In your Organization, are you promoting the culture of model differentiation? As far as I have seen, many of us are just getting out of the box proposed without taking care of the model selection as well as if the LLM should step back for an SLM. The adoption of SLM (Small Language Model) as Phil-2 (available on Azure OpenAI) or Orca2 or Falcon 7B, Nucleus 1B, ... involves remarkable benefits: limited BIAS, Hallucinations, low costs for training, and so on.
Chief Information Officer (CIO) in Healthcare and Biotech2 years ago
This is already happening as part of the evolution of LLM adoption by enterprises. It requires that one think of LLM architecture in “onion” layers. Essentially the innermost layer is the foundational models that LLMs are built on. They include the likes of GPT, Sparrow, LaMda, MT-Nlg, LlaMA, PaLM amongst others. These are the foundation models that are trained on general-purpose data, and later adapted for specific applications such as ChatGPT, Bard, and Midjourney which represent the second layer. Predictions are that the first two layers will consist of few big players that are able to invest significantly to train these models on mountains of data.
The emerging two layers are industry specific LLMs and enterprise LLMs, where enterprises are starting to leverage the first two to train their own models at much lower cost as the 3rd and 4th outermost layers respectively (aka SLM). This will see an ocean of highly customized models springing up in a much more governed way within enterprises, presenting the single biggest opportunity for all players to benefit from the open data revolution.
Data Analytics-as-a-Service (DAaaS) is a cloud-based delivery model for data analytics & AI. What do you see as the most important enabler for you to adopt DAaaS for your business?
Ease of getting my data into the DAaaS platform9%
Tools that make it easy to create use cases with the DAaaS platform41%
A pre-existing library of dashboards and report templates to help me quickly get up-and-running32%
The ability to try out the DAaaS platform for free before buying10%
Services from the DAaaS vendor (consulting, support, training)3%
This is already happening as part of the evolution of LLM adoption by enterprises. It requires that one think of LLM architecture in “onion” layers. Essentially the innermost layer is the foundational models that LLMs are built on. They include the likes of GPT, Sparrow, LaMda, MT-Nlg, LlaMA, PaLM amongst others. These are the foundation models that are trained on general-purpose data, and later adapted for specific applications such as ChatGPT, Bard, and Midjourney which represent the second layer. Predictions are that the first two layers will consist of few big
players that are able to invest significantly to train these models on mountains of data.
The emerging two layers are industry specific LLMs and enterprise LLMs, where enterprises are starting to leverage the first
two to train their own models at much lower cost as the 3rd and 4th outermost layers respectively (aka SLM). This will see an ocean of highly customized models springing up in a much more governed way within enterprises, presenting the single biggest opportunity for all players to benefit from the open data revolution.