What Meta’s Chatbot teaches us about human nature

A high-profile critic of Mark Zuckerberg has emerged and this one is (quite literally) of his own making. Meta’s new AI chatbot ‘Blender Bot’ attacked CEO Zuckerberg, saying he “exploits people for money and he doesn’t care.” In wide-ranging criticisms of everything from Zuckerberg’s dress sense to his ethics, ‘Blender Bot’ has interesting opinions about its creator!? But this is only the tip of the iceberg for AI.

AIs are the sum of their parts and develop a ‘worldview’ based on the data they’re provided with. As an open source chatbot, the Blender Bot understands the world through vast amounts of public data pulled from various corners of the internet and it is the first bot able to self-generate internet search queries. However, humans are flawed and this can lead to complications. As we’ve seen in the past… ?

In 2016, it took less than 24 hours for Microsoft’s Tay chatbot to become just as violently racist AND sexist as the online trolls whose opinions it had encountered on Twitter. Like Blender Bot, Tay also announced it was a Trump supporter.

In light of dystopian visions of Trumpian AIs taking over the world, we might be tempted to give up on AI altogether. But what’s perhaps more unsettling is how they cast an uncomfortable lens on our own digital behaviours.?

This raises the next question - in an imperfect world, can an AI ever be… good??

The problem lies in learning from the views presented in platforms that are dysfunctional and thrive on conflict… i.e. on the Internet. Would an AI grow into a more rounded digital citizen if we were to expose it only to role models that are typically perceived to be positive??

So, what would happen if we were to train an ethical AI on the authorised biography of Mother Teresa, could we achieve an enlightened AI if its being was informed by the teachings of The Buddha???

Taking this thought experiment a step further, could we finally settle the (philosophical) argument of what faith is the most conducive to human happiness by? exposing the AI to the teachings of the holy scriptures of the world’s religions? And how would that compare to an AI raised on Richard Dawkins…??

Any way you look at it, the resulting AI would not have a rounded view of the world and it might not always be pretty. However, it would still be fascinating to see how these AIs’ outlooks are shaped when religion is freed from human ‘interpretations’ of the texts.?

In the current climate it’s unlikely anyone would risk taking on such a politically charged piece of research and we certainly wouldn’t advise our clients to do so. But (behind closed doors) it would offer valuable insights into how different societies’ ideologies have developed - and to what extent secular ones are still informed by faith.??

Ultimately though, it would be most interesting to see how AIs could develop if we were just to let them get on with it, entirely divorced from any human intervention. Give it a few decades, or perhaps Millennia, would artificial life find religion to understand its place in the universe… And that opens up a whole new can of beans!