Gen AI thing Part I: Plato’s spark

A friend recently asked me to help him understand the “gen ai thing” at a level that will allow him to have discussions (and since he knows me well, he knows this comes w opinion). I decided to go a level simpler, and try to explain Gen AI in a way my mum would understand (she’s in her 80s and my recent victory was getting her to carry her mobile phone when she is out of the house). I figured out it would take me a while, so I broke the explanations into smaller more digestible pieces. Here is Part 1.

What is Gen AI?

Before we go there….

First what is AI.(with apologies to my brother, Dr. AI)

Humans are a very arrogant species, so we decided that the way we think is something worth replicating. Hence, if we could make machines think like humans, then we would have something fantastic. Basically, machines don’t get tired easily, and you can expand the capacity of a machine much faster than a human (hopefully (1)).

AI is basically that, how do we get machines to think like humans.

So, what does it mean to think like a human?

How do you think?

Let’s take a simple example (a simple application of thinking like a human), you see a piece of furniture in a shop, how do you decide that it is a chair, or a table (assuming someone hasn’t written: this chair/table for $xxx)?

Enter Plato!

This is not a new question. Plato (~428-342BC that’s close to 2500 years ago) came up with a theory of forms, and that made me fall in love with PH102. The basic idea is that there is this world where the perfect form of every item in our world exists. So, I thought that makes sense! I know if something is a chair or a table by comparing it to the ideal form: it is closer to the ideal chair, or the ideal table?

What does closer mean?

If you have read other articles by me, you will remember I love talking about distance, closer means smaller distance. An object a is more likely to be an A than a B if it is closer to form A than form/ideal B. This is easy; how you define closer is where the fun begins 😊 

Plato’s Theory of Forms

So, when I started playing with data, Plato’s theory of forms helped me a lot. The main difference is that, since I can’t access the world of ideals/forms, I have to base my version of form on what I had seen before.

The tables I had seen were 4 legged, came up to waist high (since my teens), had at the top a large flat surface so you can put stuff on top. Usually they were made of wood, although the legs could be made of metal. Chairs, were shorter, below waist high, but also usually had 4 legs, and made of similar material. However, chairs also had a back, the flat surface was not the highest point of the chair, but the back, so the person can sit on the flat seat, and rest his/her back on the back.

So, when I see a new object, I decide whether it looks more like a chair or a table, based on whether it is closer to the typical form I had in mind. Note that, I am not comparing just these words as I described table and chair, but the more complicated concept I have in mind (like an ideal form)

While humans learn from experience, machines can be made to learn. Instead of telling the machine the short ungainly description of a chair and a table above based on what I have seen, the trick is simply to give thousands of examples of things we know are chairs and tell the machine, these are chairs, and same thing for tables. So, you train the machine so that it comes up with its own view of what a chair is and what a table is. This is the training part of a model. 

In this case, we train the model by feeding it images of chairs with the label that these are chairs, and the same for tables. This is called supervised learning, since someone supervised the process by providing these presumably accurate labels.

 For now, we skip on how the machine breaks down the images, and let’s just assume that the machine now knows what chairs look like, and what tables look like. We then feed it a new image with a picture of a piece of furniture without label, and it will tell us: this is likely a chair (or a table) depending on what it has learnt. The machine has solved the classification problem, by deciding the new unlabelled furniture is classified as a chair/table accordingly.

 Now, nobody stops you from training the machine with other pieces of furniture, and animals, and all sorts of other things… Afterall, that’s how we learnt, no?

Thought experiment:

Imagine you are walking about, and from far you see something. How do you decide whether this thing with 4 black legs, and black and white splotchy pattern on the top and sides is a table or a cow or may be a dalmatian?

How would your thinking process go?

Would it be faster if you remembered you were in a field in the middle of a farm, or close to a nature inspired furniture shop?

For me, yes; based on the context (where the object is), I can make the process simpler by focusing on a smaller list of likely choices, than the whole list.

This is why you get faster, likely better results, on a specialised machine (a farm animal identifier in the first case or a furniture classifier in the second) rather than a generic machine: a machine trained only on furniture would identify the table much faster and more accurately than one that has also learnt about cows and dalmatians. However, the furniture classifier would fail if someone asked it to identify a dalmatian… Hence, machines/algos trained on a specific set of data are usually better at working on that theme/context, but will not do so well at things in different contexts. 

It should not be surprising, if someone from the tropics had never even heard of snow, he/she would be flabbergasted the first time, may be even think it was volcanic ash… But someone who has lived in the snow would even be able to tell you the type of snow (4), it all depends on what you need. Similarly, I know of many Mandarin/Cantonese/French speakers who claim that there are many nuances in their languages that are not present in English. Again, depends on what the people who use the language use it for.

If I had not seen a chair and table before, maybe I could check out in a dictionary:

• Chair: a piece of furniture for one person to sit on, with a back, a seat and four legs (2)
• Table: a piece of furniture that consists of a flat top supported by legs (3)

Then based on these definitions try and decide…

But you will tell me, wait, the human has a lot of work to do, he/she has to label the pictures.

Well, yes, for supervised learning, as a child asks adults: “what is this? And this? And this? How about this?”. But you will recognise the work the child put in: the child takes in the image he/she sees, commits it to memory in one shape or form, then later, when he/she sees a new object decides whether it is a chair, table or something else.

It is also possible to feed the machine unlabelled pictures, and it will decide by itself how many categories of objects there are (you can tell it that if you want) and it will create its own view of things and when presented with a new picture, after having been trained, decide whether that object is a chair of a table. This is called unsupervised learning.

There also is reinforcement learning, whereby the machine is given feedback on what it has predicted, therefore can continue learning by analysis what went right and what went wrong.

Now whether you choose to use supervised or unsupervised learning is up to you, there are reasons for and against using either form. Not only that, but how you choose to learn or group things also makes a difference to the output you will get and the ability of the model/algorithm to properly classify things. This is something I am geeky about, but is not for this blog post

You will agree this is a very useful thing to have in your back pocket and the practical applications are very very vast. For example, a few years ago, I found it was not too hard to build something that, once you feed it a photo of a piece of meat from a supermarket, it can identify the meat with reasonable accuracy, and you can slap on features such as estimating price (after estimating volume), freshness… You can easily do the same for fruit: auntie, no need to press-press anymore!)

Ok, but this is only classification of objects, doesn’t AI do many many more things? Is this really AI, or is it ML?

AI vs ML

AI is, as mentioned above, focused on making machines think like humans. ML is how we apply specific pieces if this to solve problems. The classification piece I used is a piece of ML, but ML is part of AI. But there is more than that.

Classification is just a small piece of what ML can do. ‘Traditionally’, ML has been used to do 3 things: classifying things as I illustrated above (think the photo app in your phone tagging the pictures by recognizing what is inside), finding out what affects what (regression) for example understanding how the weather affects price of tomatoes, and predicting things such as predicting the price of tomatoes next week.

A little diagram will illustrate what I am talking about:

So basically, while trying to explain Gen AI as it is today, I used ML, basically applied AI, and took 1 aspect (classification). I skipped over neural networks, that can be used to classify the images by say automatically varying the importance of different aspects – is height of horizontal piece mor important than number of legs – or even deep learning that basically is a more complex neural network.

But, simply by looking at the name: Neural Network, you can get a hint that the original idea was to mimic the layers of the human brain, deep learning is adding layers and other complexities. So, fret not, I am not misleading, I am simplifying. Remember, my aim is that even someone like my mum can understand.

In my next blog, I will explain the most common understanding of GenAI, ‘ChatGPT’, or basically LLM (Large Language Model) because people using it are not coding (speaking machine language) but speaking their own Natural Language (oops I slipped in NLP)

1. Elon Musk’s Neuralink has been approved to have human trials https://www.cnbc.com/2023/09/20/elon-musks-neuralink-is-recruiting-patients-for-its-first-human-trial.html
2. https://www.oxfordlearnersdictionaries.com/definition/english/chair_1
3. https://www.oxfordlearnersdictionaries.com/definition/english/table_1
4. https://en.wikipedia.org/wiki/Eskimo_words_for_snow

Singapore the land of slow fast food (fried chicken!)

 I love fried chicken, (part of one vision...) When I was a kid, it was an adventure to go to the stadium to watch a football game with my dad, followed by either a vindaye and watercress in maison bread, or the luxury of a 2 piece chicken meal from KFC (at that time popularly known as “kentucky”).

I still love the treat of friend chicken and have my favourite chicken stall at Yishun Ring Road. I also like fast food fried chicken, including KFC of course.

My last 2 attempts at having fast food fried chicken in Singapore were disasters. In both cases I waited more than 30 minutes for my fried chicken. Hence the title: “land of slow fast food (fried chicken!)”

As I am someone who is really into using analytics to make a difference, I was not very happy at how, given the data they have, the 2 fast food chains messed up.

I will go in reverse chronological order, starting with the easier problem to solve.

Texas Chicken Sengkang

We went to Texas Chicken Sengkang. We waited for 35 minutes for a standard chicken set. The thing is that there simply was no visible queue. People who were waiting to eat-in were simply inside at their tables, waiting for their orders to be ready.

It was the day of the presidential elections, so as I was checking on my order, I told a couple who was considering ordering that I had been waiting for close to 30 mins. The guy decided to order and go vote, knowing he’d be done before his order was ready, and he was right. 

The case of Texas chicken is simple, they basically had too many orders, including remote orders (online) so they basically delayed everyone.

This is a clear issue of bad planning. If they have a system to manage their capacity to deliver orders, they obviously forgot that election day was a public holiday (it was announced on August 11 that if there was more than 1 candidate the polls would be on Sep 1, no excuse there).

I also happened to see only 3 and then 4 staff members in kitchen and front-office. I can easily understand why, given the variety of the menu, including delicious biscuits, it would take good planning from a resource perspective and clearly the manager was overwhelmed. A fellow frustrated customer told me, since I was taking pictures: “if you complain, her name is XXX, that is the manager”, but to me it is not really the manager’s fault. The organization failed at equipping her with tools for planning, or did not train her well enough to deal with the data.

This is a simple resource planning issue. You can make things a little bit more complex by adding the menu variety and the orders. Simple resource optimization with constraints, then you can add more constraints.

But my point is that there is enough information to enable the manager, after all they are the ones facing hangry customers. Whether it is because of process, lack of staff, inability to access relevant data (organization willful blindness), this can be resolved.

You know it will be a public holiday (voting day), you can expect higher demand, not a normal BAU Friday morning. You also know, in advance, that you can ask staff to be available. Then you know what orders are coming in, from where, in what order. It is a simple resource allocation problem.

KFC at AMK

We went to KFC AMK, before dinner time (we expected quick meal since we were hungry) ordered a meal of 2 pieces of original recipe chicken each. While orders may not be fulfilled in the order they were received, I think you would agree it is ridiculous to see many people who arrived after you for half an hour getting their orders earlier.

After some observation, I figured out what was wrong.

The chicken was coming out from the kitchen alright, but while we had been waiting 2 batches of chicken that were prepared were both spicy, not original. (1) It was a decision by management of the branch to keep customers who ordered original recipe chicken waiting.

I have no idea how long I would have waited, since I went to the counter to change our order to spicy and my order was fulfilled within the next few minutes.

To me, this is horrible, blind management. There is so much technology that is lying around the branch, from collating the detailed orders (self-serve kiosks) to machines that track orders and display the status, but it did not help my order and that of other people waiting for a long time – presumably original recipe lovers like me. Whether it was a conscious decision on the part of the branch management, or a default setting from HQ, or some preset rules, I have no clue, but to me it is a failure to use the data they have.

Unless KFC Singapore chooses to make some customers ordering their likely flagship product wait for ages, this should not be happening.

There actually are many ways to solve this problem. At a simple level it is about optimizing the orders: minimizing the wait time, given what has been ordered, the capacity of the kitchen and the staff. This can be optimised on the spot as the orders come in, the data is already captured, do something with it...

With some more effort, the displays at the kiosks can be enhanced to inform customers about the orders. Frankly, seeing your order being prepared for more than 10 or 15 minutes at a fast food joint is ridiculous.

Failing that, a minimum prediction of how much of what should be prepared and when. You do not have to wait for the order to prepare the meal. I remember the scene in the founder where the main character checked out the fast food and was shocked to be able to just pick his order without waiting…(2). It is a simple job of predicting the food required and prepare some of it in advance. Of course there is a risk that your prediction is wrong and the food gets cold, but it is a balance between this and extra ordinary waiting times. Analytics can work with the business to optimize this.

Now if you combine all 3 components (predict what you are likely to need, optimize your orders, inform customers at point of order), you get a nice ‘living’ system whereby everyone is informed, makes decisions accordingly. If I really have to wait 30minutes for my original chicken at point of order, I am likely to switch to spicy; but at least it is a decision I make and is within my control as a customer. KFC gets food delivered to happy customers.

Many analytics solutions are actually made up of different components supporting each others, and if well designed can work by themselves, but work better with a second component. This is how I prefer to work, show results quickly, gain acceptance – includes training people to use the system- and keep growing the results and the ability of people to use the system, rather than a big bang approach. But then this depends on the ability of the organization, especially people on the ground, to adopt the use of data.

Conclusion

Data is just data, or even worse is noise, unless the people on the ground are aware of how to use it, and the organization implementing it. KFC Singapore and Texas Chicken Singapore have obviously invested in technology, generate sufficient data, but do not seem to have enabled decision making on the ground by educating the staff and enabling them to make decisions.

In this day an age, this is really not doing right by their customers and staff.

If any of my contacts, readers agree with this post and know decision makers in KFC/Texas or any other fried chicken/fast food joint that is serving slow food and is unhappy about it, please make the introduction, the rest will be up to us 😊

 

 

 

 

1. There was a country wide promotion for special flavour chicken, but the promotion was not being advertised at that branch.
2. https://www.youtube.com/watch?v=KultzqPJaJs
3. Thanks to hotpot for AI generated image

How far should going green go?

Walking from a bus terminal/mrt over the weekend, I was intrigued by the banner below:

The messaging is clear, from 1am to 5am daily, the lights that grace this walkway will be switched off; this is part of a green initiative by the Land Transport Authority.

One of the things that make Singapore Singapore, especially if you come from the 'third world' is the sheer number of lights, how well the roads and walkways are illuminated. To me, and probably to many, a well illuminated street/walkway feels safer than a dark one.

Any decision made will have pros and cons. Is there an increase in the probability of crime (low crime doesn't mean no crime!) if lights in streets/walkways are switched off, especially irrespective of weather conditions (stormy nights can get very dark)

Isn't switching to greener source of power for the lights better than simply switching the lights off?

How much of an increase in the risk of crime a price worth paying for being a nicer shade of green?

Singapore firms found arming Myanmar, how?! what analytics is missing?

Myanmar is a country that is close to my heart, I felt at home in Yangon, loved the place and the people; I left before the new junta took over. Analytics is what I do for a living. So I am a little bit ‘apalled’ (1) that firms based in Singapore, using Singapore ports have been sending weapons to Myanmar; given that the Singapore government is at the forefront of finding a solution to the Myanmar crisis. (2). Singapore has made a ‘principled position’ against the Myanmar military’s use of lethal force against unarmed civilians (3)

 

How do we know the list of firms?

The beauty of this, is that it is from the horse’s mouth. It seems the list of arms providers comes from a leak from the procurement department of the Myanmar Ministry of Defence (4). While it seems that there is some debate around which firms have been selling arms to the Myanmar military after the ‘coup’ and the following sanctions.

 

How deep does the involvement of Singapore related companies go?

First of all, these are Singapore registered companies, under the purview of corporate regulatory authority. Secondly, if they used Singapore ports, a whole range of organisations must have been overseeing their operations, from the ports authority to make sure the physical movement takes place without hitch, to customs who are in charge of the legality of the trade. Now, I am in no way suggesting that there was any complicity, nor am I suggesting all containers need to be checked. But I do think analytics could have helped.

When trade takes place, most of the time, there is trade financing and insurance. Financial institutions get involved. Most of the time someone loans money for the transaction, someone will insure the contents. Again, I am not saying that some Singapore bank or some Singapore based insurer is involved, but chances are, there was. The analytics teams of the banks and insurers must have had a look and approved the risk involved. These transactions that broke the sanctions imposed by the Singapore government were not picked by these institutions.

We must also remember that, unless the military equipment has been manufactured in Singapore, they must have been imported, stored, then exported out. Hence the involvement of these organizations listed above is doubled.

Thirdly, there are companies who actually do the moving, the storing, but the companies listed by the procurement department of the Myanmar ministry of defense could be in that business themselves. So let’s just limit to these large institutions.

 

What could these organisations/authorities have done?

Some basic checks at the government related organisations, which I actually expect to be routinely carried out, but apparently are not:

• Are these companies, by the license to operate, allowed to engage in export?
• This is a basic hygiene check.
• Do these companies regularly trade with Myanmar?
• Any company that all of a sudden starts to trade with a sanctioned country should raise a red flag or two
• How about the pattern on trade? Even if they trade with Myanmar, has the volume changed, or the frequency?
• More interestingly, I doubt they would list helicopters on the manifest, but does the manifest gel with the container volume, size weight?

These are basic hygiene models that can be implemented as the front-end systems collect the data.and feed them back to back-end analytical systems. The government is engaging in building analytical platforms and has even made it possible for services from the most common commercial cloud providers to be used by government department and government related bodies.

There is no reason for these basic checks not to be implemented. Some of these are very common in say the banking sector as they assess risk of individual customers, especially those engaged in international trade. And this brings us to the private sector.

 

What could banks/insurers involved in the trade transaction have done?

I am pretty sure that financial institutions are very aware of sanctions and are bound to flag cases where these are be circumvented. It is often the case of closing the barn door after the horses have left, but the systems are built to facilitate flagging of potentially fraudulent transactions. Tweaking them to flag sanction-busting is not rocket science.

Even better, it is not difficult to predict the business as usual flows for many corporate banking customers, I have done that myself and I am sure many others have too for various financial institutions. It is a basic tool that allows banks to know, in advance, what are the funds required, by who, and so on… Now if a company asks for financing when they are not likely to do so by the models, add to that that it is going to a country under sanction, then a second look should be required. And that should leave a trail that efforts have been made to ensure the propriety of the trade.

Hence, to me, if any Singapore bank facilitated the transactions whereby military equipment was exported from Singapore to Myanmar since sanctions have been imposed by Singapore, then these banks are guilty of, at a minimum laziness, or worse keeping an eye shut to let profits roll-in.

 

Conclusion

Did you notice something?  I did not use LLM, ChatGPT, or even AI. Everything I mentioned can be done by very basic models/algorithms. All it would have taken is someone to understand the business imperative, and someone to get it done. But then again, may be it wasn’t high on the list of imperatives. And that is the reality.

What is the cost of these infractions? May be some companies that made huge profits sanction busting will close down, directors get a slap on the manicure, but trade has taken place, financing profited from, ports used…

As in many cases, doing the analytics, implementing them is easy, understanding the need for them and the will to use them is often the stumbling block. This is probably the case in this circumstance: when there is no will, way doesn’t matter.

 

1. https://frinkiac.com/gif/S08E08/770702/772154/SSBBTSBTSE9DS0VEIEFORCBBUFBBTExFRC4
2. https://www.mfa.gov.sg/Newsroom/Press-Statements-Transcripts-and-Photos/2023/05/20230519-Comments-on-Myanmar-Report

3. https://www.channelnewsasia.com/singapore/myanmar-arms-singapore-based-entities-mfa-un-special-rapporteur-3500051

4. https://thediplomat.com/2022/08/report-claims-38-singapore-based-firms-supplying-myanmars-military/

Tech Moguls call for regulation of AI, what gives?

It’s been a while since I wrote a blog post, was neck deep in project delivery, this topic forced me to make the effort to extricate myself.

Should AI be regulated as Tech moguls such as the OpenAI CEO asked for?

My simple answer is NO.

Please read on before flaming…

 

1 Virtue Signalling

There can’t be a better example of virtue signalling (1) than that, can there?

Think of most mafia type movies when the mafia sides with the police; more often than not it is to control a threat to the mafia itself.

I am not saying the ‘insiders’ calling for regulation are like the mafia; but I do have some experience of working with an incredibly smart technical guy, who solves technical puzzle after technical puzzle and whose baby has recently been purchased and he is now able to make his mark on the world. Over beers we were discussing why would he want to (at that time, a decade ago) create and agent that automates marketing, puts it in the hands of totally non technical people, and killing the jobs of fellow technical people. His answer was simple: it can be done, and if I don’t, someone else will. He agreed that sometimes it will go wrong, but that’s about implementation.

I believe that the tech moguls also think similarly. Now that they have opened Pandora's box, the want to quickly claim that what’s coming out is not our fault, please sit on the lid.

The message is “we are using AI responsibly, but we cannot comment on how others will use it. So please protect us from evil”.

To some degree, on top of being virtue signalling, this can also be seen as being anti-competitive. Hey, and you had the impression that the USA was all about competition?

 

2 Anti Competitive

Just look back at the TikTok debate that went all the was to congressional hearings. I am not into TikTok, and have not been following closely, but form what I have read, TikTok’s defence, if I can be so rough as to summarise in 2 main points:

1. we are not affiliated to the Chinese government, so please do not worry about us passing data to the Chinese government, so we are not spying for a state.
2. what you are accusing us of, others are doing too (unsaid is: or they are dumb enough not to know how to do it effectively)

To me, asking for oversight on TikTok specifically is anti-competitive.

It is a bit the same with HuaWei. Having worked in projects in the Telco field, I have some idea of data ownership. I do not think it is an exaggerated fear that ‘foreigners’ can ‘see our data’; but that works for all foreign companies, why just HuaWei?

What the Ukraine conflict should have taught us is that conflicts can blow up anywhere.

Furthermore, while at some point in time people thought that MNCs, as pure capitalist beasts, would be purely profit driven, this has been debunked; MNCs do care more about where they are headquartered/ formed than other places. Hence, not only for HuaWei. TikTok, there is a risk of governments getting involved. But to me, the key is, any company, any government.

Add to this how governments have been changing in the recent years, agreements reneged (I am looking at Mr Trump for example), and you have dirty waters everywhere.

It is ok that if you are not China you are wary of HuaWei, but that should also mean that if you are not Finnish, you should think about Nokia for example…

 

3 Missing the forest for the trees

Whether wilfully or not, the whole argument about regulating AI is missing the forest for the trees. People keep discussing about this tree, or that tree, ignoring the fact that the whole clump of trees forms part of a forest, with all sorts of trees, and animals and birds.

This is an issue that economists are very familiar with (the difference between micro and macro economics), and that Singapore government is very familiar with (abandoning the concept of Singapore Inc)

To me AI can do a lot of good and a lot of bad (most often both simultaneously), depending on what it is used for and whose point of view you are taking. At this moment only some points of view have been considered.

Most of the debate is on a micro level. How can I reduce costs by using CHatGPT. As a Clevel, I care mainly about my own bottomline, and that is loosely tied to my employer’s. So if I use ChatGPT to replace receptionists, call centre people, developers… I save tons. The beauty of it is that as 1^st mover, while my competitors catch up, I make fantastic profits. Even when they catch up, who will undercut us?

But what of people being displaced?

Is UBI an option (Universal Basic Income) is being trialled in the UK (2) and has been trialled around the world (3).

Just think of this, if a whole chunk of people lose their jobs, unless the cost of products and services drops accordingly so people can afford the products and services whose costs have dropped due to AI replacing some humans, how will they buy stuff?

What is the point of producing more efficiently, if people cannot afford to buy what you produce? Add to this that the cost of AI, say ChatGPT is today low, will it remain so for ever? People who have used new disruptors products and services such as Uber/Grab certainly have their opinion on this.

 

4 but what is the real problem?

The real problem, in my opinion, is that AI and anything is the marketing around it and people’s expectations.

I attended Microsoft Build. And I came out of it pretty excited. Especially the safeguards they are trying to impose in terms of responsible AI. One of the very important aspects is human oversight.

However, the real issue is that people think AI is the solution to everything.

Everyone is “ChatGPT-ing”(4), expecting miracles. It is a language model, not a truth model. But the reality is, what is the cost of mistakes it will make?

As long as people’s expectations are tempered, they truly understand there will be errors, and budget for the costs of these errors, then it is ok to directly use these models to answer business questions, say process orders…

Who needs Amazon when you can ask ChatGPT where you can find the product you are looking for at the lowest price? Sometimes you will get lemons. As long as you are prepared for that…

But to trust Chat GPT with opinion pieces is a different ball game altogether…

 

5 So what is the solution?

The solution is simple: education. People, users, need to be educated about the risks, so they choose which tool to use when.

Education, not regulation.

Bottom up, not top down.

And I dare say that the tech industry has been more about marketing and jockeying for position than educating.

If we want AI to help humanity in general, this has to change.

  

1. https://en.wikipedia.org/wiki/Virtue_signalling
2. https://www.theguardian.com/society/2023/jun/04/universal-basic-income-of-1600-pounds-a-month-to-be-trialled-in-england
3. https://en.wikipedia.org/wiki/Universal_basic_income_around_the_world
4. Please note I am using ChatGPT as a convenient stalking horse, I am talking about the popular use of AI tools in general.

Do you still think AI is omnipotent?

The most important thing the metal ox could have done to benefit humanity is to deflate the AI hot air balloon. If, as we prepare to welcome the water tiger people still believe in the omnipotence of AI, then we probably deserve getting replaced by AI.

AI failed in Covid-19

Covid-19 is probably one of the worse crises of recent memory, as of date, 5.6million people have died, almost 350m infected to date. (1). I think that, since many people believe AI is capable of solving most of our problems, be it AI is the answer to everything, from guiding the doctor’s diagnosis (general practitioner (2) or specialist (3)) and hand (4), to helping predict criminals (5), to more mundane things such as predicting what you will buy next and calculating the spending of your bricks and mortar spending (6).

But, don’t you realise something? The biggest challenge we have faced the last 2 years has been Covid19. Has AI helped solve anything?

No,

• AI failed at predicting the course of the disease (7)
• AI has failed at predicting the best approach to slowing the spread of the disease

So far, a virus has been beating the finest AI minds (or at least those who have been trying to solve the problem, or those who really think they could).

Does that mean that AI is useless? May be Covid-19 is just too big a problem for AI.

AI failed in OCBC Bank Scams

OCBC customers lost SGD8.5m in December 2021 (8). The banking world is one where AI is used aggressively.

Not only that, OCBC has magnanimously decided to re-imburse all the people affected (9), do note however that the affected people have to sign NDAs (10). Not only that, both OCBC and DBS are imposing measures to minimise the impact of such events (11).

What does that have to do with AI, you may ask?

Afterall, the CEO of OCBC, in yesterday’s Sunday Times details how 100 people were working 24/7 over the whole month of December to fight the war against the scammers. May I ask, the value of the effort of these people; SGD8.5m was ‘lost’ how much was saved?

But the real question is, with so much AI in the banking sector (12)(13)(14), even advising you how to use your money, why can’t AI have stopped these fraudulent transactions; especially since, in the case of 8.5m in OCBC the fraud took place over a whole month…

But these are to take your money and invest it…

How about protecting your money?

Well, AI is full at work there too! Predicting what fraudsters will do next (15), getting academic papers written about their success (16) and even recovering the funds in 30 minutes (17).

Is all this smoke and mirrors?

What is MAS doing about this? MAS is at the bleeding edge of AI, having global challenges (18), encouraging banks to control use of ML (19), coming up with pretty papers to control use of AI (20).

But, how about making sure AI is applied properly and actually protects customers of the banks?

Of course, you can argue that SGD8.5 million is medium-sized peanuts (less than average CEO’s salary (21) in covid hit 2020), but I would like to know how this fraud managed to get past the anti-fraud AI, and still continue to do so for a whole month. Also, while the country is watching OCBC, how about the other banks? Or is it a weakness in OCBC’s systems that has been exploited?

If it is the latter, what has MAS been doing? Fines and all that are just closing the barn door after the 8.5m stallion bolted.

Even more interesting is the laissez-faire approach by MAS (22) and contrasted with European legal framework where the onus is on the bank, rather than on the customer in such cases.

I really don’t think that it is common for people to suddenly and quickly empty their accounts via money transfers; even rule-based systems would have been able to catch these…

So is AI a failure?

In simple terms no.

Gartner has the AI hype cycle (23), see below:

Do you notice something interesting?
Most AI technologies are at or on the way to the peak of inflated expectations.

Basically: STOP BELIEVING THE ‘DATA SCIENTISTS’/’AI VENDORS’ AND USE YOUR BRAIN.

Am I exaggerating?

Let me tell you a small story and you decide. I was having this meeting recently, a teams working meeting with customers, all of a sudden a customer got up and did some stretches. She explained her watch told her to do so. I laughed and told her her machine has trained her well.

Just think about this.

A device, that tracks a few parameters is deemed to know more about you than your own personal machine trained over your lifetime (your brain), and with full access to all your sensors (senses, nerves…).

I think it may be ok for individuals to make these decisions, but when professional decision makers do the same, it is time to ask questions about what is really going on.

1 https://www.worldometers.info/coronavirus/

2 https://www.nature.com/articles/s41467-020-17419-7#Sec3

3 https://www.bbc.com/news/health-50857759

4 https://www.mobihealthnews.com/news/contributed-power-ai-surgery

5 https://mashable.com/article/china-ai-crime-minority-report

6 https://www.theguardian.com/business/2022/jan/09/do-smart-supermarkets-herald-the-end-of-shopping-as-we-know-it

7 https://www.flasog.org/static/COVID-19/COVID19PredictionPaper20200426.pdf

8 https://www.channelnewsasia.com/singapore/ocbc-phishing-sms-scam-do-not-click-bitly-url-link-2407796

9 https://mustsharenews.com/ocbc-scam-payouts/

10 https://mothership.sg/2022/01/ocbc-scam-victims-nda/

11 https://mustsharenews.com/ocbc-dbs-scam-measures/

12 https://www.ocbc.com/personal-banking/investments/roboinvest

13 https://www.dbs.com/newsroom/DBS_fortifies_NAV_Planner_with_new_AI_powered_digital_investment_advisory_feature_to_help_retail_customers_make_better_investment_decisions

14 https://www.finews.asia/finance/30159-uob-launches-ai-driven-financial-planner

15 https://www.straitstimes.com/tech/tech-news/anti-fraud-experts-use-ai-to-predict-cheaters-next-move

16 https://ink.library.smu.edu.sg/cgi/viewcontent.cgi?article=6345&context=sis_research

17 https://www.straitstimes.com/tech/tech-news/uobs-anti-fraud-team-thwarts-scam-and-recovers-funds-in-30-mins

18 https://www.businesstimes.com.sg/banking-finance/mas-launches-global-challenge-to-promote-ai-adoption-in-financial-sector

19 https://www.risk.net/risk-management/7409846/singapore-banks-tighten-ml-governance-amid-regulatory-scrutiny

20 https://www.mas.gov.sg/~/media/MAS/News%20and%20Publications/Monographs%20and%20Information%20Papers/FEAT%20Principles%20Final.pdf

21 https://www.businesstimes.com.sg/banking-finance/ceos-of-singapores-big-three-banks-paid-9-24-less-in-2020-as-covid-19-hit-profits

22 https://www.channelnewsasia.com/singapore/ocbc-scam-goodwill-payout-sms-compensation-lawyers-2445061

23 https://www.gartner.com/en/articles/the-4-trends-that-prevail-on-the-gartner-hype-cycle-for-ai-2021

Allow people to safely enjoy eating at hawker centres, "site visit" for policy makers

Dear policy makers, do you actually eat at hawker centres regularly? I am asking because I am not sure that the policies you have implemented are not actually increasing the risk of Covid 19 spreading.

Unless, of course, that is precisely the plan, if we all develop antibodies to various variants, then covid19 would be truly endemic. And I am ok with this, just would appreciate some transparency.

Let me explain my argument:

1             Pack your own tray at hawker centres

I have argued against this as being something that will cost the people who are working at clearing tables at hawker centres their jobs. Yes, the workload is lower, but this means that fewer people are required for the same hawker centre.

Now I look at things from the covid19 angle.

In the past, after eating, people would leave their used crockery on the tables, sometimes with unconsumed food such as prawn shells, bones… (I personally agree this should be heavily discouraged). The cleaner would come, collect the crockery for washing, clean the table, and make it ready for the next consumer.

Now, assume that a person who ate at this table was covid contagious. The table is likely to have the virus.

When the cleaner comes, if the cleaner is properly equipped – gloves, mask (may be face shield) and disinfectant, the plates are put in the container, the table disinfected. Being more protected that the average consumer, the risk of infection to the cleaner are not terribly high.

However, under current rules, the consumer himself/herself clears the table and carries the crockery with left-overs via a tray to a collection point. The cleaners take the trays, throw the left overs, and put the plates and cutlery to wash as per usual. The risk to the cleaners does not increase by much.

But, all other consumers also visit the tray clearing area. During peak hours, these areas are packed, it is not really possible to maintain the 1m social distance. Add to this the fact that the trays are stacked, many more consumers risk getting close to crockery under this scheme.

Add to this the fact that well equipped cleaners do a much better job at cleaning the tables compared to the average consumer who at most used a wet tissue, it can be argued that even at the tables, the old system ensured more cleanliness and lower risk of infection to the next diner.

Yes, I am fully aware that, in the past there were cases of tables remaining uncleared for a while, but think about it. If the table has not been cleared, chances are you would either not go to that table, or if you cleared it, you would be much more careful, compared to an apparently clean table.

I think that the issue is the necessity to equip the cleaners, like all front-liners, with proper equipment and training on how to use the equipment. I am sure their work would be better appreciated than before.

 

2             Maximum number of people per table = 2

Again, don’t get me wrong, I personally enjoy this rule because 2 is my preferred maximum number of people at a table. However, I am in the minority. Most people, even families, are made up of more than 2 people. So what do you think people will do?

Add to this the culture of sharing food in Singapore. One of the great joys of eating at hawker centres with friends is that you get to order more dishes and enjoy a little bit of each. Do you think people will stop sharing food? Sure, the serving spoon has appeared, but is that all?

(I don’t really go to hawker centres in office areas – I wfh – so my view is limited to families)

What families do is simple. First, they try to get contiguous tables. Then they still talk to each other, but this time across tables. This means:

-        Higher volume which likely means higher droplets expelled

-        The projectiles, instead of being limited to the table, now cut across tables, into passageways.

Next they will walk across the tables, exchanging dishes. Very often this is done without wearing a mask.

Hence, to me, unless the rules are tightened to disallow ‘mixing’ of tables as in restaurants (which I really do not favour, imagine enforcement costs), it makes sense, from a point-of-view to allow more people to sit together, and only, as previously, allow people from the same household to sit together.

 

3             Barricaded Hawker centres

In order to ensure that everyone who visits a hawker centre does get tagged by the TraceTogether system, the hawker centres have been wrapped up with netted barriers and only a couple of entry points are allowed, where the tagging with TraceTogether is verified.

What this leads to is a tight entry point, even assuming social distancing is respected, given the funnel effect, the flow of people is slowed, and people spend more time in a specific area than if there was no funnel.

Furthermore, especially in the case of take away, people used to wait outside the hawker centre for their order to be ready, just to allow others to order. Now all these people cannot leave the hawker centre, increasing crowding. Add to this the fact that hawker centres, while classified as outdoor, have very varied ventilation. Some, even in pre-covid days, were quite stuffy.

So what should be done differently?

The answer is simple, beacons. TraceTogether works via Bluetooth technology. Instead of bringing people to the Bluetooth device, why not place multiple Bluetooth devices at the hawker centres? Bluetooth devices are not that expensive…

 

Conclusion

In sum, I think things could be done better by maximising use of educating and training people to risks, revising some of the rules to allow better flow and allocation of people to minimise risks, providing proper equipment (to cleaners and blue tooth beacons). The old People, Process, Technology strikes again.