Why I think we need to be able to assess Artificial Intelligence

This article also appears on LinkedIn.

I think we can all agree - scientists, businesspeople, interested beginners alike - that we're currently experiencing a hype called 'Artificial Intelligence'. We're getting blasted with CEOs proclaiming the end of human labor , the toppling of decades-old business models and in general headlines about either our salvation or damnation - all because of 'AI'.

All this, because in 2012, some very clever people used a technology invented in the last century called neural networks and solved a particular problem. This problem can be summarized as 'make a computer tell you what is on a picture". Awesome, hu?

“Monkey.” - “D’oh!”

Granted, Alex Krizhevsky, Ilya Sutskever and Geoffrey Hinton did some amazing research there, since this problem (more scientifically called 'image recognition') can be called one of the (if not THE) holy grails of computer vision. They twisted the concept of neural networks, seriously tweaked some of the ingredients, invented some really cool new ones and quite visibly destroyed every other approach out there to solve the problem. From their findings, the two times declared dead field of neural networks rose out of its grave with a mighty roar - like a particularly vocal zombie.

“GIVE US YOUR DATA!”

Photo by Nathan Wright on Unsplash

It kicked off countless new research projects and led to progress in various other fields that employed the same or even more advanced versions of neural networks - pardon me, we now call it 'Deep Learning'. Now, products are being developed that use this approach to pattern recognition quite successfully (or not).

Is this now the end of the world? Why do we now suddenly think that robots will now enslave us all? Why are suddenly AI startups valued so high, without tangible results? Why are companies slapping 'AI' stickers on their products?

To give you an honest answer, I don't really know, BUT: I have a few ideas about it.

The AI hype

Tech startups with the impact of a Facebook, Amazon or Google have not surfaced now for over 10 years - there's a reason we still call it FAANG. It's almost as if people realize that you don't need the umpteenth photo-sharing app in your life. Also, the smartphone market, a huge revenue driver and ecosystem creator over the last two decades becomes saturated and matures. So after ONLINE and MOBILE (and failed hypes like big data), what's next? You guessed it.

Over the last 2-3 years, 'AI' (there's a reason I put this in quotes) has been said to soon solve every problem there is. Healthcare, mobility, journalism, love, climate, beer and the stock market. Even in popular literature about how to improve your personal life, Artificial Intelligence is hailed as the ultimate solution to all of our problems (don't get me wrong - I love Mark Manson's new book). All that, because essentially we made a computer figure out, how to tell us what's on a picture. Billions of dollars (and RMB!) are being poured into startups claiming to use AI for whatever product or service they are offering.

My personal opinion is that this technology sparks creativity and imagination in people, while requiring quite a huge amount of tech-background to fully understand. I've had conversations along the lines of 'can't we build something that listens to meetings and then automatically transforms what we speak about into a slide-deck complete with drawings and text?'. Let that sink in. I always get the feeling that people without a computer science background think of Artificial Intelligence like a person, just one in a computer, that doesn't cost any money and basically outsmarts all of us. They are hardly to blame - pop-culture and a whole flood of news articles can make you believe exactly that.

Typical depiction of an Artificial Intelligence in popular culture.

Image Credit: http://pngimg.com/ (CC BY-NC 4.0)

The necessity to assess AI

Think about it like this: Already, in a lot of B2B-products 'AI' is being advertised. I did a quick Google search for you:

• ...intelligence engine...

• ... uses AI technology to ...

• ... with AI technology ...

• ... launches ... with AI technology ...

• ... with AI solutions ...

• ... AI-enabled ...

I’ve an honest question for you: Can you tell which ones of those are legit references to real Artificial Intelligence technology and which ones are just marketing BS? All that while a study from MMC Ventures just recently found, that 4 out of 10 startups don’t even use AI, although claiming so.

But which ones of those actually add real value to a product? Does a software benefit from being 'AI-enabled'? And how so? Is it worth the premium I need to pay for that? Wouldn't just the plain old one do?

And did you know that all AI products require extensive human-powered supply-chains behind them? That in Finland even prison inmates feed the ever-hungry appetite of machine learning algorithms?

Don't get me wrong: There's really incredible work being done with neural networks and all the other fields of AI (you DO know that there is more to AI than just deep learning, yes?). And there ARE really impressive products being thrown on the market that benefit from these cutting-edge technologies. They WILL transform our lives in the next years - as smartphones and the internet have too.

But my general feeling is that with this very fuzzy buzzword, it is easy to catch attention, leading to clicks, investor attention and customer/client decisions. Ask yourself: if a vendor approaches you and talks about how they've now improved their finance software product 'using AI' - are you able to say whether you should buy it and not just stick with the old one? And if you are not, are you able to ask further and deeper questions? Can you ASSESS and EVALUATE the proposal? Shouldn't you?

The AI market is said to contribute US$15.7 trillion to the economy in 2030. AI-derived business value is estimated to approach US$3.9 trillion in 2022 - most of that is from B2B software products, meaning, they sell it to your company and YOU need to make a decision whether it's worth it. Your current knowledge is based on news articles, hearsay and that one conference you went to where the people from above talk about how AI is going to change the world.

Do you think that'll do?

Btw.: I have just put together an online course that teaches the basics of Artificial Intelligence to non-techies in a fun and engaging way. Are you interested? :)