Use LLMs like you use a Dishwashing Machine, not an Escuelerie

Use LLMs like you use a Dishwashing Machine, not an Escuelerie

Note:? All the text in this post was generated by myself without the use of generative AI.? However, the cover image and two images in the content were generated leveraging Stable Diffusion and Fooocus, both of which can be employed in generating content in concert with the Superwise AI Operating System by Blattner Tech that is shown in the last image.

I blame David Kramer for pointing me to The Bear.? As my spouse knows, I know little about a kitchen, let alone a commercial kitchen.? However, my early resume does include being a dishwasher in a commercial kitchen and so will use the cooler word for dishwasher, namely the French word “Escuelerie”.? And The Bear is a great show, so yes, I will tie using LLMs in the same way we use dishwashing machines.? Here we go :).


Without doing any research, I would hazard to guess that just about 100% of commercial kitchens leverage a dishwashing machine.? In fact, I would say that a dishwashing machine is a critical component to a commercial kitchen.? However, the dishwashing machine is used within the following context:

  • The dishwashing machine is not used for all dish cleaning, The escuelerie leverages the dishwashing machine for what it is designed for and not what it is not.? I.e. delicate stemware may not be suitable for a dishwashing machine.
  • The dishwashing machine is not 100% perfect.? The escuelerie may prepare dishes for better results by washing large food items or baked on cheese prior to using the dishwasher.? Also, the escuelerie checks the results after use to make any corrections, i.e. a spot on a glass.
  • The dishwashing machine must be monitored and maintained.? I.e. gages are used to make sure proper temperature to kill germs are employed.
  • If the dishwashing machine breaks or a better dishwashing machine option is on the market, it must be able to be swapped out without disrupting the whole kitchen operation.
  • An escuelerie is still employed to properly input and time dishwashing operations.
  • And a sous chef still orchestrates operations where dishwashing is one of the functions that need to be sequenced in the kitchen.

Ok, so no surprise but I am going to use the dishwashing analogy to LLMs and Enterprise Operations :).? See how exactly similar the below paragraph and bulleted items are :)


With doing a lot of research and hands on building and experimenting with LLMs, I would confidently predict that just about 100% of enterprise operations will have to leverage LLMs. In fact, I would say that an LLM or even an LLM for each task should already be a critical component of an enterprise operation. However, the LLM(s) should be used within the following context:

  • The LLM is not used for all operations. Knowledgeable enterprise operators should leverage what the LLM is designed for and not what it is not. I.e. an LLM should not be used to communicate delicate information to an important customer.
  • The LLM is not 100% perfect. An operator needs to prepare prompt engineering, retrieval augmented generation, reasoning and action frameworks as well as organize data in optimized ways prior to using the LLM. Also explainability and guardrails need to be employed to gain confidence in responses returned from the LLM.
  • The LLM(s) must be monitored and maintained. Drift and many other metrics must be monitored for proper operations. Models can and will have to be updated.
  • If the LLM is no longer useful or a cheaper or better LLM becomes available, it must be able to be swapped out without disrupting the whole enterprise operation.
  • Agents must be utilized and employed to properly align the LLM to specific Enterprise Operations.
  • Humans aided by Orchestration capabilities are still needed to run enterprise operations that are powered by LLM and non-LLM powered capabilities.

With this “context” LLMs can and will be as valuable to enterprise operations and dishwashers are for commercial kitchens. Of course, we at Blattner Technologies do think about this a lot so the below paragraph and bulleted items are also identical :).


https://docs-swe.superwise.ai/

Blattner Tech makes AI Operational for Enterprises.? Our industry leading AI Operating System, Superwise, provides the context to allow LLMs to be employed in enterprise operations:

  • Superwise provides the necessary abstraction between Enterprise Operations and the LLM(s).? This allows knowledgeable enterprise operations experts to employ LLMs for what they were designed for and not for what they are not.
  • As LLMs are not 100% perfect, Superwise allows for operators and developers to easily perform prompt engineering, retrieval augmented generation, reasoning and action as well as data transformation for use with LLMs.? Also, Superwise provides explainability and guardrails to gain confidence in responses returned from the LLM(s).
  • Superwise provides industry leading Model Observability detecting and alerting on model drift and many other parameters.? Updated models are easily employed.
  • Superwise allows you to select the LLMs that you want to employ for your specific enterprise task and operations and allows you to swap out those LLMs without disrupting operations.
  • Superwise provides easy ability to create, employ and operate purposely designed agents to directly align to specific business operations.
  • Superwise allows us humans to define orchestration and insert multi-agent services into business operations through easy-to-use sequence definition, user interfaces, APIs, and SDKs.?

The Bear is a great show.? It highlights the very real human interactions that we all experience.? And for me, the setting of a commercial kitchen brings a whole new enterprise operation that I was not as familiar with to be excited about.? Yes, enterprise operations are exciting.? Employing LLMs make those operations better as long as they are employed like commercial kitchens employ escueleries.

Mark Stadtmueller

要查看或添加评论,请登录

社区洞察

其他会员也浏览了