Building on AI: How Much Risk Can You Handle?
Tom Barber (00:00)
Hi folks, just a quick one today as I'm very much on the move, but welcome to the AI briefing podcast. And I was having a think on the fly over of like the issues and the pitfalls that can come with the way that modern internet architecture is architected. For anyone who didn't see the news today, yesterday now, Cloudflare suffered a large
outage caused by some suspicious or erroneous traffic going through their system that took out a large portion of the internet. And of course, this makes me think from an AI perspective, when you've got such a reliance, just as so many customers have on cloud infrastructure, like many of the AI organizations, is what's the backup plan when it doesn't work? When
all the systems go down and you've built a product or you have a large reliance on a specific AI vendor to provide either, know, GPT support or API support or image generation support. What happens? Like what's the risk and what's your appetite for risk when it comes to working in that space? Google
the boss guy from Google today came out, it was similarly talking about the AI bubble and the potential for it bursting. Like as you build these companies that wrap so many AI services, you also have to weigh up your risk appetite for what is acceptable risk when it comes to deploying your entire business strategy around such a hot commodity. Now for a lot of people,
It's fine if you're like super startup and basically you want to create something that's in the AI space, then sure, go ahead and wrap your business strategy around open AI or Anthropic or whoever's model you want to use. if you're
a more stable business long-term been around for a while and you're considering dabbling in AI. If it's also to make it a core part of your business, you have to then weigh up the risk and the way that you want to be able to deal with that. Do you use open AI directly? Do you use the models hosted in Microsoft's AI playground? Do you self-host a model that you can get from one of the open source ones? It depends on what features and functionality you require.
as to what you can do when it comes to actually deploying that stuff. So anyway, just a quick one that was a bit more food for thought than anything groundbreaking, but let me know what you think in the comments and I will see you again tomorrow.
