Artificial Intelligence

[Fennaio]

Hello everyone,

I'm a start-up AI software business, in year 2, going quite well after year 1 spent developing the software with no income, glad I embraced the risk.

I won't plug the software I sell on here as I've never posted before and it isn't the point of this post.

I'm just wondering what everyone's thoughts on using AI are? I'm definitely seeing an increased appetite for it in some sectors but also increased scepticism about it in other industries.

I could be wrong but I still think there's a misunderstanding by many people of what AI is and how it can be used in business.

Would love to hear everyone's thoughts on AI, and also more than happy to answer any questions on it. Hopefully we can get a good debate going.

 

[fisicx]

Mostly used as a marketing buzzword for machine learning.

Why do I need an AI enabled fridge?

And the computing power needed to train these systems is consuming resources at an alarming rate.

 

[JEREMY HAWKE]

I dont have a very good understanding of it but I do see it as being a massive future benefit to business . I was talking to a customs expert and he explained that the technology would eventually streamline and simplify exporting and importing for businesses of all sizes and that this is already being trialed

 

[Fagin2021]

It would be helpful if you gave us your definition of AI and how it differs from machine learning.

Also, an example or two of how your software could help a business.

Otherwise, all you can expect are generic answers.

 

[Fennaio]

Mostly used as a marketing buzzword for machine learning.

Why do I need an AI enabled fridge?

And the computing power needed to train these systems is consuming resources at an alarming rate.

It would be helpful if you gave us your definition of AI and how it differs from machine learning.

Also, an example or two of how your software could help a business.

Otherwise, all you can expect are generic answers.

Yeah I think this is where the misunderstanding is sometimes. Data Science v Machine Learning v AI, and I agree - unless you know what you're looking for Machine Learning is often touted as AI when actually it's glorified programming.

Data Science isn't AI - although it's plugged as AI quite often - it's just statistics.

Machine Learning 99% of the time isn't AI - it's using Data Science algorithms coded into a computer/machine to generate an output based on inputs - requires a human to to create the specific algorithm and almost always they are programmed to solve a specific problem so aren't very dynamic. Much better than traditional programming as you don't tell them what to do after they've trained, but not AI as they aren't dynamic.

Proper AI is using a model of a human brain (at it's basic structural level) to solve multiple problems. Artificial Neural Networks they are called, or you may have heard the term Deep Learning, which is the same thing. The only programmed bit is the individual neuron. Link these together and they can learn from anything given to them - words, sounds, images, data. This is what I use so it's genuine AI, not Machine Learning.

In terms of computing power - that's a myth. It's not like cryptocurrency mining. Good AI can learn from enough data required for most things in less than a minute on a standard computer. For particularly large data (GBs) 30 minutes is more than enough. Image data does take more power though than standard data.

As examples, my software is used for:

• Customer churn
• Know your customer (KYC)
• Natural Language Processing
• Fraud and anomaly detection
• Crime prediction (minority report come true)
• Behaviour profiling
• HR
• Product Recommendations
• Customer Journey mapping
• Supply Chain
• Sales and Marketing
• Investments
• Diagnostics
• Fake content spotting
• Sentiment analysis
• Real time response to process events
• Cyber security
• ...loads more.

The overall power of it is it's able to predict very accurately what will happen next extremely quickly - that can be applied to pretty much all aspects of business and operations.

I can post a link to the software if allowed - will wait for a mod to give permission.

 

[Financial-Modeller]

Not much to add, but sounds like a fascinating business in a rapidly evolving space.

Perhaps pop a couple of case studies on here to help some of us luddites understand how you have used AI to solve a specific problem.

 

[MOIC]

For those that understand AI, it's the future, as are robots.

 

[DontAsk]

Yeah I think this is where the misunderstanding is sometimes. Data Science v Machine Learning v AI, and I agree - unless you know what you're looking for Machine Learning is often touted as AI when actually it's glorified programming.

Data Science isn't AI - although it's plugged as AI quite often - it's just statistics.

Machine Learning 99% of the time isn't AI - it's using Data Science algorithms coded into a computer/machine to generate an output based on inputs - requires a human to to create the specific algorithm and almost always they are programmed to solve a specific problem so aren't very dynamic. Much better than traditional programming as you don't tell them what to do after they've trained, but not AI as they aren't dynamic.

Proper AI is using a model of a human brain (at it's basic structural level) to solve multiple problems. Artificial Neural Networks they are called, or you may have heard the term Deep Learning, which is the same thing. The only programmed bit is the individual neuron. Link these together and they can learn from anything given to them - words, sounds, images, data. This is what I use so it's genuine AI, not Machine Learning.

It's still machine learning, just a more complex machine. It's not intelligence.

In terms of computing power - that's a myth. It's not like cryptocurrency mining. Good AI can learn from enough data required for most things in less than a minute on a standard computer. For particularly large data (GBs) 30 minutes is more than enough. Image data does take more power though than standard data.

As examples, my software is used for:

• Customer churn
• Know your customer (KYC)
• Natural Language Processing
• Fraud and anomaly detection
• Crime prediction (minority report come true)
• Behaviour profiling
• HR
• Product Recommendations
• Customer Journey mapping
• Supply Chain
• Sales and Marketing
• Investments
• Diagnostics
• Fake content spotting
• Sentiment analysis
• Real time response to process events
• Cyber security
• ...loads more.

The overall power of it is it's able to predict very accurately what will happen next extremely quickly - that can be applied to pretty much all aspects of business and operations.

I call BS.

What even is "customer journey mapping"?

So who is using the software?

 

[Fagin2021]

Yeah I think this is where the misunderstanding is sometimes. Data Science v Machine Learning v AI, and I agree - unless you know what you're looking for Machine Learning is often touted as AI when actually it's glorified programming.

Data Science isn't AI - although it's plugged as AI quite often - it's just statistics.

Machine Learning 99% of the time isn't AI - it's using Data Science algorithms coded into a computer/machine to generate an output based on inputs - requires a human to to create the specific algorithm and almost always they are programmed to solve a specific problem so aren't very dynamic. Much better than traditional programming as you don't tell them what to do after they've trained, but not AI as they aren't dynamic.

Proper AI is using a model of a human brain (at it's basic structural level) to solve multiple problems. Artificial Neural Networks they are called, or you may have heard the term Deep Learning, which is the same thing. The only programmed bit is the individual neuron. Link these together and they can learn from anything given to them - words, sounds, images, data. This is what I use so it's genuine AI, not Machine Learning.

In terms of computing power - that's a myth. It's not like cryptocurrency mining. Good AI can learn from enough data required for most things in less than a minute on a standard computer. For particularly large data (GBs) 30 minutes is more than enough. Image data does take more power though than standard data.

As examples, my software is used for:

• Customer churn
• Know your customer (KYC)
• Natural Language Processing
• Fraud and anomaly detection
• Crime prediction (minority report come true)
• Behaviour profiling
• HR
• Product Recommendations
• Customer Journey mapping
• Supply Chain
• Sales and Marketing
• Investments
• Diagnostics
• Fake content spotting
• Sentiment analysis
• Real time response to process events
• Cyber security
• ...loads more.

The overall power of it is it's able to predict very accurately what will happen next extremely quickly - that can be applied to pretty much all aspects of business and operations.

I can post a link to the software if allowed - will wait for a mod to give permission.

Those aren't examples - it's just a list of things you say it can do without specifying what the input and output looks like.

Take item 2 - know your customer (KYC). I have a database of 5000 customers. What will your AI do to help me understand them and predict their future behaviour ? That's the sort of information I need.

 

[Fennaio]

It's still machine learning, just a more complex machine. It's not intelligence.


I call BS.

What even is "customer journey mapping"?

So who is using the software?

It runs on a computer/machine/server but again, this is the misunderstanding between Machine Learning and AI. If a collection of artificial neurons all linked together (that builds itself) can solve multiple diverse problems and always come up with an insightful outcome thousands of times faster than a real brain can, without human involvement, then to me it's AI.

I was very sceptical too before I saw it working which is why I went from software developer to researching AI and eventually ended up creating AI. I've seen the same AI work out PI, at the same time as learning dialect, at the same time as being able to predict crime, at the same time as being able to have a conversation; very accurately. It's just a very accurate prediction machine - but actually that's how brains work - they take in lots of information, learn, predict and respond based on what they know.

Not sure what is BS about it? I'll PM you my LinkedIn profile and link to the software - it's free to use for a single user so you can see for yourself. Happy to discuss over the phone why you think it's BS.

"Customer Journey Mapping" is the ability to quickly predict what a customer will do next at every step. If you imagine the permutations of what a customer (or anyone in any process) can do if you take into account even basic info like age, postcode, last action etc (in reality there are 100s of factors) you're looking at billions of potential next steps. AI has already worked this out. If you were able to know in advance what a customer is likely to do, then you can respond more appropriately. Again it's all about accurate predictions.

I'm not publicising on here who is using the software.

 

[fisicx]

That’s not AI, that’s just machine learning.

Customer journey mapping is nothing new, companies have been doing this for years and have got very good at predicting behaviour. All the adverts you see are based on this accumulated knowledge.

 

[Fennaio]

Those aren't examples - it's just a list of things you say it can do without specifying what the input and output looks like.

Take item 2 - know your customer (KYC). I have a database of 5000 customers. What will your AI do to help me understand them and predict their future behaviour ? That's the sort of information I need.

You can have as many inputs and outputs as you like. The inputs can be things like age, postcode, salary, products-browsed, favourite colour, last twitter post (literally anything you can think of in any order and any type of data). AI will link them all together and work out how they lead to an output(s).

For an actual example which is a common use of AI - Customer Churn - how likely is a customer to come back?

In this case you get your 5000 customers in an excel/csv file and append an extra column or columns e.g. "customer never returned", "customer came back in a week", "customer came back next day". Upload that to AI. With 5000 customers and e.g. 10 inputs and 4 outputs it would learn that in about 10 seconds.

When you get a new customer, you then show AI their details e.g. via a form, or via your own system through an API and it will give you an accurate prediction of what that customer is likely to do (about 98% accuracy).

You can also ask AI to give insights and recommendations. Insights might show you that someone from a certain postcode is most likely to be your best customer. Recommendations might show you that people aged 30 behave similarly to people aged 45.

Obviously this is a loose example, but that's pretty much how it works.

I'll PM you my LinkedIn and link to the software so you can see for yourself.

 

[Fennaio]

That’s not AI, that’s just machine learning.

Customer journey mapping is nothing new, companies have been doing this for years and have got very good at predicting behaviour. All the adverts you see are based on this accumulated knowledge.

It was just one example. It's how you process "accumulated knowledge" that's the key thing. Traditional methods using database calls and programming are unreliable. Machine Learning is a lot better because it's targeted. AI (Deep Learning using Artificial Neural Networks) on the other hand is extremely powerful and thousands times more accurate.

 

[fisicx]

Great. So how will this help Dennis who has a small landscaping business in Shropshire?

That the sort of people using this forum.

 

[Fennaio]

Great. So how will this help Dennis who has a small landscaping business in Shropshire?

That the sort of people using this forum.

Sorry, I thought this was a business forum for everyone to use.

 

[IanSuth]

Have a look at qbox, i put their ceo into the proceeding company as a junior programmer (they were a web design company trying to automate marketing) over the last 18 years he came up with the idea of, managed the development of and now is CEO of the hived of company that produces an AI chatbot (or rather now they produce a tool to allow others to test and teach their own AI chatbots)

 

[fisicx]

Sorry, I thought this was a business forum for everyone to use.

It is, but you are discussing things that are unlikely to be of much use to most members. Are there any applications of AI that a small business could make use of?

 

[Fennaio]

Have a look at qbox, i put their ceo into the proceeding company as a junior programmer (they were a web design company trying to automate marketing) over the last 18 years he came up with the idea of, managed the development of and now is CEO of the hived of company that produces an AI chatbot (or rather now they produce a tool to allow others to test and teach their own AI chatbots)

Nice one. I've actually been focussing on NLP (Natural Language Processing) with my AI for the last 3 months. Took a lot of sleepless nights to get it to work convincingly. It was mainly to be used to simply read and make sense of human text in things like witness statements and fake content spotting, but we easily got the same AI to work as a chatbot. Will PM you.

 

[Fennaio]

It is, but you are discussing things that are unlikely to be of much use to most members. Are there any applications of AI that a small business could make use of?

Hard to answer without knowing about the specific business, data they hold/manage, or budget etc. I get your point though that small businesses generally won't go for AI. If you look at my OP - I just wanted to start a discussion on what everyone thought about AI. AI is relevant to everyone. I'm not here to sell it. I'm sure some people regardless of the size of their business are interested in it though.

 

[IanSuth]

Nice one. I've actually been focussing on NLP (Natural Language Processing) with my AI for the last 3 months. Took a lot of sleepless nights to get it to work convincingly. It was mainly to be used to simply read and make sense of human text in things like witness statements and fake content spotting, but we easily got the same AI to work as a chatbot. Will PM you.

If you think you did it in 3 months you have either just built off existing tech or it doesn't work as well as you think.

It took them years and a big s/w dev team based in Columbo even with major funding from Oracle

 

[IanSuth]

Nice one. I've actually been focussing on NLP (Natural Language Processing) with my AI for the last 3 months. Took a lot of sleepless nights to get it to work convincingly. It was mainly to be used to simply read and make sense of human text in things like witness statements and fake content spotting, but we easily got the same AI to work as a chatbot. Will PM you.

Also not blowing my own trumpet but i put in the original product manager for the Holmes2 project Home Office Large Murder Enquiry System, which used SGML (this was pre internet) to scan witness statements and the like and know if a detective asked for details on "a red car" it should look for any possible alternatives like "dark car", " brown vehicle" etc, and that was about 1996.

Although if you look it up online it will say written by Unisys, it was actually done by a small company nr Oxford called Status IQ who sprung out of Harwell Computer Power Limited

 

[Fennaio]

If you think you did it in 3 months you have either just built off existing tech or it doesn't work as well as you think.

It took them years and a big s/w dev team based in Columbo even with major funding from Oracle

Nope - from scratch using Python and a dynamic artificial neural network that builds itself. It was built from my own tech so strictly speaking it took 6 months as the original AI took about 3 months at its core. Tweaked it a bit. Works very well as far as I can tell.

Years and a big team sounds a lot to do this but from experience working in teams always slows things down in software dev. Too many cooks etc. Most devs work better on their own in dark rooms especially when you're not getting paid to do it.

 

[Fennaio]

Also not blowing my own trumpet but i put in the original product manager for the Holmes2 project Home Office Large Murder Enquiry System, which used SGML (this was pre internet) to scan witness statements and the like and know if a detective asked for details on "a red car" it should look for any possible alternatives like "dark car", " brown vehicle" etc, and that was about 1996.

Although if you look it up online it will say written by Unisys, it was actually done by a small company nr Oxford called Status IQ who sprung out of Harwell Computer Power Limited

Interesting. I actually know about this system as per my PM. It's not AI obviously but fantastic to be involved in that at the time.

 

[UkAppCoder]

fisicx:
Great. So how will this help Dennis who has a small landscaping business in Shropshire?

That the sort of people using this forum.

I've been excited about getting into AI for over a decade now, and I'm finally going start in a few months time. (I couldn't take the module at uni because it contained too much material which was included in my other modules.) The applications of AI are endless. For example, Dennis might be interested in the following:

Dennis finds a strange plant which looks very pretty. He takes a picture of the plant, and his mobile gardening app displays it. He taps the 'Get Information' button, and the app's AI image recognition scanner goes to work. The app's plant recognition object returns the plant's name (English and Latin), physical information (e.g. root and branch dimensions, height, spread, etc.), insects which feed on it (which could be dangerous to other plants), etc.. The module also provides a confidence percentage - i.e. a probability that the result is correct. The module then takes feedback from Denis, and uses this feedback to improve the service. The module could reject some or all of the feedback so that hackers can't ruin the knowledge base. Dennis should now be able to improve his customer service and reputation.

I'm not an AI expert, and this example is right off the top of my head, so Fennaio might be able to correct me on a couple of points above.

I've just bought this:

Hands-On Artificial Intelligence for Android: Understand Machine Learning and Unleash the Power of TensorFlow in Android Applications with Google ML Kit (English Edition): Amazon.co.uk: Veloso, Vasco Correia: 9789355510242: Books

Buy Hands-On Artificial Intelligence for Android: Understand Machine Learning and Unleash the Power of TensorFlow in Android Applications with Google ML Kit (English Edition) by Veloso, Vasco Correia (ISBN: 9789355510242) from Amazon's Book Store. Everyday low prices and free delivery on...

www.amazon.co.uk

Sounds very promising.

By the way, there's a good explanation of AI, machine learning and deep learning here:

Artificial Intelligence (AI) vs. Machine Learning (ML) vs. Deep Learning | SaM Solutions

Do you know the difference between Artificial Intelligence, vs. Machine Learning vs. Deep Learning? In this article, we will find out more about these concepts and discuss their characteristics.

www.sam-solutions.com

 

[NickGrogan]

In this case you get your 5000 customers in an excel/csv file and append an extra column or columns e.g. "customer never returned", "customer came back in a week", "customer came back next day". Upload that to AI. With 5000 customers and e.g. 10 inputs and 4 outputs it would learn that in about 10 seconds.

When you get a new customer, you then show AI their details e.g. via a form, or via your own system through an API and it will give you an accurate prediction of what that customer is likely to do (about 98% accuracy).

Not a chance. Even vague predictions aren't going to be that accurate.

This is machine learning, there's no AI here, not in the true sense of the word.

 

[IanSuth]

I've been excited about getting into AI for over a decade now, and I'm finally going start in a few months time. The applications of AI are endless. For example, Dennis might be interested in the following:

Dennis finds a strange plant which looks very pretty. He takes a picture of the plant, and his mobile gardening app displays it. He taps the 'Get Information' button, and the app's AI image recognition scanner goes to work. The app's plant recognition object returns the plant's name (English and Latin), physical information (e.g. root and branch dimensions, height, spread, etc.), insects which feed on it (which could be dangerous to other plants), etc.. The module also provides a confidence percentage - i.e. a probability that the result is correct. The module then takes feedback from Denis, and uses this feedback to improve the service. The module could reject some or all of the feedback so that hackers can't ruin the knowledge base. Dennis should now be able to improve his customer service and reputation.

I'm not an AI expert, and this example is right off the top of my head, so Fennaio might be able to correct me on a couple of points above.

I've just bought this:

Hands-On Artificial Intelligence for Android: Understand Machine Learning and Unleash the Power of TensorFlow in Android Applications with Google ML Kit (English Edition): Amazon.co.uk: Veloso, Vasco Correia: 9789355510242: Books

Buy Hands-On Artificial Intelligence for Android: Understand Machine Learning and Unleash the Power of TensorFlow in Android Applications with Google ML Kit (English Edition) by Veloso, Vasco Correia (ISBN: 9789355510242) from Amazon's Book Store. Everyday low prices and free delivery on...

www.amazon.co.uk

Sounds very promising.

Dennis can use google lens to do that already

 

[fisicx]

Dennis finds a strange plant which looks very pretty. He takes a picture of the plant, and his mobile gardening app displays it. He taps the 'Get Information' button, and the app's AI image recognition scanner goes to work.

There is already an app that does this:

PictureThis - Plant Identifier – Apps on Google Play

Identify Plants, Trees, Flowers; Personalized Care Guide; Auto-Diagnose Diseases

play.google.com

Dennis doesn't care if it's AI or some bloke busy on Google. He just knows it works.

This is the whole problem with AI and anything associated. Nobody wants AI. What they want is more business, lower costs, bigger profits and so on. How this happens is irrelevant.

The big models use millions of data points and spend months training the system. And they still get it wrong. Look at how bad facial recognition is to realise true AI is years away.

As to the statement that AI modelling doesn't use a lot of energy, this is just one of many report stating the opposite:

Energy consumption of AI poses environmental problems | TechTarget

AI energy consumption is particularly high for advanced deep learning models and is harmful to the environment.

www.techtarget.com

 

[Fennaio]

Not a chance. Even vague predictions aren't going to be that accurate.

This is machine learning, there's no AI here, not in the true sense of the word.

No, it's definitely AI, and the predictions aren't vague, they're generally over 98% accurate.

I'm glad I started this thread. The misunderstanding of AI is key feedback for me. Extremely valuable.

 

[DontAsk]

Dennis finds a strange plant which looks very pretty. He takes a picture of the plant, and his mobile gardening app displays it. He taps the 'Get Information' button, and the app's AI image recognition scanner goes to work. The app's plant recognition object returns the plant's name (English and Latin), physical information (e.g. root and branch dimensions, height, spread, etc.), insects which feed on it (which could be dangerous to other plants), etc.. The module also provides a confidence percentage - i.e. a probability that the result is correct. The module then takes feedback from Denis, and uses this feedback to improve the service. The module could reject some or all of the feedback so that hackers can't ruin the knowledge base. Dennis should now be able to improve his customer service and reputation.

Google lens and take it from there.

 

[NickGrogan]

Found the website here

eldr | AI made easy

eldr is a no code AI solution that makes Deep Learning easy. Rapidly create, deploy and integrate Deep Learning AI in any business process, app or system for AI driven decision intelligence, predictions, recommendations and insights.

eldr.ai

I'm still not convinced it's AI v machine learning. But then I think Google Lens is machine learning, and some sites are calling it AI, so maybe it's semantics?

I might throw some energy data at it and see what it thinks will happen next. No one else seems to have any idea.

 

[Fennaio]

Found the website here

eldr | AI made easy

eldr is a no code AI solution that makes Deep Learning easy. Rapidly create, deploy and integrate Deep Learning AI in any business process, app or system for AI driven decision intelligence, predictions, recommendations and insights.

eldr.ai

I'm still not convinced it's AI v machine learning. But then I think Google Lens is machine learning, and some sites are calling it AI, so maybe it's semantics?

I might throw some energy data at it and see what it thinks will happen next. No one else seems to have any idea.

Yep, that's the one.

There is definitely overlap between Machine Learning and AI, obviously they both run on computers and there is code involved. I guess from my perspective, the big difference is how they learn.

Machine Learning uses specific data-science derived algorithms and are set to solve a single problem.

Genuine AI, in my mind at least, uses models of brains made up of neurons to learn that are capable of handling all types of problems. That's the key for me.

Great feedback anyway, appreciated.

 

[Mark T Jones]

This debate about what is or isn't AI to me (who knows FA about AI)highlights how things will go.

Techies and early adopters will adopt AI for the sake of it, to wave their arms around and get noticed - often with no regard to the customer experience

Others will gradually adopt it and embed it in processes in such a way that it is barely visible

 

[Fennaio]

This debate about what is or isn't AI to me (who knows FA about AI)highlights how things will go.

Techies and early adopters will adopt AI for the sake of it, to wave their arms around and get noticed - often with no regard to the customer experience

Others will gradually adopt it and embed it in processes in such a way that it is barely visible

That's a really good point. I started off envisaging how it could be used for people and business, then to the tech side (which took a year and is key of course) and I'm now back in the mode of getting it out there and communicating how useful it is for people. All this feedback is great.

 

[fisicx]

Remove the word AI and just market an app/system/whatever that will do X. Show the benefits and reap the rewards.

Nobody cares how X is done, they just like what X does for their business.

As the saying goes: sell the sizzle not the sausage.

 

[Fagin2021]

All of the examples above are examples of Machine Learning – not AI.

Let’s try to distinguish between them without using big words.

Take for instance, Smart Bidding in Google Ads. If I ask Google to ‘maximise conversions’ it will adjust my bids for advertising space up or down depending on how likely it thinks a click on that ad’ will lead to a conversion. The greater the chance of a conversion, the higher it will raise my bid.

To assess my chance of a conversion Google needs data from at least 30 conversions in the last month (preferably many more). It analyses these conversions for customer demographics (age, location, gender etc) interests and other criteria.

Using this information, Google tries to predict who is most likely to become a customer when they search on a keyword/ keyword phrase that I’m bidding on. It then adjusts the bid accordingly.

That is Machine Learning. Predictions based on an existing set of data.

Now, I have a website that sells pearl jewellery. So my keyword phrases are things like Pearl Necklaces, Pearl Earrings and so on. Supposing someone entered the search query ‘What can I buy for my 30th wedding anniversary.”

Pearls are the traditional gift for 30th anniversaries and somewhere in the depths of it’s servers, Google knows this. If it was clever it would show my Ads for that query. But it’s not, so it doesn’t.

That’s Artificial Intelligence.

 

[Fennaio]

All of the examples above are examples of Machine Learning – not AI.

Let’s try to distinguish between them without using big words.

Take for instance, Smart Bidding in Google Ads. If I ask Google to ‘maximise conversions’ it will adjust my bids for advertising space up or down depending on how likely it thinks a click on that ad’ will lead to a conversion. The greater the chance of a conversion, the higher it will raise my bid.

To assess my chance of a conversion Google needs data from at least 30 conversions in the last month (preferably many more). It analyses these conversions for customer demographics (age, location, gender etc) interests and other criteria.

Using this information, Google tries to predict who is most likely to become a customer when they search on a keyword/ keyword phrase that I’m bidding on. It then adjusts the bid accordingly.

That is Machine Learning. Predictions based on an existing set of data.

Now, I have a website that sells pearl jewellery. So my keyword phrases are things like Pearl Necklaces, Pearl Earrings and so on. Supposing someone entered the search query ‘What can I buy for my 30th wedding anniversary.”

Pearls are the traditional gift for 30th anniversaries and somewhere in the depths of it’s servers, Google knows this. If it was clever it would show my Ads for that query. But it’s not, so it doesn’t.

That’s Artificial Intelligence.

Yes, that's exactly what my AI software does. If you typed "what can I buy for my 30th wedding anniversary" into my AI, which learns through an Artificial Neural Network, as opposed to Machine Learning algorithms it would definitely come up with "Pearl's" if that's what it knows.

Again, this is misunderstanding the difference between machine learning and artificial neural networks.

I'll say it one more time. Machine Learning is using data-science derived algorithms to solve specific problems.

Artificial Neural Networks use the same mechanism as the brain - they link literally every piece of data together via artificial neurons without being told what to look for.

It's the mechanism of learning that differentiates AI from Machine Learning.

I can guarantee that all the examples above when using my software are AI - the key thing is that my AI can do all them at the same time using the same artificial neural network. You can't do that using traditional machine learning.

Please read up on Artificial Neural Networks.

I'd love to show you.

 

[fisicx]

That’s not really a useful example. Everyone knows pearls are for 30 years so I can’t see where the AI bit comes in.

Can you set up a demo/walkthrough on your site showing how your software answers the question about what to buy for a 30th anniversary. And then how it would help a small business.

 

[Fennaio]

That’s not really a useful example. Everyone knows pearls are for 30 years so I can’t see where the AI bit comes in.

Can you set up a demo/walkthrough on your site showing how your software answers the question about what to buy for a 30th anniversary. And then how it would help a small business.

I'm putting together a load of them explainer videos you see on YouTube in the next month or so but because I can't afford a voice-over artist, they'll be narrated by my proper down to earth northern voice and not some overexcited american. Could be comical. I'll let you know when they're available.

 

[fisicx]

Great. But don’t mention AI or ML anywhere in the videos. Leave out anything about the technology that makes it work. Focus totally on the benefits to a small business of using your tools. Show an example of how you can help Fred sell more bikes.

 

[Fennaio]

Great. But don’t mention AI or ML anywhere in the videos. Leave out anything about the technology that makes it work. Focus totally on the benefits to a small business of using your tools. Show an example of how you can help Fred sell more bikes.

Yep, I get your point and I've taken it on board - has been useful to think in terms of end user requirements .

There will be a mixture of videos ranging from technical to end user, to decision makers. 30 planned in total.

The software is aimed at various groups: software developers, tech leads, ceos, ctos, big business as well as small business so it will be a case of targeting them to the appropriate audience.

Dennis the bike seller won't care about the tech behind it as you have pointed out, but Cuthbert who wants to add AI into his cybersecurity business without spending 180k developing it will want to know the tech and how it slots into his systems.