The watch itself for sure not. Garmin Connect will probably introduce AI at one point. But as Garmin is usually lacking behind regarding software, I am quite sure it will not be in the near future.
I'd expect adapted training plans to utilise machine learning at some point in the near future. It'll likely take in more parameters than the current recommendation engine, meaning better tailoring around sleep, resting heart rate etc.
People freak out about AI, but Garmin's use, at least in the short term would mainly involves machine learning and statistical models.
You mean will it feature "generative AI".A lot of apps already use some form of AI to detect activities. I.e. automatically detecting if someone is walking and not swimming, or riding a bicycle, or detecting walking without knowing the location change like indoors, or determining "fitness age" etc.
The best use of AI I could benefit from at the moment would be the strength activity. Once it gets to know your routines and movements it should be better at guessing the exercise and what kind of weight you use. Then you would just have to tweak the values rather than have to input everything manually in Connect as you have to currently.
I think AI is still too energy demanding (many floating point operations) for continuous use on a watch, but propably AI will be starting out analysing the data off-line (phone or pc), and perhaps provide even better reports. I actually think Garmin's software is pretty neat, efficient and tends to focus on what is relevant.
To have AI device has to either have a constant connectivity to a server where it can delegate processing or have a powerful CPU on its own. None of that applies to the current generation of Fenix in order to save the battery power.
To have AI device has to either have a constant connectivity to a server where it can delegate processing or have a powerful CPU on its own
Maybe in the case of generative AI, but machine learning can be used to make inferences against a small model pushed down to the watch via Connect. This is technically under the umbrella of "AI". The very same CPU we current have could be used.
An example of a simple model could be training recommendations based upon a < 20 parameter model which includes sleep, HRV, chronic load etc. In fact, they may already be doing this rather than explicitly writing out all of that logic in code.
So we want to take a teeny tiny processor with tiny amounts of addressable memory thats locked in a water tight container thats already dealing with HR, accelerometer data, updating the watch face etc and add a cpu intensive process? At least people will have something new to complain about.
