How HPE can get you started with AI

We talked to two of HPE’s experts about how the growing use of AI solutions is fuelling demand for the vendor’s HPC starter packs.

A growing number of organisations are looking to use AI as a way of accelerating digital transformation and getting a competitive advantage. But typically, the people who want to deploy such solutions are not in the IT department. They are often data scientists or other experts and while they might have a solution they want to use, they expect IT to provide an appropriate platform.

Martin O’Sullivan, HPC and AI Business Development Manager, HPE

More often than not, the IT department does not know where to start. This is where the challenge for the customer starts and the opportunity for HPE partners begins. As Martin O’Sullivan, HPC and AI business development manager at HPE explained: ‘There is a disconnect between the data scientists and IT and that’s where we are presenting our HPC Starter-Kits as the ideal solution.’

Customers are starting to take AI seriously because it can make a real difference to business outcomes, said Matt Armstrong-Barnes, chief technologist for AI at HPE.

‘It’s been projected that around 75% of all digital transformation projects this year will include some form of AI. Today, we produce around 1.2 quintillion bytes of data a day, and we are getting to the point where traditional, rules-based approaches to tackling big data challenges have run out of steam. We are having to bring in new techniques to address the massive amount of inbound data and help organisations extract value from it.’

AI as a team sport

Matt Armstrong-Barnes, Chief Technologist for AI, HPE

There are a couple of key points that everyone needs to understand about AI, he stressed. The first is that no one organisation can deliver it in isolation. ‘AI is a team sport. The only way to be successful is through partnership – between suppliers and the customer and you need to find partners you can work with long-term.’ This is important he stressed, because most AI projects evolve and develop over time, and you need to be able to go back in and update solutions.

‘There is an old Africans proverb that certainly applies to AI: “If you want to go fast, go alone. If you want to go far, go together”. You can make some quick advances with AI on your own. But if you bring in partners, you are going to go much further.’

The collection of partners needed varies. Clearly, the customer and/or a specialist partner will understand the data set and the business problem. Then, someone has to design an AI solution that addresses the challenge – and that’s clearly specialist work. Only then will the platform come into consideration. But the choice made at this point will be critical to the immediate and long-term success of AI projects.

Hyper parameter optimisation

The other important thing to understand about AI is something called ‘hyper-parameter optimisation’ Matt told us. ‘If you think about AI being a black box, and there are a collection of knobs and dials on this box that make it better, faster and more accurate – and improve the ‘F1 scoring’ [a way of measuring the effectiveness of machine learning]. It can be very complicated and difficult to get those settings right.’

Once companies start on their AI journey, their competitors will never catch them up

It's important to get the parameters just right in order to maximise efficiency. But on a server platform that is not tuned for the task, it can take a week of running an AI solution before you can tell if a particular set-up is really working or not.

This is where a lot of projects start to flounder. There are two possible solutions, said Matt Armstrong-Barnes. ‘You can bring in real experts and computer scientists who know exactly what to do, or you can offload the challenge and put in place an infrastructure that will give you a very swift answer, so you can experiment and tune settings much quicker.’

Effectively, this is what HPE is providing with the HPC Starter-Kits. They have been designed for customers who are perhaps venturing into AI for the first time, and to enable them to start small and then scale up as required.

Not a “one and done”

Scalability is important, said Matt, because once organisations set out on the AI path, they tend to continue. ‘AI is not a “one and done”. Once companies start on their AI journey, their competitors will never catch them up. As long as they stay on that path, they will always stay ahead. Also, once they start, they realise that it’s not actually as difficult as they thought it was going to be – and once you have built one AI [solution], you will soon build another.’

HPE and TD SYNNEX can help you to find the right partners

Having the right partnerships in place is key though and will help you to overcome challenges and pitfalls you are certain to encounter. Here too, HPE and TD SYNNEX can help you to find the right partners. HPE has established relationships with many ISVs and HPE Pointnext services can be applied to the AI solutions as well.

With the pre-defined HPE / NVIDIA starter kits providing the foundation platform. They have been designed and come with the services that are ready to run, so partners can depend upon them entirely, said Martin O’Sullivan. ‘They come with all the services and software, and they have been pre-validated, tried and tested to deliver the most secure, scalable and performant solutions at point of delivery.’

Putting yourself in the frame

The vendors have created these solutions now in response to demand – and it’s important for partners to put themselves into the AI frame with their customers, said Martin.

Our Starter-Kits come with services elements built-in and are ready to deploy

‘IT departments may well reach out to their trusted partner for AI solutions. But a lot of the time, partners are not actively demonstrating that they have an AI capability and customers will look around for other partners who do have an involvement in AI. Our Starter-Kits come with services elements built-in and are ready to deploy, so partners can pro-actively offer them to their customers, even if they don’t initially have the knowledge or skill-set.’

That’s key said Matt Armstrong-Barnes because partners need to be ready. AI projects usually have board-level approval, and the budget will be there to acquire the necessary platform to run them. Partners must be ready to meet that need. With the HPE / NVIDIA Starter-Kits they can be – and the TD SYNNEX team are geared-up and ready to support partners in learning about the HPE’s AI solutions and taking them to market.