The Zurich release has arrived! Interested in new features and functionalities? Click here for more

Willem
Giga Sage
Giga Sage

Is Gen AI the opportunity of 2024? It's a question that resonates with many as we stand at a new era in artificial intelligence. This post aims to shed light on Gen AI, with a particular focus on Chat GPT, Large Language Models (LLMs), and introducing you to some of the key players in this exciting field.


Chat GPT: A Phenomenon in Two Months

Let's start with an intriguing fact: Chat GPT gained a staggering 100 million users in just two months!  This incredible growth signifies the rapid adoption and widespread recognition of this technology.

 

To put it into perspective, consider how other digital products fared in their journey to 100 million users.

  • Google Translate, for example, took 78 months to achieve this milestone.
  • Instagram accomplished the same feat in 30 months,
  • TikTok did it in just 9 months.

Willem_0-1704386473983.png

 

But what is Chat GPT?

Chat GPT, short for Chat Generative Pre-trained Transformer, is a chatbot interface developed by OpenAI. It's an interface that can provide answers based on your questions. When you engage with Chat GPT, it responds by leveraging the GPT architecture.


Deciphering GPT

But what about the "GPT" in Chat GPT?

It stands for Generative Pre-trained Transformer. It's a deep learning architecture that excels at understanding and generating human language. This architecture relies on a neural network that has been (Pre-)trained on a massive dataset of text, enabling it to generate coherent and contextually relevant text.

Willem_1-1704386473987.png


LLMs in the Market

Large Language Models (LLMs) play a central role in the world of AI. These models are at the forefront of the AI revolution, and several leading companies have harnessed their power. OpenAI, Microsoft, Google, and ServiceNow are currently the most relevant LLMs for the ServiceNow ecosystem. More on this in a later blog and video.

  • ServiceNow has it's domain specific Now LLM,
  • Open AI and Microsoft use GPT,
  • Google has Bard

Willem_2-1704386473993.png

 

ServiceNow has made many acquisitions and partnered with Hugging Face and Nvidia to accelerate their Large Language Model (LLM) development.

 

There are many companies that are developing or have developed LLMs, see below (not an extensive list):

Willem_0-1704392239175.png

Source: https://www.cbinsights.com/research/large-language-model-operations-llmops-market-map/

 

For a quick and high-level insight into this topic, I have also created a short video that complements this article. You can watch it on my YouTube channel.

Comments
DorianK
Tera Expert

Don't forget all the LLM's that are open source and multi modal ones that exist as well 🙂

Naval Singh
Tera Contributor

ServiceNow LLM was built on a specialized version of the 15 billion parameter StarCoder LLM. Giving attribution to the StarCoder alliance and the code base of Hugging Face is a nice thing. Also, the benchmarking data is also available here: https://huggingface.co/blog/starcoder 

FYI, ServiceNow also uses open source for its backend, i.e. MariaDB. https://mariadb.com/resources/customer-stories/servicenow-massive-scale-with-mariadb/ 

Willem
Giga Sage
Giga Sage

Thanks @DorianK

 

You are right; the LLM landscape is huge! I have clarified a bit more why I highlighted the 4 companies and have added a part on other LLMs with a reference that provides more info.

If you have a better source or would like to contribute, I am happy to include and collaborate. Let me know if you are up for that! 🙂

BhupeshG
Tera Guru

There is  below query from our customer.  Can some one advise is it applicable to Virtual agent if we are not using Now assist.

We have checked with the AI COE as the solution appears to use a training model that includes AI, and they did not have this on their approved capabilities list.  It is unclear with the level of documentation provided by the vendor, including the details in the security checklist from what we could find, if it is simple text matching or using some cognitive services.  

 

There is a concern from AI COE that the training model could be using a non-approved LLM engine such as Cloud9 or Bard/Gemini that could introduce responses that could violate our professional business policies.   We had exempted the design review and will need to roll back this exemption based on the model identified by ServiceNow.

Version history
Last update:
‎01-04-2024 10:30 AM
Updated by:
Contributors