During The Watch Scholar Authentication courses, the questions I most get asked are, ‘Can’t AI do this?’ or ‘Surely AI would know the difference immediately?’. It was only when we broke down what would be necessary to train an AI on a single model reference from a single brand that the scale of the task became clear.

Remember the Metaverse? We were all going to work, shop, and socialise in Mark Zuckerberg’s virtual world. $36 billion later, this disruptive technology is officially dead.

Web3? Its plan to overturn the online giants and decentralise the web to give us back ‘control’ is now limited to Crypto and some random supply-chain applications.

Now AI is hyped as either humanity’s next evolutionary step or an existential threat to us all, and yet, so far, our day-to-day experience seems limited to annoying chatbots, quirky but unrealistic images, and bland, inaccurate copywriting, served with a side order of ecological disaster.

More money has been invested in LLMs and AI since 2022 than in any other technology…ever, and yet the payoff is always delayed. ‘It is hoped that’, ‘Is expected to…’, and ‘May one day…’ are phrases found in almost every AI-related press release, and yet the common public opinion seems to be that AI is already an omniscient tool for everything.

Going back to Watch Authentication, the practical challenges of making such a model suitable for public interaction are equally steep, and, given that it can be easily tricked and that the whole process lacks transparency, the results cannot be trusted.

Students were left with a greater appreciation of what the human eye and brain, coupled with knowledge and understanding, can achieve, and that the skills they were beginning to develop would be safe from AI replacement for many years to come, if not forever.