AI better than dermatologists at detecting skin cancer

Hello Future Curious Folks,

It’s been a while since my last post. Life and travel gets in the way sometimes, and sometimes a bit of time at yet-another-airport-lounge is the perfect time for some future gazing.

Today I’m going to do one more AI post, before moving onto new topics. Whilst AI is a huge field with many implications, one of the most immediately striking is the area of image recognition and the massive strides we’ve taken in this area over the last years. As always, this emerging powerful capability has potential to drive massive benefit and also gives us some potential warnings.

As a first example, AI accuracy is already matching and often beating diagnosis accuracy in pathology, radiology, and skin cancer. I’ve referenced this in the past, and the capabilities continue to progress at pace.

What does this mean? Certainly I would question the choice of specialisms for anyone training towards these impacted careers. There still seems to be a need for seasoned professionals to complete tasks beyond the ‘narrow’ diagnostic process, however this clearly needs a lot of thought.

Looking a little further into the future, I wonder how quickly we can move to fully automated self-serve systems for basic diagnosis. Radiology and pathology are tricky due to the process of gaining the sample in the first place, however it’s fairly easy to imagine uploading images of moles / freckles and having an AI agent diagnose these at scale and at negligible cost. I would assume that even radiology and pathology sampling could be automated over a longer period, and AI could of course be used for data other than images. I can see a future health screening machine / booth that gathers a full range of samples for screening and analysis without human intervention, for example.

The exact same kind of technology is being used for facial recognition and vehicle recognition. On the positive side, there are some clearly positive security outcomes available. Unfortunately, one can also envisage Orwellian abuses, with restrictive governments exerting iron-clad control over citizens.

In the linked examples below, you’ll see facial recognition being used to monitor immigration ports. You’ll also see car recognition moving far beyond number plates and towards full make, model, decals, features (e.g. roof racks), etc. These are both very detailed and complex problems that are only now able to be solved via advancements in AI.

Fast forwarding in time, it’s clear that tracking the movement of individuals will become immensely simple. I can see identity detection being blended with millions of encrypted personal data points that can be verified, shared, or blocked from the various machines, services, and people around us (for an example, look up ‘Self-sovereign identity’). This means that we can live in a world that dynamically adapts and reacts to us as an individual, opening up a huge array of future services & capabilities.

We can look forward to completely seamless international travel with no immigration stops at all, self-driving car fleets that already know your destination and preferences without saying a word, rooms and offices that automatically adapt to personal needs, highly personalised information at every step of our journey, and meals that are automatically tailored to your taste and health needs at that moment in time.

All of this would be possible with no need to provide any active instructions at all … I wonder what we will do with all of our spare time?