Getting to the heart and soul of AI
Artificial Intelligence (AI) is a common term we hear, but what does it actually mean? The challenges that are before us are not so much the AI technology, the heart and soul will be the moral and ethical decisions that need to be made behind the innovations.
For several years we have started to see limited use of AI in our everyday lives from Sat Navs to self-driving cars.
Hollywood has depicted AI as a potential threat to humanity from “War games“ in the 1980s which saw AI Ready to start world war 3 to the ruthless AI of the terminator series or the augmented capacity of Neo in Matrix allowing him to learn Kung fu by uploading a program.
I believe the future of Ai is going to be a collaborative technology that can benefit everyone. Thankfully we are a long way off from the Hollywood version of a superior AI. But there is a huge amount of military research and development in this area.
What stands out is the need for appropriate regulation and moral and ethical codes behind this technology.
A brief history of the Industrial Revolution
The last three centuries have seen Industrial revolutions starting with rapid advancements in technology:
- The first industrial revolution occurred when steam engines technology came into existence allowing the mechanisation of machines and the mass production of goods and saw big changes in textile production, chemical processes and development of machine tools from 1760-1840.
- The second industrial revolution started when a way was found to convert pig iron into steel. That led to mass production of a new strong structural material. Which consequently led to innovations in chemistry and the start of the petrochemical industries. From the 1890s on this led to the rapid industrialisation around the world. That continued to develop plastics and new lighter alloys for engineering.
- The third industrial revolution was the invention of nuclear fission technology from the atomic bomb in World War 2 to the subsequent nuclear race. A consequence of this was the need to develop communications that could survive a nuclear attack. A young engineer called Paul Barran pioneered a communication system that could keep running even if part of it was knocked out by a nuclear blast. This eventually led to the internet that we know today. Now, we take for granted the internet that is widely available and contains a huge wealth of information. The rise of the internet started the Information Age. Today the only limitation to learning is your own desire and hunger to learn more.
- Are we now entering the intelligence age, the fourth industrial revolution? Where a combination of raw computer power, vast amounts of data and combination of new technologies such as neural networks and 5G technologies allow the real-time acquisition and response to information that before was not possible.
A heart of AI
Broadly speaking there are two main types of AI, the majority of the technology that we use in our everyday lives is covered by Narrow AI:
Narrow AI – Is very good at specialised activities such as playing music based on your request, challenging you to a game of Chess or directing you to a destination by sat nav. There are three main types of narrow AI:
- Classic AI – This is a specific set of instructions combined with various variables that are logically executed. i.e Playing Chess against a computer
- Expert system – This could be a virus protection program that analyses threats to a system based on previous attacks. It then dynamically protects the system from coming to harm. The system then also updates itself as more knowledge is found.
- Machine learning – Planning the best route to a destination via your sat-nav avoiding traffic jams by crowdsourcing data from other drivers, weather, road conditions to calculate the best route. Or even driving a car to the destination through automated driving systems.
Artificial General Intelligence (AGI) – This is the potential for a machine to face a completely new set of circumstances it has never encountered before and know how to respond. This is currently the realm of science fiction and whilst current AI may be good in a narrow field it is not able to respond as a human being to a new set of circumstances.
The soul of AI
Today machines have abilities far beyond human abilities in certain areas i.e. numerically a computer can outperform a human in calculations. However, AI has a lot to learn and is still not very good with certain key human interactions:
- Lack of empathy – understanding the human’s emotions or mental state. It can deduce something based on previous outcomes but it is very limited in being intuitive of compassionate about the human’s condition from facial expressions to tone and language.
- Lack of manual dexterity – Robots using AI are not very good at operations that need manual dexterity. A car could be built by a robot but if it breaks down it needs to go back to a human to diagnose what the fault is and repair it. An interesting example of this is the one-handed robot that can complete the Rubik’s cube.
- Lack of ethics in AI – How does a self-driving car decide whether to save the pedestrian or the driver in a crash situation. There is a fascinating study in Nature that did a survey across different countries and cultures comparing what ethical decisions should be made depending on the social status, age and cultural background. These are the kinds of decisions that autonomous vehicle manufacturers are going to have to start making.
- Understanding nuances in natural language processing, i.e. consider the phrase “The children didn’t eat the grapes because they were bad” the AI might have trouble understanding who the “they” is referring to. It sounds simple but the combination of syntax and semantics make it very difficult to create software that can understand human speech correctly.
Gary Kasparov (Chess Grandmaster) lost to Deep Blue (IBM Computer) in 1997 reflecting on this loss he realised that in certain areas computers will always have the advantage. The heart of the issue is looking at how we will collaborate with artificial intelligence moving forward.
More recently in 2016, the reigning human GO champion Lee Sedol was beaten by the Alpha GO computer. This was 19 years after Deep Blue had beaten Kasparov. So whilst it might appear that AI is beating humans this is only in a very narrow field.
The game of GO is significantly more complex than Chess even though it looks relatively simple. “GO” is played on a board of 19 by 19 squares and each square can have three possible states. Therefore, there are 3 raised to (19×19=) 361 power of combinations which is more than the estimated number of atoms in the universe! It also requires creativity, forward-thinking and above all a desire to win. So the win over the human in GO marked an impressive milestone in machine/human collaboration.
Getting to the heart and soul
But the trend for computers and AI to start to understand the characteristics of winning a game through strategy and planning is one thing. It is much harder to emulate creativity, personality and human thought.
AI discussions tend to talk about the evolution of technology in Darwinian terms. But at the root of human existence is the mind, body and soul. Something that not even a machine with advanced AI can simulate. Humans are created in the image of God and endowed with creativity, talents and consciousness. AI does not have a natural sense of right or wrong it needs a human creator to set out the moral code.
The biggest question with the rise of these technologies are not going to be technological but they are going to be moral and ethical that is the heart and soul of the challenge.
AI skills economy
In the first industrial revolution, 40% of the workforce were working in agriculture but now less than 2% do. Unemployment hasn’t gone up just people have moved into different jobs.
You can see in supermarkets the transition to cashier-free checkouts and people are comfortable with using the technology. Or the adoption in fast food chains such as McDonalds of self-service checkouts. That means fewer cashier jobs but it does mean more skilled jobs.
One of the realities in the technological age is the differential between incomes between poor and rich. Executives in the tech industry are earning way more than the manual workers who are building the machines.
With the rise of the AI economy, it is going to be more important to work collaboratively with machines and technologies and retrain people in new skills. This could create an economic crisis if people are not able or willing to retrain with new skills and huge pay differential between the top and bottom ends of the pay spectrum.
AI Data protection morality
The rise of AI needs a new approach to the protection of data. One of the common ways to create a type of artificial intelligence is to feed in data and to produce a machine learning of common responses or identify inconsistencies. For example, Google translate gets a variety of phrases and human translation of those phrases and comes up with a reasonable approximation if it has a large enough source of information.
The question then becomes who owns the data and who has the right to process this information?
One of the key achievements of the GDPR regulation is that starts to enshrine in law the digital rights of the individual.
Unfortunately, we can see radically different approaches to the legal use of data in US, Russia and China as compared to the European Union. We need protection from friendly and unfriendly governments who are harvesting data without consent. This all raises ethical and legal questions where the law has not kept pace with technological developments.
DNA data breach
There was the recent case of an Ancestry site that is selling DNA kits for people to identify their family tree admitting to giving DNA records to the FBI without consent. Also recently a similar site Veritas genetics also suffered a data breach. It is one thing to have your credit card number stolen that you can change but when it comes to your DNA you can not change that. There is already DNA technology for cloning animals, but currently, it does not seem possible to clone humans but again the ethical and moral questions still need to be explored.
Law enforcement use of data
The recent protests in Hong Kong have seen the widespread use of AI power facial recognition to monitor and control protestors in horrific ways.
Edward Snowden’s revelation of the widespread collection and analysis of personal data of innocent citizens are all using AI technologies. The legal and ethical implications are still being debated. We can assume that law enforcement know a lot more about our digital lives through the digital trail we leave every second of every day.
Social media use of data
The Cambridge Analytica scandal and the way Facebook is using data for advertising and profiling of users. Selling data for marketing purposes to third parties for who knows what purpose.
We need to accept that in certain areas AI will always outperform humans. Like in a game of Chess or GO, they just need to make fewer mistakes than a human. We need to embrace this technology and see where we can add value, be curious and use the AI technologies to enhance our daily lives.
Humans are more flexible, intuitive and creative and can see potential in certain strategies whereas computers will repeat failed strategies relentlessly. There are difficult questions that need to be addressed in morals and ethics. AI is only going to reflect the strengths and weaknesses of their creators and society as a whole.
Most technology startups have a fail-fast philosophy this is not a good idea for AI. What happens when autonomous vehicles mistake a floating plastic bag for a child and crashes a car killing the driver. Or a robo-trader sells your portfolio after mistakenly misunderstanding a news headline and wipes out your retirement savings. Failing fast could have real and devastating consequences, who is going to be accountable for the outcomes?
The challenge is ahead of us and collaboration is going to be about thinking constructively with those that think differently to us.