AI and Job (in)Security

The increasing capabilities of AI has created concern and even apprehension in the Natural Intelligence community.  The concern begins with our limited understanding of evolution; where we assume that all AI will behave like humans and go about destroying mentally inferior species such as the homo sapiens. But then, even if it doesn’t go about annihilating us, it will still take our jobs.  So people are writing learned articles about which jobs will still be around when AI is as commonplace as a smart phone.

The doom of driving (e.g. taxi drivers) is predicated to be just around the corner as AI starts driving autonomous cars. Beyond that we imagine smart helpers relieving our every day chores, maybe even doing our shopping and paper work for us.

The common prevailing conception is that low skill jobs would be the most endangered. But in terms of return on investment it seems the greatest impact would actually be in highly skilled jobs. Google already is toying with the idea of creating art with AI but how about engineering.  Can we replace engineers with AI ? Or at least augment their capabilities ? and how about inventors ?

Engineering consists of a toolbox of methods to solve a problem and a weight function to determine the worth of one solution compared to another. The challenge is often there is no ideal solution and the optimization consists of trade offs.  Moreover, the models are partial, the data collection limited and the decision process fraught with estimations.  However, engineers can solve problems because they apply prior knowledge and recognize similarities with past problems.  This is exactly where AI excels. If it can play GO, it can probably solve engineering problems.

But can AI invent things ?  It probably can.  Not only that, since there are already systematic approaches to inventing its surprising that we don’t already have AI that replaces the innovation leads and pathfinders in the R&D.  Google should probably be developing AI for inventing and not for doodling.  Or maybe its done that already.



Last slide standing

How do you end your slide deck ?

With a big thank you slide ?  

Have you really thought about that thank you slide ?  

The most valuable real estate in presentation is devoted to being polite. That’s certainly a worthwhile trait but realistically, by the time you are proudly being polite people have already forgotten your presentation. 

So if you feel compelled to say thank you, than put it in the summary slide. 

The ABC of presentation is tell what you are going to tell them (outline) tell them (body) and tell them what you told them (summary). 

The last slide standing should summarize your presentation and if you feel compelled add  a thank you as the last line to the last slide. 

Marketing mosquito traps

While mosquitos are small in size, their impact on our well being is disporptionaly large. From malaria to zika, or just being a nuisance late at night mosquitoes are a problem waiting to be solved.  

The mosquito fight has shifted from wide range chemical warfare, with a broad and severe environmental impact, to local skirmishes. As usual, the individual consumer is left alone in the night, combatting the buzzing sound of a hungry mosquito. Hence it is no wonder that many companies have jumped into the void and offer mosquito traps or repellents. 

The amusing fact is that reading the existing studies reveals that these traps or repellents don’t really work. But the traps are an amazing study in marketing products. 

Most traps work with UV light. Studies have shown that many insects are attracted to UV light but not mosquitoes. Mosquitoes are actually attracted to heat and CO2. Some traps utilize TiO2 which is a white material commonly used in paint and sun protection lotion and claim when UV light iradiating TiO2 generates CO2. 

So there is some science in place. Now placing a trap in open space results in bucket loads of insects. It’s actually impossible to discern mosquitoes from other pests but the sheer amount provides proof the trap is doing its job. 

The lesson to be learnt is that while its best to solve the job at hand, selling is sometimes about demonstrating results. 

Automated shade system: a crowd funding story for August

Inventions come from necessity and many times as I find myself spending some time outside I wish for some shade. This got me thinking about a drone assisted solar powered shade system. 

Four drones carry a canopy which has a solar panel. The drones position the canopy in the sky to create a shade above the user. 

Crazy as it sounds its actually feasible. ( the less physic’s inclined can skip this part). A ten meter square canopy made or organic solar cells should be able to generate around 50 W and yet be light. Furthermore the canopy can be shaped as a wing so the drones don’t actually lift it but only define the angle and speed providing an elevation force for the panel. The drone battery should suffice for the initial elevation but once in place in the sky the sun power and canopy lift will take over. 

Positioning is done using a camera on a drone. Once in the sky, the user marks the location to be shaded on his cellphone and the drones do the rest.

So who will take up the challenge and go for a Kickstarter campaign. 

Why Waze shows you the past

As usual, when driving back home during rush hour, I turned on Waze to get insights on the road conditions.  After a few minutes Waze diverted me to a side road to avoid a traffic congestion.  However as I looked at the highway, traffic was flowing freely while I was stuck in a traffic light on the diversion.  It then dawned on me that the success of Waze in directing people to the diversion, means that Waze is blind to the changing conditions on the highway.

Various solutions to this problem are possible.  One example is A/B testing, i.e. having some users diverted while others are not diverted.  In this manner Waze can keep up with traffic conditions, but at the expense of its users and it risks loosing its users trust if they are used as experimental pawns.  Another example is reducing the persistence time of events, i.e. in the absence of active reports or driving speeds Waze reverts back to normal conditions.

While the consequences of this in Waze are mainly loss of time, a future with autonomous cars may be more complicated.  Autonomous vehicles will pass through three phases

  • a minority of autonomous cars and limited functionality
  • a majority of autonomous cars and car to car and car to device communications
  • only autonomous cars allowed on roads

Its reasonable to assume that there will be at most a handful of car operating systems and probably all will meet some basic criteria.  This means that the diversity of driving patterns we see on the roads today will disappear. So we have a situation of similar objects with many degrees of freedom and interacting in a non linear way.  In physics such system are invariably chaotic.  A simple way to understand the complexity of the system consider a single road.  When the distance between cars is large and there is no interaction the traffic flows.  However as more cars populate the road they start to interact.   When one car brakes so do all the others.  We see this even in human driving, as highways become congested the traffic is erratic even without visible obstructions.  For autonomous cars this will be accentuated since all cars have the same underlying logic, so certain situations are prone to trouble.  Since this is a road specific and multi car scenario, it requires extensive and accurate simulations to create safeguards in the car operating system which identify and change the interaction rules to delay the onset of chaos. These situations will be exasperated in mixed driving conditions when humans try to take advantage of the predefined rules of the autonomous vehicles.  For example, if an autonomous car always keeps it distance, there will always be a human driver cutting in, which will push the autonomous car further back in line.  The mixed driving period will be challenging in every aspect as even innocent behavior can have far reaching consequences raising questions of liability and fault.




Those who can, do

In the world of trading stocks, a well known adage is

If you have a good trading system you trade your money, if it’s mediocre you trade other people’s money and if it’s lousy you write a book about it. 

In inventing as in stocks, to make money, talent alone does not work and luck is an important factor.  But to come up with ideas is certainly a learnable trait. I have met many people who teach how to invent. One would imagine that such teachers can generate ideas effortlessly, however the reality is that inventors need at least two aspects of knowledge to be prolific. The first is indeed systematic methods of inventing. The second, which in today’s lean and quick world is often forgotten, is a wide knowledge base. If we compare inventing to quick reading, than the methods of systematic inventing are methods of quick reading, but the knowledge is the books.  

The biggest driver for knowledge is curiosity, which is why curious people make good inventors.