Matisyahu's Blog

I was watching the following video talking about AI, LLMs, the research occurring at the big technology names regarding their plans on going beyond LLMs to develop new ways of implementing AI that address the short comings of LLMs:

I am reminded of Microsoft making the decision many years go to exit from markets they tried to compete in but failed with the most high profile being their purchase of Nokia and investment into Windows Phone (then later Windows Mobile) which was later unwound by the new CEO Satya Nadella. The reason? they were spending billions and getting very little in terms of marketshare as a result (the reason for that failure is a whole new blog post its own right which I may write in the future) – an acceptance of the fact that they were late to the market and the idea of throwing money at something that produced very little in the way of return on investment made little sense.

What was done instead? Microsoft pivoted and instead focused on what they’re good at which is their cloud services, their productivity software an do so on. It made sense to go where the ball is – the platforms of Android and iOS were more or less a full gone conclusion so why not spend the time instead investing where it made sense which is in the enterprise and education then making sure that both Android and iOS got the best experience when using Microsoft software and services. Don’t focus on fighting battles that were lost long ago but instead focus on future opportunities instead.

What this reminds me of where Apple is when it comes to AI development in context to the above video. There are rumours that Apple is going to use licence Gemini from Google and run it on their own hardware but there is much talk about it being something that they’ll pay per year for with the idea that Apple will eventually move to their own in-house LLM some time in the future. Personally I am not convinced that Apple will create their own LLM for the same reason why Microsoft walked away from Windows on the mobile – the battle was already lost so instead the focus should be on future opportunities.

In context to the above video does it make sense for Apple to invest time and resources into their own LLM or are they better of spending the resources on developing a model that moves beyond LLMs. There are already plenty of discussions about creating models that are far better suited for what Apple want to achieve with Siri so maybe that is the long term goal – keep paying Google so then they have something that isn’t behind the competition all the while working on something better than the current LLMs out there.

Regarding the long term view of AI, I think we’re going to find that as we hit the limitation of LLMs that the hype will subside and those that will be successful are those companies that a) have a extensive portfolio that they can integrate their AI offers into and b) have custom hardware that is tuned to the absolute maximum efficiency possible so that the cost of delivering those AI enhanced services are a rounding error in the grand scheme of things (to the point that they can bundle it with their product offerings at no extra cost and thus making their products more competitive when compared to the competition) c) are able to push beyond the existing models to something even better.