What does technology do to society?
When technology and human ingenuity gets together, everybody in society profits. If you look at graphs of GDP / capita over long time periods (850 years+), the trend is always upwards. The only things that drag this progress down are severe periods of sickness (like the black death) and widespread war (such as WW2). Overall though, technology makes everyone healthier and richer.
At this point in our society’s history – technology has never been more important to the global economy. 5 out of the top 6 companies in the world are technology companies – Apple, Alphabet, Microsoft, Amazon and Facebook – the non tech company being ExxonMobil. Only 10 years ago though, Microsoft was the only tech company in the top 6 – clearly technology is only growing in importance.
We have moved on from the age of agriculture, steam, oil or finance to the age of technology, as Azeem Azhar showed us at ProductTank London.
The most important thing in technology is AI
There are three types of AI:
- ANI – Narrow AI – A system that doesn’t have to be specifically programmed for all of its outcomes and can learn from its experience. It can still only solve a defined set of problems and cases though.
- AGI – Human-ish AI – What most people mean when they think about the Turing test. These types of systems can solve lots of different problems without being programmed specifically for them.
- ASI – Super AI – Intelligence that can think like us and solve general problems while defining its own goals. The key characteristic of this AI is that it will work out how to make itself smarter and will do so exponentially.
In reality, ASI is a very long way off and nobody really knows what it will look like when it gets here. The most important research work is being done to build general problem solving systems that are in the AGI bracket – and even this is largely at the mathematical modelling stage.
ANI systems are in production and use right now – from working out what objects are in a picture to predicting sports scores – these are the systems that most digital professionals will end up employing very soon if not already.
AI is a blanket term
Traditional AI is made up of things like decision trees, natural language processing and predictive analytics. Machine learning starts to take these inputs and make more out of them than we had anticipated at the start of the system’s programming. In order to describe the situation we’re in, it’s best to use AI as an umbrella term for all these activities.
AI is already here
The most well known examples of recent AI success:
- We beat GO – an incredibly complex game with 10^170 move combinations which as recently as 1999, computer scientists thought we were 100 years away from beating using AI. Last year the world champion human was beaten 4-1 with the machine even creating a new type of move by going against conventional wisdom.
- Siri, Cortana and Google are all on the market providing voice recognition software to perform simple tasks for us that didn’t exist 12–15 years ago.
There are three main reasons for the explosion in the AI industry right now:
- Moore’s Law is continuing to drive processing power to build these machines on demand and shows no sign of slowing down.
- We are able to provide these systems with enough Data and Information to make decisions.
- Businesses have been digitised and interfaces have been created to access their services using technology.
2012 was the first time that GPUs were being combined with deep neural networks (a 40 year old technology) to allow us to do very interesting stuff and solve previously unimagined problems.
Progress has come rapidly from that point with http://image-net.org/ seeing computers overtake humans in image recognition 2014. Speak recognition saw the same milestone in the last 18 months.
Pretty much anything that a normal person can do in less than a second, we can now automate with AI
Where Will we be in 5 – 7 Years?
- Conversational interfaces will run certain types of processes with people
- Safe autonomous driving systems will become more commonplace
- Power optimisation in data centres (and problems like it) will be optimised by machines where humans can’t keep up
The Product Improvement Loop
Once you put AI into a product you should see an ongoing improvement of that product through the learning that the system can now generate. This leads to more people using the product which means the data that it has at its disposal to optimise itself increases, and you are hopefully set up with a self fulfilling cycle of improvement that leads to exponential success…
Product Manager Considerations
Everyone needs to be considering how AI fits within their product or marketing space. The challenge for many people is that these technologies work best at a large scale with big data sets and are still very expensive to run at production levels. There are options available though:
- Buy off the shelf capabilities such as IBM Watson, Google Compute or AWS
- Experiment with creating scale even as a startup
- Find a niche to explore and exploit, and use that to build your product based on data sets that aren’t broadly available
Finally, don’t forget that you are now building products that fundamentally control how people experience and interact with the world. More likely than not you will end up making decisions that affect how happy or successful people are or could be – these need to be considered and taken seriously.
One example of this is the way that Facebook looks to mediate its content for suitability based entirely on AI – leading to challenging situations which we haven’t worked out a way to solve at scale yet. These weren’t foreseen when their product was being designed, but need to be considered so they do not just focus on one person’s opinion or political views.