Steve Ranger’s Post

When you see how many GPUs it takes to train an LLM - 24,000 Nvidia H100s for Llama 3 and they are very definitely not cheap - then you can understand how and why AI is driving Nvidia's revenues. And if that's not enough to keep the chip giant's production lines busy, watch out for the rise of 'AI factories' next, and maybe another, more profound shift that could change computing stacks all over the world. For ITPro I took at look at how Nvidia took the world by storm https://lnkd.in/etUJfdhi

As everyone knows, in any gold rush, it’s the people selling picks and shovels that get rich. In the AI frenzy it’s the GPU companies like Nvidia who are selling the tools to train large language models that are doing well right now. When the generative AI frenzy moves past the hyperscalers into broader enterprise, there's going to be another big wave of spending, too. I spent last week chatting to a lot of Cisco execs and analysts about where the networking giant goes next. Cisco is reckons the next AI wave will be for big businesses like banks, healthcare and automotive – as well as government agencies – to build their own AI infrastructure. And the networking giant thinks that it is well positioned to help. For ITPro I had a look at what this means.

I enjoyed the long conversation behind this interview, which has some great thoughts on the role of the CIO, routes into the tech, how the industry is changing - and why it's always a good idea to ask questions.

The latest set of examples of generative AI are a lot odder than I expected. Already this week OpenAI's latest ChatGPT-4o has been flirting and laughing at dad jokes. Yesterday I watched a video of two AIs trying to charm each other while role-playing a customer service scenario, complete with phrases like ‘Got it’ and ‘Cool’ and ‘Great, thanks’. Now it's Google's turn, and in one of the examples it shows, its new AI manages to come up with a name for band, which it did. That fact that the band members were a) a dog and b) a cuddly toy didn't really seem to register, or matter. So where are we heading? Rather than competing just on the sheer power of the LLMs, tech companies are switching to how these tools they interact with humans. Expect more oddness ahead.

Innovation isn't always about having bags full of money to spend. Sometimes innovation happens because you've got no other option. I had a really fascinating conversation about the drivers of innovation and the reasons for the adoption of new technology this week. Not with a big tech company but with a local council working hard to do more with less. For ComputerWeekly.com - why generative AI is taking off where you might least expect it.

If you can’t measure it you can’t fix it. For ComputerWeekly.com I had a really interesting chat with the Open Data Institute about net zero data sharing and breaking down silos to help us understand the carbon footprint of the built environment https://lnkd.in/eMdApG3i

OK, so tech isn't ever going to be much of a vote winner. And so I doubt there will be much about tech in the political manifestos when they are published. But that's a symptom of a wider problem. For lots of reasons, the political classes have consistently underestimated the potential impact of technology. That means, as a nation, we are more exposed to risks and less open to opportunities than we really should be. For my latest ITPro column I've put together a list of five tech policies the next government should be thinking about. Most of the big challenges the next government will face will have a tech dimension; if the people at the top don’t understand that, they limit their own options to respond.