The Sunday AI & ML Review
Welcome to this week’s review of the latest news in Artificial Intelligence and Machine Learning.
It’s been a busy week at TDMB (nothing new there!) so I was pleased to let off a bit of steam with my family last night. We took my daughter to the arcades and bagged a treasure trove of tat from the 2p coin pusher machines. Then we all went for pizza before coming back to mum’s to watch a movie… guess what film it was? Well, AI, of course!
And on that note, it’s high time we got on to working through the top five socially-shared news from the world of artificial intelligence this week. So, here we go…
This week’s most socially-shared article comes from Wired, and it’s a really cool one.
Rather than layering sounds of different instruments, this machine learning algorithm produces entirely new sounds using the mathematical characteristics of the notes that emerge from those instruments, creating sounds that have never been played before… ever. The software is called NSynth, developed by Jesse Engel and Cinjon Renick at Google Magenta. Though the idea in itself isn’t new (orchestral conductors have worked with a similar approach for a long time), the way NSynth gets it done and the results it yields could take music to new places, developing music beyond what we know. Find out more on the Google blog.
Google’s new AutoML project is developing an AI that can assist in creating other AIs. The idea is to make machine learning more accessible to non-experts. So, as we know, deep learning is kinda complicated, involving layers and layers of neural networks. So, Google thought, why not create an AI that could do it for them? Why not, indeed. As Google explained on their blog:
“If we succeed, we think this can inspire new types of neural nets and make it possible for non-experts to create neural nets tailored to their particular needs, allowing machine learning to have a greater impact to everyone.”
Futurism strikes again, this time talking about OpenAI, the non-profit AI co-founded by Elon Musk. So, OpenAI has announced that they have made an AI system that can learn to complete a task in real life after seeing a demonstration of the task in a simulation. Basically, this means the system can successfully replicate human behaviour – but learn even quicker. Sound intriguing? Read more here.
A new Google cloud computing service is coming, offering exclusive access to a new AI chip designed by Google’s own engineers. Before the end of 2017, any business or developer will be able to build and operate software accessed online, via the new Google cloud service (named TPU 2.0 or the Cloud TPU), that taps into hundreds or even thousands of these new Google processors. Currently, the dominant Silicon Valley chip maker is nVidia, and they must be feeling a little worried that the Google megalith is stepping up to compete.
Google, Google, everywhere. And here it is again. That’s four out of five of the top socially-shared articles of the week focusing on Google. I kinda like this, in a ‘bring-on-the-robo-pocalypse’ sort of way. The crux of the article is this: Tech giants (like Google) are the ones building our artificially intelligent future – not the government. Yes, our governments are being left behind; how do you fancy being governed by Google? Can’t be much worse than it is now, right? – or could it…?
So that’s the top 5 socially shared. Now, here’s my pick from my inbox this week:
- Home Monitors Are Getting Smarter (and Creepier) – (MIT)
- If we’re living in a simulation, this UK startup probably built it – (Wired)
- Intelligent Machines: An AI Ally to Combat Bullying in Virtual Worlds (MIT)
- Perpetuating Bias: Why We Should Think Critically About Artificial Intelligence in Marketing (Skyword)
- The Relentless Pace of Automation (MIT)
- The Surprising Repercussions of Making AI Assistants Sound Human (Wired)
- Nvidia CEO: Software Is Eating the World, but AI Is Going to Eat Software (MIT)
- Durham Police AI to help with custody decisions (BBC)
- Why Google’s CEO Is Excited About Automating Artificial Intelligence (MIT)
I’ve been having a few conversations recently about the difference between artificial intelligence and machine learning, including a chat this week with a Doctor in Computer Science who was adamant that the two phrases were pretty much interchangeable. I’m not sure… what do you think? Answers on a tweet to @msmichelebaker.