Hello World by Hannah Fry. Read this book.
My take
If you are having any issues explaining algorithms, data or the impact to anyone from a student to the CEO. READ this book ! It is brilliant.
I highlighted more in this book than most
Hannah brings out that algorithms and data bias is part of society, however the scale is now something we cannot hide from. Past controls were hidden but now are out in the open, but sometimes we don’t have a clue how it happens
We have a choice every time we use a service - be lazy or be in control. Neither is better however, to become dependent on the algorithm is not about the loss of control but about the loss of identity
Control of the algorithm is not limited to the code, but the weaknesses of the machine, design and data.
Love Hannah’s take on AI, will be writing up another book on AI soon - so will keep thinking on AI for that.
When people are unaware that they are being manipulated, they tend to believe that they have adopted their new thinking voluntarily (Epstein) …. We may believe we are immune; we are not.
Algorithms provide a convenience of authority
Only humans can feel the weight of responsibility for their decisions
We have no idea what the data can tell us or why - but the relationship can prove powerful
Targeting works (efficiency) but are not always effective - the algorithm is not the problem, the data and what it is designed for vs the outcome expected greats a gap
The end of inferring about you and your likes without your consent is real and soon to be enforced - this changes the model, changing the algorithms is far harder
Incentives vs privacy is a better debate
We cannot be predicted based on our data - we are too messy, irrational and impulsive ( see blog on the mind is flat)
We still need to understand why we put more faith and trust in the output of an algorithm than our own judgement
Since we are so unfair and biased - why and how can we expect algorithms and data not to be?
Machine Learning - we get the tech, we know the data bias but unsure how to feedback false positives and false negatives to create better outcomes.
Subtle signals about health lie in the tiniest signals and therefore the most unexpected fragments of data - it will take time.
Interesting point that we may try to solve problems that the body or system will solve, data does not always lead to the right outcome (health and cancer detection)
People appear less worried about hacked data (assumption it already has been) than now that the machine may project (make assumptions) them in a light that they don’t wish for.
Privacy vs public good - this needs so much more work. Are you a citizen first or an individual. Far better read is The Future of Capitalism by Paul Collier ( blog to follow) who addresses the policy on why this dilemma exists and what we can do about it
Signals ( and noise) are the number one data issue.
Human ability and machine performance are related and causal
People are predictable ( how this conflicts with previous thinking that we are too random to be predictable) - this plays to Daniel Ariel’s work on Predictably Irrational )
Given we cannot agree on good, beauty, quality or anything else that involves human emotions (feelings) why are we asking algorithms to come up with solutions. However this thinking plays to Antonio Damasio work on the Strange order of things; this explores the link between chemistry and feelings and why there will be no objective measure (yet)
We need to get far better at assessing algorithms and working out how to buy them.
Maybe we should build algorithms that are contestable? The transcript of how they came to an conclusion
My big take away
Hanna suggests the need for algorithm regulation - however I would not be in favour of regulation (I am not a fan of regulation and its role in a new or competitive market) whilst I like the thinking there are better policy instruments.
However debate is needed about how much reliance we should have on self governing algorithms ( both the performance of the algorithm and the reliance on the data that built the algorithm) - this is corporate governance issue 101 for 2020.
---
If you like this book - you need to also read Artificial UnIntelligence by Meredith Broussard
----
From the publisher
“You are accused of a crime. Who would you rather determined your fate – a human or an algorithm? An algorithm is more consistent and less prone to error of judgement. Yet a human can look you in the eye before passing sentence.
Welcome to the age of the algorithm, the story of a not-too-distant future where machines rule supreme, making important decisions – in healthcare, transport, finance, security, what we watch, where we go even who we send to prison. So how much should we rely on them? What kind of future do we want?
Hannah Fry takes us on a tour of the good, the bad and the downright ugly of the algorithms that surround us. In Hello World she lifts the lid on their inner workings, demonstrates their power, exposes their limitations, and examines whether they really are an improvement on the humans they are replacing.”