This article looks at what the evolution of AI means for business and society in general and things we can look forward to in the next decade.
If you think of primitive humans, we have always as a society aimed to be “tool builders” as Steve Jobs famously said. He called the computer a “bicycle for the mind” giving a famous example from Scientific American. The example goes as follows…
If you look at the energy expended to travel a distance from A to B, the Condor is the most efficient whereas humans are languishing somewhere to the bottom of the list. But a human on a bicycle is far more efficient than a Condor. Humans, Jobs reasoned, are inherent tool builders to augment our natural capabilities and the computer extends the natural abilities to think, create and compute.
The reasoning makes sense. One of our primitive forefathers may have used a stick to pick a fruit from a tree, unable to reach it himself. Therefore he extends his natural reach using a tool. This tool building extended to using rocks, knives, and even the wheel.
Fast forward a few hundred years and we are now in the midst of a tech revolution unprecedented in society. Arthur C Clarke, the famous scientist and science fiction writer who has inspired amongst others Elon Musk wrote:
“Any sufficiently advanced technology is indistinguishable from magic.”
This particular quote has been pondered over by many famous technologists and scientists. We are in the midst of an AI revolution that is mimicking and even surpassing human-level thought in fields as diverse as language generation, cognition, classification, and regression or extrapolation tasks. AI has become the proverbial Oracle which has answers to all our problems. From robots that can navigate such as those from Boston Dynamics to drug discovery, generating art, and self-driving, AI is at the cusp of an unprecedented revolution that will only exponentially increase in the coming decade it would seem as more dollars are allocated to this area of computer science.
With this in mind, it is time to revisit and perhaps extend the Turing Test. Alan Turing famously postulated:
“A computer would deserve to be called intelligent if it could deceive a human into believing that it was human,”
Turing wrote in 1950 defining his now-famous Turing Test.
The Turing test has since then been the subject of much research and competitions with a definitive “victory” for AI coming with IBM’s Deep Blue beating Gary Kasparov in Chess. Although that was not technically self-learning AI, and was a rules-based model of sorts, DeepMinds AlphaGo which beat a Human at Go is far more a “self-learning” AI learning from scores of games and improving its playing style till it becomes unbeatable. By now, Chess had also been cracked as Chess Engines can have far superior ELO ratings than humans can. As an example, Magnus Carlsen, the world’s supposedly best Chess Player has an ELO rating of 2859 whereas Stockfish, a Chess Engine can get an estimated ELO rating of over 3500. To the point, Carlsen has gone on record saying “he doesn’t enjoy playing against computers”. For a professional chess player to be beaten by a Chess Engine is death by a thousand paper cuts it would seem.
Elon Musk has taken this analogy even a step further to propose it is possible we live inside a computer simulation. He says “Many years ago there was pong. Two dots on a screen. How we have photorealistic games which can be played by millions of people around the world simultaneously. It is not inconceivable for games to become indistinguishable from reality”.
So now in the context of this, I would propose, somewhat hesitantly, a new type of Turing Test as follows:
If computing technology in a specialised field can beat a Human Master “consistently” – i.e. by a 99.999% margin, we can safely say that the technology has passed what I would call the “specialised Turing Test” in that field of expertise.
Let’s say this takes the form of 10,000 rounds of Human Vs AI. Has this been achieved ?! Let’s look at case studies.
- Chess – we can safely say Yes.
- Go – this looks like it will have the same conclusion.
- Self Driving – this isn’t quite there yet. Machines can’t drive better than humans.
- Diagnosing disease from medical records – for X-Ray scans of say Lung Cancer and other forms like Breast Cancer we can say yes.
- Robotics – It is not inconceivable a Boston Dynamics Robot can beat an elite soldier on the ground in a variety of tasks such as landmine clearing or reconnaissance.
- Decision Making – it is also not inconceivable for AI to make better decisions than Judges. They are not swayed by biases which are almost impossible to weed out in a human judge, and can possibly clean up the legal process.
- Business – Can AI make better decisions than Steve Jobs at Apple or Elon Musk at Tesla? No business leader has not made their fair share of blunders – under immense workload and time pressure. However, an AI can “assist” a human in decision making playing the role of a consultant or a secretary. In that case, it is not inconceivable for it to have a superior “strike rate” for better decisions than high-quality management consultants given it can be fed all Annual Statements and business case studies of all public and private companies in the world. AI could also theoretically make better accountants making fewer errors and certainly not being “prone” to committing fraud if tuned properly.
- Medicine – DeepMind has already solved the protein Folding problem that eluded scientists. It is also not inconceivable for robots to be able to do advanced surgery under the stewardship of humans. AI can also recommend the right dosage of medicines for patients and provide aftercare to patients after they leave the hospital as well as to the elderly at home.
- Art – AI is generating ever more realistic paintings and sketches often taking art from classic greats such as Da Vinci and Van Gogh. It is certainly possible for AI to create high-quality art in the same way diamonds are grown in labs rather than mined from nature in many cases.
- Government – can AI make public policy better than humans can? This is certainly possible given AI can churn through all public records, the history of a country and its evolution, and current trends in politics gleaned from News Events to generate Public Policy recommendations for humans to ponder and debate over. It may stop the misallocation of capital which has been a problem for almost all governments. This would ensure funds are allocated to the highest value public cause.
In almost all of these cases, we can see AI achieving parity and beating humans in specialised pockets and domains as per the proposed “specialised Turing Test”.
It is important for the author to mention specialised because it is still not conceivable for AI to do everything properly for the dream of AGI – Artificial General Intelligence – an all knowing all thinking machine to exist. However, once the specialised Turing Test is passed in a critical number of domains, they can be fed into another general AI that allocates problems to each of these subdomains depending on the need.
What would humans do in this case ?! Humans then become Stewards of AI, course correcting the algorithms from making errors such as making sure AI accountants are not prone to fraud or Government Policy generating AI are not funded and therefore biassed towards right or left leaning governments.
A general AI governing body becomes critical as AI picks up momentum and most countries if not all are behind the curve already on this. This, according to the author, is a critical intervention needed for AI to prosper in the coming years. As with Nuclear technology, AI has the power to improve lives exponentially but also cause destruction to society and commerce at a scale unprecedented.
To contact the Author for speaking or consulting opportunities, please contact at Ibrahim.Mukherjee@outlook.com, OR for
AI consulting at :- https://thegeneralaico.carrd.co/
