Trust, bullshit and AI hype

Trust for the day - a nice combo of my trust and bullsh*t strands of work here. Active distrust the best mode for AI!

A useful run through of AI from Gary Marcus for the lay person which makes AI look a bit crap and overhyped, which it is. But I think he underplays the real and current harms of bias, misuse in predicting things it can’t effectively predict, over reliance on bad data and other uses eg autonomous weapons we already see.

But interesting too how if you take the word AI out of all this, it is just a load of examples of companies hyping and lying about what their products can do and not checking properly if they are safe and do what they say they will do. Or whether what they want them to do is even a good idea in the first place. Same old same old! But if they can put the words AI in somehow we and investors are all more likely to get excited or frightened.

So let’s stop getting bamboozled by the AI word and ask for evidence that the products are sound and uses are ethical, just like we would any other new service or product. But most importantly, as he says, make sure we are funding lots of NGOs, academics, think tanks and regulators to help see the hype from reality of technical possibility, navigate the ethics and call out and prevent the bullsh*t tech before it does harm.