The Reading Room

Artificial Intelligence: A Guide for Thinking Humans

Melanie Mitchell

Published: 2019, Pelican

Questions such as will AI take over the world, cause mass unemployment, make humans obsolete or create a Marxist utopia(1) feel like they are in abundance today. We all must have a view but the relatively recent nature of machine learning compared to other occurrences during our history, combined with its fast developments and general complexity often results in such a view being uninformed. Enter stage left Melanie Mitchell with her new book Artificial Intelligence – A guide for thinking Humans. It covers the intellectual history of the field with characters such as Alan

Turing and Ray Kurzweil, alongside the practical successes (and continued challenges). There is also a healthy dose of involved but digestible overviews of the technical aspects driving AI achievements. For instance the explanation of techniques such as image recognition software or natural language processing using “deep learning” through convoluted neural networks (or convnets for short!), was certainly worth persevering with.

The most valuable aspect of the book was how she contextualised the overall discussion with an examination of intelligence itself. How can we answer some of the questions above if we don’t actually understand what constitutes thinking? Put differently – to understand artificial intelligence we first need to understand what understanding itself means and how we do it as humans.

The history of AI since its beginning in the 1950s has been somewhat seasonal. Certain developments about a machine programme performing a human function create significant high-profile promise of impending world changing developments which attract considerable funding and media coverage. The resulting developments don’t materialise or underwhelm, causing an AI winter of intellectual and financial disinterest. We are certainly in the midst of considerable promise and investment, and within the context of mass availability of data and significant improvements in computing power, there is a compelling argument at the very least to say that the length of the cycle of interest and disinterest is narrowing. Computers can beat human beings at Chess and Go, recognise images, automate large aspects of driving, and predictions from certain quarters believe that a general level of intelligence amongst machines could be with us by the end of the decade.

The reality according to Mitchell is that AI is a long way from being at a general human level. How long? “Take your estimate, double it, triple it, quadruple it. That’s when” is the view she shares of Oren Etzioni, Allen Institute for AI (launched by Microsoft co-founder Paul Allen). But what about Ray Kurzweil’s view of the singularity whereby machines learn to do everything better than humans – crucially including making machines? This is based, amongst other things, on Moore’s law of the doubling of computing power every two years resulting in an exponential lift off at some point, due to the power of compounding. If these trends continue, a high-street computer will “achieve human brain capability (1016 calculations per second) …around the year 2023. At that point, Kurzweil believes that achieving human-level AI will just be a matter of reverse engineering the brain.” Whilst much of the successes of AI have been based on neuroscience, our understanding of the brain is too superficial. Moreover, a new-born baby’s brain does not have human intelligence until it has interacted over a period of time with a complex world. Kurzweil has a bet with Mitchell Kapor that a computer will pass the Turing Test by 2029 on the website Long Bets. Kapor asserts that without the equivalent of a human body, and everything that goes along with it, a machine will never be able to learn all that is needed to pass his and Kurzweil’s strict version of the test.

This is compelling when viewed in the context of the types of mistakes that AI currently makes. For instance, despite the significant advances in translation software when Google Translate tells you in English that the French diner “forgot to ask for his piece of legislation” rather than “forgot to ask for the bill”, it reveals more than an amusing mistake. The artificial intelligence programme is just following pattern recognition in its software and by getting the context so wrong shows it doesn’t truly understand the task it has been asked to solve. Until it can understand the context of a situation it can never be viewed as intelligent – at least as we humans define it.

Beyond poor transferability, AI software can be manipulated to detect things hacking software wants it to detect. Having initially felt that that our entire legal, financial and Governmental bureaucratic systems could effectively dealt with by AI, I am now more cautious due to learning how easily manipulatable simple image recognition software is. For instance stickers which are undetectable to the human eye can be placed on stop signs causing driverless cars to continue rather than come to a halt. Despite the sign being large, red and having STOP written on it the software detects the image and understands it differently to how we do. We understand from a very early age that a large red thing in that shape probably means something important even if we never came across the word STOP before.

Our understanding of more complex concepts, particularly abstractions, is made possible by using more earthly ones in the form of metaphor. Mitchell references George Lakoff’s book Metaphors we Live By to demonstrate how AI would struggle to understand the world in its current unevolved form. For instance, to understand something like time we use the more concrete concept of money. Time can be spent, saved, wasted, worthwhile, or is often used profitably. We often use physical directions to understand emotional states – we fall into depression, have lifted spirits, feel up or down. It’s difficult to programme software to learn like this. For instance would a computer think to use temperature to better understand social interactions? Icy stares, cold shoulders, and warm welcomes, allow us to quickly describe a photograph whilst AI struggles to simply identify the objects therein.

For those thinking that they can use recent advances as a way to predict stock markets or solve inefficient government should note that considerable engineering muscle power goes into programming AI. In terms of directing that power “it’s almost like an artform to get the best out of these systems…there’s only a few hundred people in the world that can do that very well”, according to Demis Hassabis, CEO of Google DeepMind, there’s a lot of human driven trial and error involved, calling into question the nomenclature of “machine learning”.

Microsoft’s Eric Horvitz describes recent advances as taking us only to the stage of alchemy, but what’s clear is that the availability of big data and ever-increasing computing power has been a game changer for progress in attempting to improve the ability of machines to do things that were the sole preserve of humans. Reading the book has given me a better understanding of why and how this has come about, whilst informing me of the dangers and opportunities that lie before us as we embrace developments that the vast majority of us are blissfully unaware of. To stand a better chance of not making catastrophic mistakes, having an informed opinion is of vital importance and Melanie Mitchell has made an excellent effort to allow us take the first steps.

Amul Pandya
July 2020

(1) Artisan carpentry in the morning, theatre going in the evening.

The information contained above and in other entries in the Ocean Dial Book Review Series is intended for general information and entertainment purposes only, and should not be relied upon in making, or refraining from making, any investment decisions. No information provided herein should or can be taken to constitute any form of advice or recommendation as to the merits of any investment decision. You should take independent advice from a suitably qualified investment adviser before making any investment decisions.

Leave a Reply

Your email address will not be published. Required fields are marked *

Africa Is Not A Country: Breaking Stereotypes of Modern Africa

Dipo Faloyin

Published: 2022, Vintage Publishing

Brimming with humour and intellect, Faloyin’s fascinating volume mines the rich and varied span of histories and cultures of the continent’s many countries, delivering a punchy corrective against lazy sterotypes of Africa.

Read more

Chums: How a Tiny Caste of Oxford Tories Took Over the UK

Simon Kuper

Published: Published: 2022, Profile Books

The lack of academic effort generally put in by Oxford undergraduates is an important theme running through this book, penned by Financial Times columnist Kuper. It links this work-shy attitude to the centuries’ old dominance of Oxford by the top English public schools spawning “top tory toffs with a born to rule attitude”.

Read more

What I Learned Losing a Million Dollars

Jim Paul and Brendan Moynihan

Published: 2013, Columbia Business School Publishing

Jim Paul’s meteoric rise took him from a small town to governor of the Chicago Mercantile Exchange, yet he lost it all in one fatal attack of excessive economic hubris. In this honest, frank analysis, Paul and Moynihan revisit the events and examine the psychological factors behind bad financial practices in several economic sectors.

Read more



This website is provided by Ocean Dial Asset Management Limited, whose Registered Office is at 13/14 Buckingham Street, London WC2N 6DF and Registered Number 05583807. Ocean Dial Asset Management Limited is authorised and regulated by the Financial Conduct Authority (“the FCA”), with registered number 447424.

The information contained herein does not constitute a distribution, an offer to sell or the solicitation of an offer to buy any securities in any jurisdiction in which such distribution or offer is not authorised. In particular, the information herein is not for distribution and does not constitute an offer to sell or the solicitation of any offer to buy any securities in the United States of America or to or for the benefit of any U.S. Person as such term is defined under the United States Securities Act of 1933, as amended.

No investments or services mentioned on this website are directed at U.S. Persons who are not Professional Clients or Eligible Counterparties as defined by the UK Financial Conduct Authority (FCA) Handbook or Qualified Purchasers as defined under the Investment Company Act of 1940. The information contained herein does not constitute a distribution, an offer to sell or the solicitation of an offer to buy any securities in any jurisdiction in which such distribution or offer is not authorised.


ODAM does not manage money for retail clients. Eligible investors only may access ODAM’s expertise through the funds or investment companies for which it acts as investment manager.

This website is not directed at you if Ocean Dial Asset Management Limited (“ODAM”) is prohibited by any law of any jurisdiction from making the information on this site available to you, and is not intended for any use which would be contrary to local law or regulation. This website and its contents are not intended to sell services or products over the Internet, rather for Internet viewer convenience and information purposes.

Information displayed in this website contains material that may be interpreted by the relevant authorities in the country from where you are viewing the website as investment advertising or an offer to purchase securities. Accordingly the information on this website is only intended to be viewed by the persons who fall outside the scope of any law that seeks to regulate investment advertising in the country of your residence or in the country in which the website is being viewed. Users of this website are therefore assumed to have the knowledge and experience in business and financial matters to enable them to evaluate the merits of, and risks of, investing in the investment funds referred to in this website. Persons of any other description should not enter this website nor rely on any of the information contained therein. If you are uncertain about your position under the laws of the country in which the website is being viewed then you should seek clarification by obtaining legal advice from a lawyer practising in the country of your residence or in the country in which the website is being viewed before completing this question.

Risk warnings

Investments may fall as well as rise and investors may lose a substantial portion or even all of their investment. Past performance is not necessarily a guide to future performance.

Investments denominated in foreign currencies can result in the risk of loss from currency movements as well as movements in the value, price or income derived from the investments themselves.

Investment in emerging markets involves risks which may not be typically associated with investing in more developed markets. Investment in small- and mid-cap stocks may also involve a higher degree of risk as these markets tend to be more volatile than their larger capitalisation counterparts.


To the extent permitted by relevant law and regulation, ODAM expressly disclaims any and all liability which may derive from this website or any other information provided by ODAM in connection with this website, and any errors therein or omissions therefrom. The description of the investment management services and processes provided by ODAM in this website is written in general terms. The services and investment funds referred to in this website should not be regarded as an offer nor solicitation for such services or investment funds in any jurisdiction where such offer or solicitation is unlawful. The terms and conditions applicable to individual investors investing in investment funds managed by ODAM are set out in each relevant fund’s offering document. This website contains information designed only to provide background material for determining whether or not to consider investment. This website should be read in conjunction with the relevant offering document for each fund. All potential investors must carefully read the offering document which will contain significant additional information needed to evaluate investment in the fund and important disclosures regarding risks, fees and expenses. The offering document is the sole document upon which potential investors should rely. Recipients in any doubt about the content of this website should seek advice from an independent professional adviser.

This website is based upon information that is considered to be reliable but ODAM does not represent or warrant that it is accurate or complete, nor should it be relied upon as such. Views and opinions expressed within this website including comments on sectors or individual stocks or companies are those of ODAM and its Mumbai based affiliated company Ocean Dial Asset Management India Private Limited (a company incorporated in the Republic of India) as at the date of publication. Subsequent events may cause these views to change. The investment policies and procedures described are guidelines and not rules. Occasional deviation from the parameters given may result.

Use of links
The ODAM website may contain links to websites operated by other parties. ODAM does not control the content or accuracy of information on such websites and does not in any way confirm or endorse the material placed on such sites. The links are provided for your reference only.

The copyright and all other rights in all of the material on this website are owned by the company listed on the web page or the material is included with the permission of the rights owner. You may review and copy material on this website for your own private or domestic use only. All other copying or use is prohibited.

Governing law
You agree that in the event of any dispute of any nature that may arise between you and ODAM, these terms and conditions shall be governed by and construed exclusively in accordance with the law of England and Wales and you agree to be subject to the exclusive jurisdiction of the courts of England and Wales.

Privacy policy
Any personal information that you supply ODAM via the website and other means will be held in accordance with the General Data Protection Regulation – please see our privacy notice for more details . If you would like to be removed from our database at any point, please email your request to

Use of cookies
For the purposes of gathering data to measure traffic and patterns of usage on our website, we need to use your IP-address. In addition, for example, we use cookies to ascertain as to whether you have agreed to our disclaimer.
A cookie is a small file held on your PC. If you do not want to accept cookies, you will need to alter the setting on your internet browser. We will not use your IP-address or any cookie to identify you personally.