As I continue to travel around New Zealand, this week we have a “Things I like/things I don’t like” AI edition, which is copied off Zach Lowe’s NBA column. We will take the next week off.
We start with the things I don’t like, because the things I like are quite nerdy.
Twitters Ad Targeting
I don’t know how Twitter does its targeting, but sometimes it’s downright awful. The other day I was served an ad for Lowe’s 20 volt power tools, which, as a renter in NYC, has almost no relevance to me. In hopes of making their algorithm worse, I clicked on their ad, clicked around on the site and put the most ridiculous item I could think of in my cart, a sander, and then closed the tab. My hopes are that all AI enthusiast, millennial, city dwellers will get served ads with sanders in the future.
Amazon announcing something and then taking forever it release it
In the past, when Amazon had their re:Invent conferences and announced things, they would roll them out very quickly. Well, now its two months after the conference and things like Textract and Time Series Forecasting are still in the “preview” mode.
Also, Amazon announces re:MARS, a conference dedicated to machine learning, robots and space, essentially a huge nerd fest, and then they leave us hanging about how much it costs to attend, what the sessions are, etc. It sounds great, but can I afford to go? Should I call in sick to work to fly to Las Vegas? What illness would I have? Will the sessions be good? Who is going? Some many questions I need answered.
Things I am neutral about
Nvidia’s recent guidance
Nvidia updated its recent financial guidance, which is now 17% lower than previously stated. causing the stock to dive by 13%. Nvidia cited excess mid range inventory, weak demand in China for gaming CPUs, a lower uptake of its new top of the line chips and slower closing of data center deals.
Lot to chew on there.
First, excess capacity of mid range graphics cards is good for people like me, who have been debating buying a GPU instead of relying on the cloud. The bad news is if companies are slow to adopt the newest AI products, overall enthusiasm for AI/ML may be tempering off, which is bad for the long term nature of Cloudy. It could just be a hiccup on Nvidia’s quest for GPU dominance, but we will see.
Things I like – all Google related
I can’t put my finger on it, but for some reason it feels like Google Cloud is just more simple to use than AWS. I first started playing around with GCloud because their GPU instances are almost 50% cheaper than AWS, which at $0.97 vs $0.47 an hour doesn’t sound like much, but adds up quickly when doing large computations.
Google Cloud has this nice feature where you can launch an instance of your terminal right in the browser window. No more managing those pesky .PEM keys with Amazon and connecting through the terminal, you just select, open and go!
And then it opens.
They even have an option to allow you to upload files right to the instance, which is super convenient.
Google has this very nice feature where you can create Python notebooks that are hosted by Google in the cloud free of charge. They even let you use GPUs or TPUs so if you are prototyping a machine learning project, you can use this hardware for free! As long as you have a Google Drive account, even the free version, you can access it. You do get weird errors every once in a while and have to reauth every time you want to link your Google Drive to it, but for the most part, I would recommend.
Watching videos of Waymo’s cars drive
There was an article about Shawn Metz, who was selected to be part of Waymo’s self driving program in Arizona. He posts videos of the car in action, which I find to be strangely interesting…and very boring at the same time.
New Zealand – Yea, its pretty cool