technologies that work on what AI generates. They are like the ‘forks and knives’ you need to enjoy a good meal. Such ‘forks and knives’ in generative AI could be for:
- Integrating with other data sets,
- Ensuring up-to-date information,
- Testing for accuracy,
- Mitigating bias and hallucination,
- And much more.
Currently, several emerging tools fall into these categories. These are in the early stages and will evolve and solidify over time. They include:
- Agents for integrating generative AI solutions into your workflow,
- Retrieval-augmented generation (RAG) tools for using LLMs with your data,
- Vector databases for preparing the data for agents and RAG,
- Prompt engineering tools for efficient prompt management,
- Responsible AI tools to counter inaccuracies, biases, and hallucinations,
- And more.
Examples of these tools include Langchain, LlamaIndex, MS Semantic Kernel, and HuggingFace Transformer Agents (agents and RAG tools); PineCone and Weaviate (vector databases); MS Semantic Kernel, Promptable, Prompt Studio (prompt engineering tools); and Credo AI, Fiddler AI, SolasAI, Interpret ML, Arthur AI, Fail ML, Hive (responsible AI tools).
This list is not exhaustive, but it paints a picture of this emerging landscape.
As a technology leader in an enterprise or tech firm, you should focus on ‘forks and knives’ in addition to ‘picks and shovels’. For some of you, the focus could be solely on using ‘forks and knives’ when you not building your own generative AI. By doing so, you can gain a wider perspective of the arena and harness the power of this powerful technology to its fullest potential over time.
What do you think? What other tools have you found beneficial so far? Where are the gaps? Thank you.
Leave a Reply