Here we are on Day Three of our Autumn/Fall Sale, which started on Monday morning and ends sometime during the night this coming Sunday.
So far, approximately fifty club members have bought online Italian lesson credits at 20% off the usual price!
But sale weeks (we do four a year) aren’t so interesting here at the club. There’s quite a lot of routine work involved – sending orders, customer service, and so on.
Which means a lot of hanging around waiting for something to come in. And less time for the fun stuff, like editing ebooks and doing my own language-learning.
Ho hum, at least I can get through the backlog of emails which, when they first arrived, I rated ‘non-critical but interesting’ .
This morning, for instance, I had one from Amazon Web Services (who own the servers on which we store the audio recordings for EasyItalianNews.com).
Would I like training in ‘Artificial Intelligence’, how to get machines to learn, that sort of thing?
For free? No previous experience required?
Say one thing for Amazon, they know how to do marketing. But no, I think that this week I won’t have time to teach myself machine learning strategies.
However, it did get me thinking…
From what little I already know about the subject, we’re basically talking about algorithms (programs) which are then tuned/adjusted/perfected with practice over time.
So, for example, you want to get your system sorting images from your CCTV system, so as to distinguish criminals, terrorists and so on from the non-threatening citizens who are just going about their daily business?
First you need to program in likely threats – anyone carrying a weapon, for example, speeding vehicles in pedestrian areas, people with brown skins and bushy beards (that’s a joke…)
Next you test your algorithm with a set of data you already own and understand – say a few months worth of images from an area where you know suspect activity was taking place.
Can your A.I. system pick out the bad guys and gals from the ordinary shoppers?
Likely it’ll get some of the black hats but miss a few – the lady with the nuclear device stuffed into the pram she was pushing somehow slipped by, as did the guy wearing a smart business suit and carrying an AK47.
More problematically, the system will likely have identified a bunch of innocents as people of concern.
Thousands of people passed in front of the CCTV cameras each day, virtually none of them malintentioned. It’s therefore inevitable that there’ll be plenty of false positives, that is to say, cases in which a threat is identified where no threat actually existed.
Tweak the algorithm a little and feed the data in again, perhaps you’ll get better results? If not, then you’ll need MORE DATA.
In fact, here’s where the biggies of the Internet (Amazon, Google, Facebook, etc.) have a huge advantage – they already have stacks of data, and so can train their computers more effectively than the little guy (or charge the minnows for access to the data…)
Beh anyway, eventually you may have a system that can spot threats from CCTV images more cheaply and more effectively than police officers could. Or not. Who’s to say what’s possible these days and what’s hype?
But, the basic process – start with a trial algorithm, expose it to data, check the results, tinker some, try again – that’s not just how machines learn, is it?
That’s how you learn, too.
People (especially Italian people) seem to think that the process of ‘teaching’ is predominently one of ‘explaining’, with a little ‘testing’ for good measure.
Personally I’ve always found it to be mostly about getting the right materials and activities in front of the students at the right time, in a suitable sequence, and in sufficient quantity.
If the teacher doesn’t do anything obviously WRONG, while providing sufficient opportunities for learning, it’s likely the students will do so.
With or without explanations and testing.
In fact, forget the teacher. As long as the student has access to materials/learning opportunities which are (keyword coming) ‘appropriate’ to her stage of learning, then assuming she makes time to exploit them, she will learn.
Just as A.I. algorithms get better when exposed to bigger data sets and given feedback on their mistakes, so do people.
It’s all about the INPUT.
Imagine you’re sitting on a desert island with no internet, no radio, no books and no shipwrecked companions. Of course, you decide, well hey, just to pass the time, I think I’ll teach myself Italian!
Genial idea as Italians say. But there’s a flaw…
With no input at all, not even a soggy, salt-encrusted grammar book, you’re clearly not going to get very far, are you?
What WOULD make a difference to our lonely Italophile?
He could probably learn a lot from a book, given enough time and absolutely nothing else to do.
Or imagine he had a radio!
Better? A coursebook, with audio recordings, and a good supply of coconuts?
He’d be sorted.
The years would pass and, having no other distractions, he’d get more and more able with the language.
The human brain is a wondrous thing.
Unlike a computer, you and I have our language-learning algorithms pre-installed.
All we need to learn is sufficient input and the motivation and time to process it.
Our team of writers and editors carefully craft them so they are ‘appropriate’ for learners at different levels.
Students get text input, and audio too!
We even try to make them interesting and fun.
Autumn/Fall Sale info.
Don’t forget, everything in our online shop is a fifth cheaper this week!
Use this coupon code to get your discount: Autumn-Sale-2019
Make your selections of ebooks and online lesson credits from our online shop.
- Coupon code Autumn-Sale-2019 will reduce your cart total by 20%!
- It’s valid until midnight on Sunday 29th of September 2019
- Spend as much or as little as you wish
- The coupon code can be used multiple times during the offer period
- And it’s valid for items which are already discounted, for example ebook multipacks
- N.b. the coupon code cannot be used cumulatively with other coupons
- Secure payment systems include Amazon and Paypal