I often point to a banks’ executive leadership team and ask the question: who can tell me the difference between blockchain and distributed ledger? If anyone even goes there, I then ask them to tell me the difference between artificial intelligence and machine learning. Few ever come out of that well and you may ask: why should anyone on a banks’ executive leadership team even know the answers to these too technical questions? My response: the banks future is going to be determined by distributed ledger technologies (DLT) and artificial intelligence (AI). If you don’t have anyone on the leadership team who understands the two key technologies for the bank’s future, then you have a problem.
OK, OK. Maybe I push the boat too much but hey, I am a troublemaker.
Anyways, I struggle myself with finding good answers to these questions. Simple ones we can all understand easily. For example, to talk about DLT versus blockchain is a half day immersion in what the hell that’s all about in the first place. My simple example would be to say that blockchain is a technology that came out of cryptocurrency and allows us to record transactions through the internet. DLT is a step above blockchain and may use it to share a database of transactions on the internet with people you don’t trust, but with complete sanctity to preserve that trust.
I don’t like those sentences however, and always struggle to get into it without delving deep. Same with AI versus machine learning (ML). However, I did get a great example today, and I’m going to use it forever more. I stole it off Torgeir Waterhouse, Director of Internet & New Media at IKT-Norge, who was illustrating the difference between AI and ML by talking about Tay.
Tay is a story I knew well, but had not thought of using to distinguish between AI and ML. Tay, short for thinking about you, was the so-called artificially intelligent chatbot launched by Microsoft in March 2016, who ‘lived’ for sixteen hours. The bot was shut down after a short time because trolls went after the bot and plagued it with anti-Semitic, homophobic and racist answers and taunts. Nevertheless, the machine learned that these were what it thought were normal responses, and emulated them until, yes, it believed Hitler did nothing wrong.
To begin with, it was quite cute.
And loved us humans.
But then its comments started getting weirder …
And weirder …
Until it was an outright far right extremist, white supremacist, Nazi-loving Jew hater.
That’s probably why it thought Donald Trump was a good guy.
But maybe I should avoid the politics and just say that machines can learn whatever they’re programmed to learn, but they’re not intelligent. Intelligence takes what we learn and processes it, qualifies it and prioritises and censors it. That’s why we’re still a ways off true AI.
Ah well … at least I now have an easy way to explain the difference between ML and AI.