For the longest time, I have been wanting to dive deeper into AI and share some thoughts. This article barely scratches the surface - it is merely an exploratory article on some of the capabilities of AI.
Artificial Intelligence (AI) is arguably one of the most important technological revolutions of the next decade. We live in exciting times! we are already starting to see AI introduced more and more into the elements of the gadgets/websites that we use on a daily basis. It is no longer a concept that exists only in sci-fi movies but it is becoming a reality around us.
However, the internet seems to be painting a somewhat misleading picture of the current powers of AI which I wanted to clarify further. I could be wrong, but I believe we are a bit far from the point where AI takes over the world (we still need to develop software’s consciousness and motivations). Don’t get me wrong - AI is transforming industries like advertising, commerce, finance and translation …etc. but even though major developments have been achieved so far, but there still further work needed.
The creation of “Universe” by Elon Musk, Sam Altman and others at OpenAI, is in itself an affirmation of my above point but a great step forward. A platform / environment that can train systems (AI) to accomplish multiple tasks and adapt accordingly to various challenges. e.g. Have one AI learn how to master War Craft (game) then have it shift to play another game. Through which the AI can apply that experience it learned from the first game to get itself up to speed on the next. And that’s the sort of training AI needs to learn how to tackle different problem sets over the next decade across various industries from healthcare to space travel.
AI can currently be applied only to specific types of applications and do that very well - but then applying that same AI to another non similar application does not guarantee success.
To make this clearer, I’ll borrow the example made by HBR; AI’s recent progress so far is through one type, in which simple input data (A) is used to quickly generate some simple response (B). Example:
- Photo tagging: Input A: Picture >> Response B: Are there human faces (Yes or No)
- Loan Approval: Input A: Loan Application >> Response B: Will they repay the loan (Yes or No)
Current systems today can do the above very well through huge amounts of data but that’s still not enough for AI to overtake humanity. Engineers still need to simplify inputs (A) and identify outputs (B). Our brains are designed to think beyond just A to B systems.
AI is still being approached with a narrow view of solving a specific need/problem - Which is fine and could be game changing; but not that super power that will take over the world – yet, it could be used to do evil.
The amount of room for automation of A to B systems is already revolutionizing many industries including self-driving cars but I believe there is still a lot of room for creativity to go beyond these methods to combine one or more algorithms to solve a single problem set (ensemble methods). That is what engineers at Google for example are doing to tackle translation, trying to build a single algorithm that can adjust its own structure to tackle problems (Google Neural Machine Translation) – not having to teach the computer all translations, yet, having it succeed in doing a good job with incomplete sets of data.
These are some of the interesting applications/industries AI is being applied in and what I view lies ahead:
- General Intelligence:
Moving to a more General intelligence might have us encounter an era of Behavioral AI. Not all humans have the same response to one question. Building personalities is something I haven’t seen many startup’s tackle. E.g. take the robot from the movie interstellar; having a humor setting …etc. (In a way neural networks might not be the answer)
- Messenger Bots:
Programmers are still fascinated with the idea of building responsive consumer targeted Messenger Bots using off-the-shelf algorithms. I believe this will still continue to grow – but would be a short lived fad. I absolutely agree with Bill Gates when he says that through AI we can build one very powerful personal assistant e.g. Alexa with its voice commands – but I don’t see apps dying and morphing into messenger bots. And I don’t see consumer bots contributing too much to AI development – as they are more designed to be conversational rather than perform an actual task.
- Self-driving cars and Traffic Control Infrastructure:
Self-driving cars are something that are inevitably coming. People are even predicting that some highways would only allow self-driving cars to get on (means we would always be stuck conceptually in the back seat). Google’s self-driving car has traveled an impressive 200 Million miles (320+ Mill Km) while Tesla’s autopilot exceeded 140 Million miles in just the past few months alone.
With that said, traffic infrastructure might need to be redesigned to cope up;
Robin Li; the CEO and Co-Founder of Baidu (China’s Google) said that AI can be used to adjust traffic lights and optimize urban traffic congestion (so instead of having set timers for traffic lights – they became a constant variable linked to the flow of cars).
But not only that – imagine all cars are connected and communicating with one another – then imagine all cars having real-time feeds on traffic situations and are able to control their speeds and match the best routes and communicate with traffic lights – we might be theoretically be able to have cars designed to maximize green lights all the way (assuming no congestion).
AI and self-driving cars are also about to disrupt the auto-insurance industry. Tesla is now aiming to sell car insurance as part of the car’s final price. Autopilot contributes to the fact that cars are becoming much safer than human drivers (40% lower crash rates). This trend will push insurance premiums negatively lower to adjust for safer conditions.
- Logistics and Warehousing:
Richard Liu; the CEO and founder of JD.com (Jingdong Mall) – one of the leading e-commerce websites in China (a major Taobao competitor); announced in Beijing this year that they will be establishing the first real uninhabited fully automated warehouse. From storage, to shelves, to scanning to packaging to loading all automated. No more people! 99.9% accuracy – insuring same day deliveries.
- Robo-advisors and lawyers:
Roboadvisors have made their foray into the Fintech scene with wealth management applications. Providing automated algorithm based management advice without the use of human financial planners. The question that sparks debate – does this mean that rich companies will get richer with access to these super intelligent forms of machine learning advisers? Does that mean companies betting on AI like Google, Amazon, IBM, Microsoft ... are here to stay with access to massive amounts of data and super computers?
On another hand does this mean people will have a better chance of pushing for a universal basic income – since computers will be doing major investing decisions on their behalf?
- Country Policies:
Countries that don’t start considering AI in their policies are about to fall behind the rest. Take one of the most recent reports released by the White House on AI saying that “it is expected that machines will come to reach and exceed human performance on more and more tasks” (warning of job losses – which they seemingly view is not a big of an issue as long as productivity can be increased).
Mark Cuban recently said; “Artificial Intelligence, deep learning, machine learning — whatever you’re doing if you don’t understand it — learn it. Because otherwise you’re going to be a dinosaur within 3 years.”
One of the things that I found interesting in a recent interview by one of the guys leading Facebook’s AI team (Joaquin Candela); was that he breaks down the applications of AI into 4 areas: vision, language, speech, and camera effects. All of those, he says, would lead to a “content understanding engine.”
These systems are proving to be existential in playing a role in how Facebook’s newsfeed ranks posts and becomes more user relevant – unlike twitter’s cluttered noisy environment.
So I think going back to my question – Could AI be the Future of the Future?
I think the answer is clearly YES.
20 Years ago there was no Internet and it changed everything as we speak.
I am excited to see what not 20 but just 10 Years of AI advancement would lead us to in the next decade.
But to expand on my question – would AI be the Future of the Future to all countries/companies equally – or would it provide an advantage for some over others?
Which draws my attention to Mark Zuckerberg’s 6,000 word manifesto letter detailing how FB will create the “infrastructure” that will help solve some of the world’s biggest problems – helping understand content; detect terrorism, fake news, polarization of societies …etc.
These tools need time to develop but could prove to be crucial for a 21 century government to manage economic and political affairs. Ask yourself – with the dominance of Facebook world wide – people like Zuckerberg stop focusing only on quarterly profits for shareholders – but they look beyond for the next 5,8,10 years horizon on what tools will help companies stay relevant and ahead of the game.
Remember globalization benefits weren't shared equally across the world - could the same happen with AI?
That’s all for now. Until my next post :)
If you enjoyed reading this please help to share with your friends.
Have anything to add? Please leave a comment below.