Dipping a toe in the Udacity pool:
I received an email from the UK/Europe Business Growth Manager of Udacity, looking for people with knowledge of/experiencein Machine Learning to lead and/or support face-to-face training sessions in Machine Learning in London. The email included the link to Udacity’s Nano Degree in Machine Learning.
While I’m not yet skilled or experienced enough at this time to lead a session, I felt really excited that I might have an opportunity to be a session supporter and assistant. That would be cool as I’d be exploring and networking in such an important growth area in tech skills, as well as supporting the growth of MOOCs here in London, which is something I feel really passionately about.
Why so passionate about MOOCs?
Well, firstly as someone who always struggled with learning in traditional settings/formats (it’s not the academic level that’s the problem, I just got so easily bored by the trad. education system), I really value the opportunity MOOCs are giving me to learn in highly engaging, multi-media formats and hands-on project-based learning. I also really appreciate the financial accessibility that MOOCs offer. But above all, I really believe that this kind of focussed, personalised learning is the future of education – especially since we now live in a world (and an economy) where techonolgy and business environments are rapidly changing, forcing us to be contually learning and growing and developing/extending our skillsets, our experiences, our persepctives.
MOOCs – especially those which prioritise practical/project-based and community aspects to learning – are, I believe, the future of accessible, affordable education for those of us who want to be nimble, flexible and forward-looking in our careers.
So why Udacity?
That email triggered it and I went online to check out their nano-degree in ML. I hadn’t thgouht to do so before, assuming this ‘nano-degree’ thing was just a marketing ploy, just a fancy buzzword. I was wrong! The syllabus of the nano-degree is one based almost entirely on projects, some as provided by the course, others which the student picks and develops themselves.
Based on how I approached the Python programming course from Coursera – which I heavily extended by setting and working on my own suite of projects as I went along – I can definitely see how I can benefit from the Udacity nano-degree. And any areas where I feel I might be lacking in direct knowledge, they provide a whole raft of free courses to revise and reskill as you go along.
The platform is neat, clean, easy to use and – on first inspections, at any rate – looks like it will be the best of the three MOOC platforms I’ve so far experienced (the other two being Coursera and Udemy).
What knowledge will I need to start the Machine Learning nano degree?
I’ll require basic knowledge of maths and stats, like calculus, statistics and linear algebra (here’s some basic revision I did earlier). I’ll need to be able to calculate standard deviations, derivatives and matrix determinants. I’ll also need to have intermediate level Python programming skills. The maths and stats are fine – and I will be doing some revision and recaps as I go along (similar to my post on simple linear regression here). Maybe my Python programming is not fully up to scratch yet, but hey, I am pretty confident I will be able to get through by Gooling, searching StackOverflow, and so on. And I’m really up for the challenge of moving on to the next phase of my Personalised Training Plan.
What is Machine Learning in a nutshell:
Machine Learning can be thought of in a number of different ways, or looked at from a number of angles. Firstly – and most commonly viewed and spoken about these days – it’s a branch of Aritifical Intelligence which seeks to address computational requirements for handling large sets of data. Key to this is the idea that the computer ‘teaches itself’ as it goes along, rather than relying on a human programmer to have pre-programmed all decision-making into a program up-front. In this way, the computer is being programmed to act as an agent which can process and understand data, and react intelligently to that data, as it goes along.
But machine learning can also be thought of as a sub-set of Data Science, as a form of applied statistics – applying advanced formulas to compare, analyse and summarise data – and built on a set of advanced algoritms which elegantly handle such large volumes of data input, processing and output. As we’ve mentioned in a number of different posts, the time for machine learning (and data science) is very much now, due to a number of factors. Firstly, computer processing speeds have increased significantly at costs which are far more accessible to most. Second, the availability of large data sets has grown since so much data these days is being collected automatically as more of our personal, business and financial interactions are now taking place online. Third, the cost of storing all that data has fallen markedly (both on local servers and in the cloud).
Machine learning is particularly useful in predicting outcomes, so has many applications across many different domains, both business and public sector. In finance, it can be used to predict stock prices to assist in investment decisions. In business, it is useful for predicting customer actions and buying patterns, thus helping businesses to optimise sales. In education (an area of particular interest to me), it can be used to build intelligent agents to provide personalisied learning support to students (as well as – in larger, more traditional institutions – predict which students are expected to do best or worst in any given class or year).
It’s also worth bearing in mind there are a number of different approaches to AI and machine learning, some more algorithm- or statistics-oriented (such as supervised learning, reinforcement learning, representation learning) while others follow a more cognitive or neuro-based approaches to research and practice (e.g. artifical neural networks, deep learning).
Why my interest in Machine Learning?
One of the main reasons I got interested in Machine Learning, and indeed in Artificial Intelligence generally, is the threat that all our jobs will be taken over by machines in 20 years’ time. While this debate rages on and on as to whether a given job or profession will be at risk from AI or not (e.g. “Will AI/ML replace accoutnants”), I note that the technological innovations and advances continue apace. Debaters and the opinionated ‘educated masses’ (often those with the most to lose by accepting it’s going to happen, so stubbornly refusing to admit it) don’t seem to get the point. While others of us (I’m in this latter camp) accept the inevitability and step up to the challenge and excitement this new wave brings.*
I want to know: how can I be a part of the change, rather than bury my head in the sand and pretend it’s never going to happen (until the day comes when it’s too late to do anything about it). So I start to learn and grow my skills and knowledge. Then, in doing so, I realise there’s a whole range of other applications for this technology that I hadn’t previously thought about, which turn out to be even more interesting to me. And so it goes…
Googling accountancy + machine learning brings up that Quora post linked above, and the viewpoints of a whole bunch of accountants convinced they’re far too skilled and clever to ever be replaced by computers, “ERP systems haven’t been able to do our jobs in the last 30 years so there’s not reason to think they’ll be able to do them in the next 30 years”, and so on…
I beg to differ (see the footnotes for a little more on this). We’ve already seen applications coming through in the legal profession, and we’re starting to see applications now within accountancy. Given my experiences this year in helping to design/implement a full-scale integrated business and finance ERP system for an international SME, I can see how ERP systems (certainly at the mid-market level) are not technologically advanced at all (they’re basically just pootling along on the back of 1980’s/1990’s technology) and are definitely due a significant refresh!
So while my key interest in this field is education (particularly a highly-personalised, accessible/affordable version) there’s also plenty of progress that can be made in the accountancy/finance and business systems space. So I would never rule out getting more into that arena too. I’m really excited about all the possibilities!
* Footnotes (some background on the expected loss of accountancy jobs/profession to machines):
It’s worth reflecting on the significant academic study conducted by the Oxford Martin School of Oxford University in 2013 which ranks Accountants, Auditors and Budget Analysts all as having a 94% probability of being replaced by machines in the not-too-distant. (Just for clarity, yes that does mean the highly-qualified ones who continue to think they’re immune to all this.) Lower-level roles such as cashiers and payroll clerks (97% probability), book-keepers and accounts clerks (98%) and tax preparers (99%) fare even worse. You can access the report as PDF here; check the Appendix for the full list of jobs and probabilities.
The UK government also conducted similar research (published early 2014) which can be accessed as PDF here. They predict Accountancy will be one of the first professions to feel the change (others being the insurance industry and the legal profession), with an expectation that both law and accountancy firms will “shed a significant percentage of their highly skilled workforces”. Their timeline for this expected change is throughout the 2020’s, with these professions and firms having been significantly disrupted (and down-sized) by 2030.
Read more like this:
This post kicks off the Machine Learning stream of my Personalised Training Plan. More posts in the series will be available under this tag:http://deborahroberts.info/tag/machine-learning/.
I’ll be recording my responses to and learnings from Udacity’s Machine Learning nano-degree program (linked below) which is presented in partnership with Google and the Machine Learning faculty of Georgia Tech as well as other research I do and projects I undertake in this field.
References:
Machine Learning Engineering by Google nano-degree program on Udacity