BadgeIsBack

I agree that the science has been around for awhile and that we need PMs, and anything the demand will increase.

As for dishing out work, yes, :-) but the Pm is still accountable. In that role, I've done a few late nights making sure it's all covered when a project team member could have done it. Don't mind backing the people though. There is still an art form to it.

Beoz

Lets say you do decrease the need for PM's. Lets say today 6 out of 10 automation and tech projects need a PM. Lets say tomorrow 5 out of 10 need a PM because some form of automation kills one PM job.

The thing the luddites forget is tomorrow there won't be 10 auto and tech projects. There will be 50.

That's 25 PM's needed.

BadgeIsBack

You are v tech minded. Programme mgrs can look at traffic light colours and earned value but still need to approach the PM to ask what's going on. Remedial action is still manual.Dashboards and metrics can be fudged and obfuscated. The more I think about it the more I think the old PmO will survive !! Who'd be a Pm eh..

GarryP

Actually, if you parse what I said carefully, you'll find I'm not too happy with the conventional process and qualification view of project management. In large part because it empirically doesn't work. However if you are C-Level and do want it - you can automate your chosen methodology and it will probably work at least as well as your fleshy machines.

And if you turned it around and took an approach that could probably work, you've probably find that automation (of a different sort) would work well.

And don't get me started on metrics.

GarryP

This is kind of interesting, and worrying, and explains a lot.

Americans think robots will take everyone else’s jobs but not theirs

Looks like 50% of jobs are under threat, but single digit percentages of people think they'll be affected. That means a vast number are likely wrong - and due a rude awakening.

I don't think people really get what's already done. You can already integrate image classifiers, etc. into your app/device/etc. by including a few lines of code and giving Google a few cents. It'll tell you what it is, and where - down to the level of recognising which breed of dog it is. Same with OCR, speech recognition, understanding text meaning, etc. And they are as good as humans at the task. When you actually see where they are today, you begin to realise how many jobs are going to go by-by as these blocks are joined together.

If you want a play, this lot do a classifier demo where you can feed in your own image :

https://www.clarifai.com/demo

For instance, given this image



it gives back

• Donald Trump 0.312
• administration 0.980
• leader 0.975
• home 0.960
• people 0.959
• chair 0.937
• politician 0.934
• adult 0.931
• portrait 0.929
• one 0.892
• election 0.888
• man 0.883
• business 0.880
• league 0.864
• room 0.862
• banking 0.861
• meeting 0.857
• parliament 0.842
• capitol 0.811
• democracy 0.803
• indoors 0.800

which is probably better than most people could do.

BadgeIsBack

Big Fail? : Trump was matched low against others like indoors and chair?

Granted Trump was recognised

GarryP

It's the combination of two different classifiers (celebrity/general) - so the %age probabilities cannot be compared. Think how hard it is to recognise trump and not have someone who looks a bit like trump register highly. Next to that it's recognising that it's indoors with an 80% probability from ... what ?

Stuck in Auckland

as an Australian

Shard

This is spot on. I think it's because anyone over 40 probably remembers the false dawn of AI from the Seventies, and thinks that the potential debunked then, is somehow a natural law.

Another factor is when the tech is introduced, it tends to be a bit wonky (think speech recognition 20 years ago, passable, but glitchy) and then people don't appreciate the incremental changes. Look at Siri (etc.) and how relatively unremarkable consumer reaction has been to a long promised interface. It's that kind of insidious tech creep that is going to catch a lot or people out, IMO.

ozzieeagle

I've been using voice recognition a lot lately and it's improved unbelievably, problem I have with it is when using it for say forums etc, you dont really want the whole world around you to hear what you are posting. I can think of many other areas where the same applies etc, messaging, emailing etc etc.

So at best it's probably only useful as a command center and currently it lags behind on that. The rigours you have to go through to turn a tablet into totally voice activated on all apps is still quite daunting. I was going to give it a go, in fact I bought a hi spec tablet just for that purpose...But installing 3 different software apps just to get it working basically, is over the top. It's a long way from done and dusted yet.

GarryP

In part some of the reason why speech recognition has gotten better over the last few years is the application of deep learning - so it's all connected. There was a day when google turned on their new DL model and people noticed how much it got better overnight, without knowing the why.

BTW, they are working on the translation aspect similarly, with similar results :

https://www.nytimes.com/2016/12/14/m...awakening.html

Here's your opening paragraph in Finnish

and here's what you get when you run the Finnish back into English

It isn't perfect, but it's pretty damn close to getting the sense of the original rather than the words, certainly enough to understand. And it learns from simply getting large gobs of text thrown at it - which is why they have over 100 languages now, including Scottish Gaelic ("Mar ma tha duine sam bith a bha a 'cur dragh."). It's going to be a piece of piss to connect speech recognition at one end and speech synthesis at the other for a star trek translator.

Now imagine that rather than obscure languages, they threw previous insurance claim forms and assessments/settlements at it ....

Things like Amazon Echo and Google Home ain't perfect, but they are useful enough to keep around, and will only get better. Knock off the corners and really connect things together properly and I can see how they are going to take over the home - which is why both of those companies are trying to own the space for home purchasing.

Personally I think someone (probably google) is going to come out with a voice controlled/AI OS, rather than trying to bolt things on to the phone/tablet experience (which is itself derived from the Xerox WIMP interface) - give it a few years.

And when you take into consideration the ability to take in and understand visual AND audio data, coupled with at a minimum an understanding of particular niche domains - I think that even customer facing type roles are going to fall. Call centres (obviously), but shops (online & offline merging), fast food (similarly), etc. - basically anywhere where you've tried to turn humans into robots with process laden training and scripts.

Shard

A long read, but an excellent piece. Thanks.

Beoz

Hold on. Voice is just another input method. Same as your keyboard, mouse, hololens. We already replace call centres, shops, takeaway food, etc with those input methods.

What voice does do is allow you to make those purchases, cancellations, alterations, in situations where you can't use other input methods - like when you are driving your car giving the user more opportunities to make transactions. More transactions more jobs in other areas.

What are those other jobs? Well it doesn't replace the audio sales method, the complex scenario that needs sorting, the human touch. Imagine Foxtel allowing you to cancel your account with one click, one voice command. Not a chance. The call centre will get in the way before they allow that, then try and on sell you whatever they can.

Shard

Maybe not replace entirely, but how about a 95% reduction in human touch. VR goes far beyond call centres. For example teaching and healthcare systems could be voice driven.

GarryP

You deleted the bit before

It's not just voice as an input, it's the combination of interacting with humans via human mechanisms AND being able to at least look like an intelligent entity.

And the reason is bound up in your Foxtel example. I'd much rather deal with an AI that's programmed to appear to be human and compassionate than I would the average Foxtel call centre operative (the AI would be smarter for a start). It would also be much cheaper for Foxtel.

What's more the divide between 'online' and 'real world' gets blurred. Imagine walking into a shop, having the shop 'size you up' ('looks like a bogan, non-cashed up variety') and then meet your needs and sell you a product taking that into account, delivered to your home. Again, probably better, certainly more knowledgeable, and cheaper than your usual JB Hifi type.

Thing is, in that niche it's relatively easy to create something that can pattern match how the best human salesman works, and deliver it consistently. And learn to be better. All with less people involved, so less jobs.