Let’s imagine you are the Chairman of the Board for a large international company. During a board meeting, the CEO and his sales team are presenting a market thrust in Germany. You’ve been losing market share to a major competitor and the plan is to pump €40M into the market to take back that market share and the market leader position. After the presentation a lively discussion arises on the pros and cons of the plan, likelihood of success and the potential ROI. The board is relatively supportive and is approaching a decision when the computer sitting (do computers sit?) at the end of the table says “May I object? Your investment of €40M would have a higher likelihood of success and would give a significantly better ROI if it was invested in Kazakhstan.”
How do you handle a suggestion like this from a computer?
(This question was posed by Erik Korsvik Østergaard to Martin Risgaard and myself over a Tuborg at Tuborg Havn.)
But let’s start somewhere else…
How do you handle a suggestion from Google maps when you enter a destination and ask for directions? Do you doubt that Google gives you the best solution? Maybe a bit. You maybe have had an experience with Google maps where it sent you on a freeway that ended up being filled with traffic…
That’s probably the position I had until a few weeks ago. I was driving to the other side of Copenhagen for a meeting with a potential client. Google maps sent me out on the freeway – and the traffic was more than just a bit heavy. I asked Google again – just to be certain – and it confirmed – stick to the freeway. I was going to be a couple of minutes late if I did.
So David’s brain kicked into overdrive – ahhh – Google seems to have forgotten that little back road that I’ve used so many times. We’ll take that one instead – we’ll beat Google. Things worked brilliantly – there was still traffic – but less than the freeway. And then suddenly the traffic all disappeared – beautiful! Until I realize that the reason there was no traffic on this road was that the road was CLOSED!
So what did I do? Checked back with Google which could now tell me that I would be 9 minutes late for my appointment – and then gave the client a call to let them know that I was running late.
So – do I still think I’m smarter than Google? My confidence in Google’s “omniscience” on the traffic situation has certainly grown and I don’t think I’ll be suggesting alternatives to myself in the near future. 🙂
Let’s take a look at another bit of input to the AI discussion.
In March 2016, DeepMind’s “computer” AlphaGo was playing the strategy game of Go against Lee Sedol – the undisputedly best Go player of the past decade. AlphaGo won the match 4-1 (Lee had predicted he would win 5-0 or 4-1).
During game 2, move 37, AlphaGo made a move that baffled all the professional Go players who were watching the game and it so unnerved Lee that he got up from the table and took a break for a couple of minutes.
What was so unnerving about #move37?
It was unnerving because no human being had ever made that move and none of the professionals (including Lee) could understand why it had made that move. Either it was a mistake or else AlphaGo was so “smart” that we didn’t understand why it made that move.
The correct answer is the latter.
It was only after the match that the programmers were able to go behind the scenes and understand what AlphaGo was “thinking” when it made #move37.
AlphaGo had been fed thousands of professional games and thus had learned from the best. But this wasn’t a move that it had ever seen a professional make.
It had also been taught to play the game against itself – and learn from these games. This is where it becomes interesting. It was able to collect data from these simulated games and the result was that it had taught itself #move37 as a viable option – the winning move!
So – back to the boardroom!
We still have a ways to go. If you want to know more about where we actually are, then get in touch with Erik Korsvik Østergaard – he plays with some of this fun stuff.
In reality – what IS our response when the computer starts beeping?
After discussing with Erik and Martin my own response was that I would request an explanation from the computer – just as I would today if it was one of the other board members that suggested a move into Kazakhstan instead of Germany. If the explanation was sufficient and could be validated, then I would choose the Kazakhstan option.
Now remember – this was pretty much my thought about Google maps – I didn’t even actually ask Google (not sure I know how…) – I just assumed that I knew better and took a shot at a shortcut. Today – I don’t even ask – I just shut up and do as Google tells me to do! 🙂
It only took me one mistake to figure out that I don’t actually know better!
How many mistakes would I have to make before I just accepted what the computer suggested?
There is no doubt that there could be a temptation to let the computer make those decisions at some point in time – however, the real value-creator here is using AI to augment our own intelligence. We need to always ask critical questions of AI – just as we need to do so with our colleagues when making significant business decisions.
There is no doubt that the future will be highly impacted by Artificial Intelligence. Do we need to fear? I choose not to! I choose to make use of AI to augment intelligence – and ultimately make better decisions.
But where will AI fall short? What will the role of leadership be? I’ll be back… 🙂