That wasn’t Google I/O — it was Google AI

on

|

views

and

comments


Three years in the past, the corporate pressured out Timnit Gebru, the co-lead of its moral AI crew, primarily over a paper that raised issues in regards to the risks of huge language fashions. Gebru’s issues have since develop into mainstream. Her departure, and the fallout from it, marked a turning level within the dialog in regards to the risks of unchecked AI. One would hope Google discovered from it; from her. 

After which, simply final week, Geoffrey Hinton introduced he was stepping down from Google, largely so he’d be free to sound the alarm bell in regards to the dire penalties of speedy developments in AI that he fears might quickly allow it to surpass human intelligence. (Or, as Hinton put it, it’s “fairly conceivable that humanity is only a passing section within the evolution of intelligence.”) 

And so, I/O yesterday was a far cry from the occasion in 2018, when the corporate gleefully demonstrated Duplex, showcasing how Google Assistant might make automated calls to small companies with out ever letting the individuals on these calls know they had been interacting with an AI. It was an unbelievable demo. And one which made very many individuals deeply uneasy.

Repeatedly at this 12 months’s I/O, we heard about duty. James Manyika, who leads the corporate’s know-how and society program, opened by speaking in regards to the wonders AI has wrought, notably round protein folding, however was fast to transition to the methods the corporate is considering misinformation, noting how it might watermark generated pictures and alluding to guardrails to forestall their misuse. 

There was a demo of how Google can deploy picture provenance to counter misinformation, debunking a picture search successfully by exhibiting the primary time it (within the instance on stage, a pretend photograph purporting to indicate that the moon touchdown was a hoax) was listed. It was a bit of little bit of grounding amidst all of the awe and marvel, working at scale. 

After which … on to the telephones. The brand new Google Pixel Fold scored the largest applause line of the day. Folks like devices.

The telephone might fold, however for me it was among the many least mind-bending issues I noticed all day. And in my head, I stored returning to one of many earliest examples we noticed: a photograph of a girl standing in entrance of some hills and a waterfall

Magic Editor erased her backpack strap. Cool! And it additionally made the cloudy sky look much more blue. Reinforcing this, in one other instance—this time with a baby sitting on a bench holding balloons—Magic Editor as soon as once more made the day brighter after which adjusted all of the lighting within the photographs so the sunshine would look extra pure. Extra actual than actual.

How far can we need to go right here? What’s the top aim we’re aiming for? In the end, can we simply skip the holiday altogether and generate some fairly, fairly footage? Can we supplant our reminiscences with sunnier, extra idealized variations of the previous? Are we making actuality higher? Is all the things extra stunning? Is all the things higher? Is that this all very, very cool? Or one thing else? One thing we haven’t realized but?

Share this
Tags

Must-read

Nvidia CEO reveals new ‘reasoning’ AI tech for self-driving vehicles | Nvidia

The billionaire boss of the chipmaker Nvidia, Jensen Huang, has unveiled new AI know-how that he says will assist self-driving vehicles assume like...

Tesla publishes analyst forecasts suggesting gross sales set to fall | Tesla

Tesla has taken the weird step of publishing gross sales forecasts that recommend 2025 deliveries might be decrease than anticipated and future years’...

5 tech tendencies we’ll be watching in 2026 | Expertise

Hi there, and welcome to TechScape. I’m your host, Blake Montgomery, wishing you a cheerful New Yr’s Eve full of cheer, champagne and...

Recent articles

More like this

LEAVE A REPLY

Please enter your comment!
Please enter your name here