This week I was at HP’s Amplify Partner Conference. It’s one of my favorite events because I’ve worked on partner programs before and HP’s impresses because it has replaced EMC’s as the best currently in market. This year, HP made a number of interesting announcements, including that it is taking product leadership in AI PCs and workstations. Intel had the most design wins, but those based on AMD’s processors, according to HP, provided 30% higher performance, and mobile workstations had significantly better battery life than Intel workstations. This is important when you consider that all LLMs (Large Language Models) are first created on workstations.
But what really caught my eye was a software offering from HP called AI Studio that was created with NVIDIA. AI Studio corrects what has been a huge gap in the generative AI creation process. AI Studio uses AI to manage permissions and responsibilities between team members building generative AI models.
Given it costs around $100M to create an LLM, anything that makes the effort better, faster or cheaper (let alone all three) would be incredibly valuable if it were uniquely useful, and AI Studio appears to be just such an offering.
Let me explain.
What Makes AI Studio Unique
Up until these recent HP announcements, there were just two AI leaders providing complete AI solutions in the market: IBM and NVIDIA. Both companies started on AI decades ago and have increasingly collaborated with each other to mutually advance their offerings. Before these latest HP workstations, NVIDIA even had to build its own unique workstation offerings because no one else wanted to, and IBM was out of the workstation business having sold it to Lenovo.
But LLMs are incredibly complex and have unique security and collaborative requirements that make traditional software management offerings inadequate for the task. And even though HP is the most aggressive company in its class about deploying AI internally, it was still new to AI, and it isn’t known as a software company (an image it wishes to change by way of Jim Nottingham’s Advanced Compute Solutions effort).
So HP partnered with NVIDIA, one of the two companies that had the knowledge of how to create a collaboration and coordination tool for AI. With NVIDIA’s help, HP created AI Studio, backing it with an AI engine that automates the underlying process because you want your very expensive data scientists focused on the project, not the paperwork, which I expect they hate as much as managing the related effort and reporting progress.
I’m not a data scientist, so I’m relying on Jensen Huang’s advocacy and support for this product as my reference point for this offering, including its uniqueness in the market, because I doubt there is any other CEO more versed in what is needed (given, he has had to build his own custom hardware in the past to create AI tools) and what is missing.
Reported benefits for this tool include:
- Fewer false starts by forcing early rigor into the process and assuring adequate planning before the actual creation of the LLM, including responsibilities and well defined permissions (so that people who shouldn’t be mucking with the LLM won’t be).
- Fewer iterations to correct mistakes or efforts to align the project with changing requirements by forcing the requirement process up front while forcing collaboration, management and creation process control over the effort.
- Integrated components that allow the users to select, manage and validate the LLMs being created.
- Integration with related AI-focused workstations that are designed to build LLMs (NVIDIA once had to build these itself because no OEM wanted to build them for NVIDIA until HP).
- A related AI Certification program for HP’s sales and services channel to assure those in this channel know how to train others to build generative AI models.
While I doubt this will be the last tool of its type, it is the first, and backed by NVIDIA, it should be the best for some time to come given how few seem to fully understand or embrace AIs yet.
Wrapping Up:
Until I saw AI Studio, I didn’t realize it was a missing link in the AI creation process. But now that I’ve seen it in action and confirmed NVIDIA’s participation in the tool’s creation process, I believe HP’s AI Studio currently stands alone when it comes to coordinating data scientists and engineers working on a generative AI creation project. Since these projects tend to cost in the $100M range, anything that makes them better, faster and cheaper should pay for itself. It is brand new, however, and the training and certification programs just kicked off, so as with all new tools, this one may need some time to fully mature. When it does, I expect it will be a game changer.
