Good morning, my little cyborgs!
It’s December!! 22 days to go until the big day. I’m aware that this isn’t necessarily what everyone wants to be reminded of, but personally, I am quite excited this year. Indie and I put our tree up a day earlier than planned, on Thursday evening, then watched our favourite Christmas movie, Elf. I do love the cosiness of Christmas time, even though the cold is horrendous. You know what else is horrendous? Other people’s festive enthusiasm. Time to shut up and get on with the week’s AI news.
So, ho ho ho! Here we go…
A positive move from everyone’s favourite Silicon Valley overlord. They may not have eliminated hate speech or violent videos, but they are taking steps to try and prevent suicide.
The PR juggernaut that is Sophia rolls on.
So an AI created a ‘child’ AI (maybe Sophia’s ‘dreams’ may come true). Using reinforcement learning, Google researchers have automated the design of machine learning models. AutoML is a controller neural network for its ‘child’, NASNet, a computer vision tool. And it does it better than humans have been able to manage.
The Google news brings us obliquely on to the latest from our favourite scaremonger, Elon Musk. He reckons our chances of making AI safe are between just 5 and 10%. “There’s a lot of risk in concentration of power,” Elon told his staff at Neuralink. “So, if AGI represents an extreme level of power, should that be controlled by a few people at Google with no oversight?”
My Picks of the Week
So, that’s enough of that for this week. Plenty of reading for you… and the above are all very fascinating, so you’re going to have an enjoyable time this Sunday.
Until next week… so long, suckers!