Decoding AI: Behind the Technology That's Changing Everything

The extraordinary evolution of artificial intelligence from academic theory to global phenomenon has forever changed how we view technology's potential. 

Drawing on insights from the World Travel and Tourism Council's comprehensive guide, partnered with Microsoft, we (or rather Google Notebook LMs digital AI podcasters) unpack the remarkable journey of AI from Alan Turing's foundational question "Can machines think?" to today's reality where AI systems create art, compose music, and revolutionize medical diagnosis.

TLDR:

• AI's evolution from the 1956 Dartmouth Conference to landmark moments like IBM's Deep Blue defeating chess champion Garry Kasparov in 1997
• Recent breakthroughs including Google DeepMind's AlphaFold revolutionizing protein structure prediction and AI creating art displayed in prestigious museums
• Critical challenges including "hallucinations" where AI presents fabricated information as fact, the digital divide with 2.6 billion people never having used the internet, and a skills gap with only 13% of workers receiving AI training
• The need for careful governance to ensure AI aligns with human values and creates a future that's greener, safer, more prosperous and more free for everyone

We explore the three fundamental pillars powering this revolution: sophisticated learning algorithms that find patterns humans might miss, the unprecedented explosion of data (projected to reach 175 zettabytes annually by 2025), and the specialized computing hardware making it all possible. 

From IBM's Deep Blue defeating chess champion Garry Kasparov to ChatGPT becoming the fastest-growing consumer product in history, we trace the key milestones that brought AI from science fiction to your smartphone.

The conversation doesn't shy away from critical challenges. We examine the very real problem of AI "hallucinations" - where systems confidently present fabricated information as fact i.e., make sh*t up. 

We confront the widening digital divide that threatens to leave billions behind, with 2.6 billion people never having used the internet. And we address the urgent skills gap, with IBM estimating 40% of the workforce will need AI-related reskilling in the next three years, yet only 13% of workers currently receiving AI training.

Beyond technical explanations, we wrestle with profound questions about intelligence itself. If AI can learn, create, and even "hallucinate" almost indistinguishably from humans, what does that mean for our definition of thinking? 

More importantly, how do we ensure this powerful technology aligns with our human values to create a future that's more sustainable, equitable, and free? 

Join us (or rather Google Notebook LMs digital AI podcasters) as we decode the technology reshaping our world and explore how we can consciously steer this revolution toward truly human-cantered progress.

Chapters

• 0:00 Can Machines Think? Turing's Question
• 2:10 The Explosion of AI in 2023
• 4:30 AI's Evolution: From Chess to Creativity
• 6:20 ChatGPT and the Global Tipping Point
• 10:30 The Three Pillars: Algorithms, Data, Compute
• 14:58 Types of AI and Their Applications
• 18:32 Challenges: Hallucinations, Regulation, and Skills
• 20:48 The Future: Steering AI Responsibly

Transcript

Speaker 1: 0:00

Can machines think? It's a question that well. It's puzzled humanity for decades. Alan Turing, the celebrated computer scientist, codebreaker, he posed it way back in 1950.

Speaker 2: 0:12

A profound question, then?

Speaker 1: 0:14

definitely, but it feels incredibly urgent now it really does, Because you fast forward to 2023 and artificial intelligence, AI. It just exploded into the global consciousness.

Speaker 2: 0:26

You couldn't escape it.

Speaker 1: 0:27

No, everywhere. We heard everyone, from you know, king Charles III, to the UN Secretary General, even the Pope, calling AI unprecedented, one of the greatest technological leaps.

Speaker 2: 0:36

It really did feel like that phrase.

Speaker 1: 0:37

Yeah.

Speaker 2: 0:38

Gradually, then suddenly.

Speaker 1: 0:39

Exactly Like the internet before it just burst onto the scene.

Speaker 2: 0:43

And that's why we're doing this deep dive. Our mission is really to unpack the technology behind this explosion. We're drawing on a great guide from the World Travel and Tourism Council, WTTC, partnered with Microsoft.

Speaker 1: 0:54

Right. We want to help you understand how AI actually works, where it came from the different types.

Speaker 2: 1:00

And, crucially, what it means for you, for industries like travel and tourism, because AI, fundamentally it's an engine, isn't it?

Speaker 1: 1:07

It is, and the goal has to be ensuring it fuels us where we want to go A future that's, you know, greener, safer, more prosperous, more free.

Speaker 2: 1:16

Alan Turing himself said back in 49, we might be seeing only a foretaste of what is to come, only the shadow of what is going to be.

Speaker 1: 1:22

And wow, what a foretaste of what is to come only the shadow of what is going to be, and wow, what a foretaste it's turning out to be. So let's maybe trace that journey. Ai's official birth, you could say, was the Dartmouth conference in 1956. That's where the term artificial intelligence was actually coined, but for a long time for many people it stayed kind of sci-fi Until those big moments, right Like IBM Deep Blue beating Gary Kasparov at chess in 97.

Speaker 2: 1:45

Yeah, that wasn't just technical. It really shattered assumptions about what machines couldn't do. It was a psychological blow in a way.

Speaker 1: 1:52

And then IBM Watson winning Jeopardy in 2011. That felt different. Again Trained on huge amounts of data.

Speaker 2: 1:57

Another huge turning point. It forced people to see AI not just as theory, but as something real, capable of well complex intellectual tasks.

Speaker 1: 2:07

Okay, but now this is where it gets really interesting, isn't it the last decade? Yeah, the applications are just mind-boggling.

Speaker 2: 2:13

Take Google DeepMind's AlphaFold Predicting the 3D structure of proteins. Almost every protein.

Speaker 1: 2:20

Which is huge for medicine, right Vaccines, diseases like Alzheimer's.

Speaker 2: 2:24

It's more than just a tool. It's like a paradigm shift in how we do science, unlocking biology computationally. Things that took decades now take months. It just speeds up human knowledge.

Speaker 1: 2:35

It really makes you wonder where are the boundaries. Now, AI is moving into areas we thought were uniquely human, like creativity.

Speaker 2: 2:43

That's a massive question. What is creativity if a machine can do it?

Speaker 1: 2:46

Remember that AI painting Edmund de Bellamy sold for what? Over $400,000 back in 2018?.

Speaker 2: 2:52

Caused a huge stir, but 2023 just accelerated everything.

Speaker 1: 2:56

Like the museum in the Hague.

Speaker 2: 2:58

Yes.

Speaker 1: 2:58

Swapping Vermeer's girl with pearl earring for an AI piece girl with glowing earrings temporarily but still.

Speaker 2: 3:04

And apparently a lot of visitors didn't even notice the difference at first.

Speaker 1: 3:07

And then the Sony World Photography Award Boris Eldakson wins.

Speaker 2: 3:10

Only to refuse it, revealing his photo Pseudomnesia, the electrician was pure AI.

Speaker 1: 3:16

The key thing here, I think, is that this kind of advanced software it's available to everyone now.

Speaker 2: 3:24

It's not locked away in labs. Exactly, it's on your phone, your laptop. Democratized in a sense.

Speaker 1: 3:28

And it's not just visuals Music too. The Grammys saying AI music is eligible.

Speaker 2: 3:33

And that final Beatles song Now, and Then using AI to pull out John Lennon's voice Incredible.

Speaker 1: 3:39

Plus artists like Grimes, actively inviting people to use an AI version of her voice, sharing royalties.

Speaker 2: 3:46

It's fundamentally reshaping creative industries, no question.

Speaker 1: 3:49

But there's always a but isn't there. With this kind of power comes risks, concerns.

Speaker 2: 3:54

Definitely. We saw it very clearly with the Hollywood strikes in 2023. Actors, writers worried about AI taking jobs, creating digital doubles.

Speaker 1: 4:03

That really throws the spotlight on things like job displacement, the need for reskilling. It's urgent.

Speaker 2: 4:08

And then there's the whole disinformation problem. It's serious.

Speaker 1: 4:11

Those fake images the pup in a puffer jacket, trump being arrested they looked real.

Speaker 2: 4:15

Generated easily, quickly and they go viral. It just shows how badly we need well guardrails, regulations, codes of conduct, something.

Speaker 1: 4:24

Although, interestingly, not everyone sees it purely as a threat. You mentioned Lionel Messi.

Speaker 2: 4:29

Right Using a deep fake of himself for PepsiCo ads Personalized messages, so some are embracing it.

Speaker 1: 4:35

Okay. So what was the tipping point? The moment it went from niche tech news to, well everywhere.

Speaker 2: 4:41

Undeniably, it was OpenAI releasing ChatGPT November 2022.

Speaker 1: 4:46

That felt like the moment yeah, Fastest growing consumer product ever.

Speaker 2: 4:50

Hit 100 million users in two months. It just sparked this global wave of interest. Put AI in the hands of ordinary people. Very different from the 90s, when you needed a supercomputer.

Speaker 1: 5:00

And now we have Google, bard, microsoft Copilot, baidu's Ernie. These sophisticated chatbots are becoming commonplace.

Speaker 2: 5:07

Right. So the big question why now? Why this sudden explosion?

Speaker 1: 5:12

Well, you said, ai is, like electricity, a multi-use tech, and its recent boom comes down to three key things working together right.

Speaker 2: 5:19

That's a good way to put it. Think of it as algorithms that's the how. Data, that's what they learn from. And Think of it as algorithms that's the how. Data, that's what they learn from. And computing power that's the muscle making it all happen fast enough.

Speaker 1: 5:28

OK, let's unpack those. Pillar one Algorithms, the brains. How is AI learning different from normal programming?

Speaker 2: 5:36

Well, traditional programming is all about explicit rules. If this happens, then do that LSE, do this other thing.

Speaker 1: 5:43

Like your simple digital doctor example. If fever then prescribe drug X. Good for predictable stuff like a tax calculator.

Speaker 2: 5:50

Exactly, but AI systems, they don't follow rules we write down explicitly. They learn from data.

Speaker 1: 5:57

So the AI digital doctor trained on millions of scans, it learns to spot tumor patterns itself without being told look for this exact shape.

Speaker 2: 6:05

Precisely. It identifies patterns, sometimes ones humans might miss or ones too complex to code as simple rules.

Speaker 1: 6:11

And the results are pretty staggering. That Swedish study in 2023, AI helping screen for breast cancer.

Speaker 2: 6:16

Yeah, as good as two radiologists cut workload by half and detected 20% more cancers early. That's saving lives.

Speaker 1: 6:24

But it brings up that black box issue, doesn't it? We see the result cancer detected, yes or no, but we don't always know how the AI decided.

Speaker 2: 6:31

That's a huge challenge. The AI figures out the interim steps itself, so there's a whole field now explainable. Ai trying to make these systems transparent turn black boxes into glass boxes.

Speaker 1: 6:42

Which you need for trust, especially in medicine or finance.

Speaker 2: 6:45

Absolutely so. How does it learn? We hear terms like machine learning, deep learning.

Speaker 1: 6:50

Can you break those down simply?

Speaker 2: 6:52

Sure, machine learning ML is the broader concept. Systems learn from data to make predictions or find patterns. They get better over time with more data. Okay, Deep learning DL is a subset of ML. It uses these complex structures called artificial neural networks Think layers of virtual neurons.

Speaker 1: 7:09

Kind of inspired by the human brain.

Speaker 2: 7:11

Inspired. Yes, but it's really about complex math, adjusting weights between connections to recognize incredibly intricate patterns. That's what powers the really advanced stuff image generation, drug discovery.

Speaker 1: 7:22

And the pioneers here, the godfathers of AI.

Speaker 2: 7:25

Jan LeCun, Jeffrey Hinton, Yoshua Bengio. Their work in the 90s, early 2000s, foundational for deep learning. They won the Turing Award for it.

Speaker 1: 7:34

The Nobel Prize of Computing named after Alan Turing himself. His Turing test from 1950, still relevant today.

Speaker 2: 7:40

It really is and in terms of how they train. There are main ways. Supervised learning you give it labeled data this is a cat, this is a dog, got it. Unsupervised learning you give it unlabeled data and it finds patterns itself, maybe groups similar animal pictures together.

Speaker 1: 7:54

Okay.

Speaker 2: 7:55

And reinforcement learning trial and error. The AI gets rewards or punishments for its actions. Learning the best strategy over time like learning to play a game.

Speaker 1: 8:04

Right, so that covers the brains, the algorithms. Now the fuel data. Data is the new oil. We hear that all the time.

Speaker 2: 8:11

And it's true, it's essential for modern AI. It's what the algorithms learn from, find relationships in, make predictions based on.

Speaker 1: 8:18

The amount of data is just mind-blowing. That stat from 2013,. 90% of the world's data created in the last two years.

Speaker 2: 8:25

Then and it's only accelerated by 2025, the estimate is 175 zettabytes a year 175 zettabytes.

Speaker 1: 8:32

What does that even mean?

Speaker 2: 8:33

It's huge. Think 175 billion high-end laptops worth of data, or about 21 laptops for every single person on Earth per year.

Speaker 1: 8:43

That's staggering. It's getting so big. We needed new words for it we did.

Speaker 2: 8:47

In 2022, scientists agreed on the ronobite and the quetobite, First new prefixes for data size in over 30 years.

Speaker 1: 8:53

A ronobite is 127, quetobite 1030. A one with 30 zeros.

Speaker 2: 8:58

Wow, and they think we'll need them sooner rather than later. That's how fast it's growing.

Speaker 1: 9:01

What's driving this explosion?

Speaker 2: 9:03

Three main things. One more devices Smartphones, internet of Things, sensors nearly 30 billion network devices in 2023.

Speaker 1: 9:10

IoT that's like smart fridges, watches, sensors and cities.

Speaker 2: 9:18

All generating data. Two social media Think 100 billion WhatsApp messages daily back in 2020.

Speaker 1: 9:21

Tiktok, youtube Huge data generators and three.

Speaker 2: 9:22

Just new ways to collect and store it all. Better sensors, cheaper storage, all adding up to this phenomenon we call big data.

Speaker 1: 9:28

Big data.

Speaker 2: 9:29

That's about the volume, velocity, variety, the Vs Exactly, and this big data is precisely what fuels modern AI. It provides the massive data sets needed for well accurate self-driving cars, reliable medical diagnoses better predictions for businesses. Personalized shopping and natural language, processing, understanding and generating human language. It all relies on massive amounts of data.

Speaker 1: 9:55

So how do businesses actually handle all this data? Where do they keep it?

Speaker 2: 9:59

Cloud computing is key for most and you generally see two main storage approaches Data warehouses first.

Speaker 1: 10:07

Warehouses sound organized.

Speaker 2: 10:08

They are. They source, structured data, spreadsheets, transaction records Great for specific questions like show me Caribbean bookings last year.

Speaker 1: 10:14

Okay, and the other?

Speaker 2: 10:15

Data lakes. Think of these as well big, less structured pools. You can put any type of data in there social media posts, videos, customer emails, sensor readings.

Speaker 1: 10:25

Even if you don't know how you'll use it yet.

Speaker 2: 10:26

Exactly. You store it now, figure out how to analyze it later. And often businesses use a hybrid, a data lake house, combining the best of both.

Speaker 1: 10:35

And for travel and tourism. Combining AI with these data stores, that unlocks a lot.

Speaker 2: 10:41

Huge potential Optimizing prices, predicting what travelers want, understanding sentiment from reviews, improving customer service, truly personalized experiences.

Speaker 1: 10:51

Like a superpower travel agent. Okay, so we have algorithms and data, pillar three computing power, the muscle.

Speaker 2: 10:59

And you need serious muscle Training. These big AI models means processing literally hundreds of billions of words.

Speaker 1: 11:06

You mentioned, the Bible has less than a million words, so this is vast.

Speaker 2: 11:10

Vast. On a normal computer it could take hundreds of years.

Speaker 1: 11:12

And learning everything on the internet. 5.75 billion pages would take a human 11,000 years.

Speaker 2: 11:20

Roughly yeah. But these advanced supercomputers, using specialized chips and deep learning, they can digest that amount of info in months.

Speaker 1: 11:26

That speed up is incredible. It links back to Moore's law right. Gordon Moore's observation in the 60s.

Speaker 2: 11:31

That was foundational the doubling of transistors on chips. But for AI's recent leap, we needed something more specific than general processors.

Speaker 1: 11:38

Which brings us to GPUs.

Speaker 2: 11:40

Exactly Graphics processing units. Nvidia really popularized them in the 90s, mostly for video games.

Speaker 1: 11:45

Hollywood effects Making things look realistic.

Speaker 2: 11:48

Right, but the key was how they worked. Processing tasks in parallel, lots of calculations simultaneously, not one after another.

Speaker 1: 11:56

And that turned out to be perfect for AI.

Speaker 2: 11:58

Perfect AI training involves millions of simultaneous calculations. Gpus were ideally suited, and that's why NVIDIA became a trillion-dollar company in 2023.

Speaker 1: 12:07

This need for specialized chips. It's become a huge geopolitical issue, hasn't it?

Speaker 2: 12:12

Massive. Only a few places can make the most advanced ones. So you see countries scrambling, china's Made in China 2025, the EU Chips Act, the US Chips and Science Act.

Speaker 1: 12:22

Billions being invested to build local capacity. Intel CEO said no innovation without semiconductors.

Speaker 2: 12:28

He's right, and the concentration is stark. About 70% of all chips, and even more of the most advanced ones, come from just two companies.

Speaker 1: 12:35

TSMC in Taiwan.

Speaker 2: 12:36

With about 55% market share, and Samsung in South Korea Around 15% this concentration.

Speaker 1: 12:45

Well, we saw the impact during the pandemic, with car production grinding to a halt because of chip shortages and it raises huge questions about supply chains resilience, especially concerning Taiwan. It's a major global vulnerability, Definitely.

Speaker 2: 12:54

Which is also why big tech companies are now designing their own AI chips. Global vulnerability, definitely, which is also why big tech companies are now designing their own AI chips. Yeah, amazon, apple, google, microsoft, tesla.

Speaker 1: 13:01

Like Google's TPU Tensor Processing Unit.

Speaker 2: 13:05

Exactly Custom designed for AI tasks. They're incredibly efficient at the matrix calculations core to neural networks. It speeds up machine learning dramatically, powering things like Google Translate.

Speaker 1: 13:16

But building the hardware to run these things, that's another level. The supercomputer Microsoft built for chat, gpt.

Speaker 2: 13:22

Immense scale 10,000 GPUs, millions of dollars took up warehouses.

Speaker 1: 13:27

Which raises that concern again the digital divide. If this costs so much, does it mean only the richest companies or countries can really develop cutting edge AI?

Speaker 2: 13:35

It's a very real risk and it connects to basic access too. Remember the ITU estimates 2.6 billion people have never used the internet.

Speaker 1: 13:42

A third of the world completely cut off from this technology.

Speaker 2: 13:45

And affordability is a huge barrier.

Speaker 1: 13:47

Yeah.

Speaker 2: 13:47

In low-income countries, just getting basic mobile broadband can eat up nearly 10% of average income.

Speaker 1: 13:53

Six times the global average, so 64% of people in the least developed countries are offline.

Speaker 2: 13:59

It highlights that AI could easily widen global inequalities if we're not careful.

Speaker 1: 14:05

Are there efforts to counter this?

Speaker 2: 14:06

There are initiatives. The UK and Canada announced an AI for development plan focusing initially on boosting skills and computing power in Africa.

Speaker 1: 14:15

Trying to make sure the future, as William Gibson said, gets distributed a bit more evenly.

Speaker 2: 14:20

That's the hope, Because right now it's definitely not evenly distributed.

Speaker 1: 14:24

Okay, but for businesses, say in travel and tourism, they don't need to build their own multi-million dollar supercomputer, right?

Speaker 2: 14:31

No, thankfully not Cloud providers like Google, microsoft, amazon. They offer access to pre-trained AI models and data tools.

Speaker 1: 14:40

So companies can take these powerful base models and fine-tune them, adapt them with their own specific business data.

Speaker 2: 14:46

Exactly it makes advanced AI much more accessible than it might seem. You don't need to start from scratch.

Speaker 1: 14:51

So, wrapping up the pillars, it's algorithms plus data, plus compute. That's the engine.

Speaker 2: 14:56

It's algorithms plus data plus compute. That's the engine, that's the core equation driving what we're seeing today and also where we see some of the biggest challenges, like that digital divide.

Speaker 1: 15:04

OK, let's shift gears a bit and talk about the types of AI. How do we categorize them? One way is by capability.

Speaker 2: 15:10

Right. First you have artificial narrow intelligence, or ANI. This is the AI we have today, mostly.

Speaker 1: 15:17

Designed for specific tasks.

Speaker 2: 15:19

Exactly Smart speakers answering questions, AI playing chess systems, detecting cancer on scans. They're very good at one thing or one narrow set of things. This is what businesses are using now for automation.

Speaker 1: 15:31

And the other type, the one we see in sci-fi.

Speaker 2: 15:34

That's artificial general intelligence, agi, the theoretical idea of an AI that can perform any intellectual task an average human can.

Speaker 1: 15:42

Thinking, reasoning, learning across domains.

Speaker 2: 15:45

Yes, we are definitely not there yet. Some chatbots show maybe tiny sparks of generality, but true AGI is still hypothetical. It would likely need embodiment to a physical or virtual form to interact with the world.

Speaker 1: 15:57

Okay, so that's by capability. What about by functionality? How they work?

Speaker 2: 16:01

Well, for years, a key type was AI expert systems.

Speaker 1: 16:04

Simulating human expertise.

Speaker 2: 16:06

Pretty much Imagine a travel expert system. It has a knowledge-based destinations, hotels, rules and an inference engine to reason with that knowledge.

Speaker 1: 16:15

So it could give personalized recommendations integrate with booking systems Right.

Speaker 2: 16:20

Offering consistency, cost savings drawing on a vast pool of expert knowledge digitally.

Speaker 1: 16:25

Then there's predictive AI.

Speaker 2: 16:27

Using past and present data to forecast the future, Like an airline predicting flight delays based on weather maintenance, airport traffic.

Speaker 1: 16:34

Allowing them to be proactive, maybe rebook passengers early.

Speaker 2: 16:37

Exactly, minimize disruption. It fits into that broader analytics picture.

Speaker 1: 16:41

Yay.

Speaker 2: 16:41

Descriptive what happened. Diagnostic why. Predictive, what will happen. And prescriptive what should we do?

Speaker 1: 16:47

Okay, and then the one that really grabbed headlines recently generative AI, gen AI.

Speaker 2: 16:53

Yes, the relatively new form that burst onto the scene with ChatGPT, its power, is creating new content.

Speaker 1: 17:00

Text images, video music based on a user prompt.

Speaker 2: 17:04

Correct, and these often use what are called foundation models Huge models trained on broad data, adaptable for many tasks writing, marketing, copy, building websites, coding.

Speaker 1: 17:14

The investment in research here just exploded, didn't it?

Speaker 2: 17:17

Phenomenal growth. Scientific papers up fivefold VCs, pouring in billions just in the first half of 2023.

Speaker 1: 17:24

And underlying many of these generative systems are large language models, LLMs.

Speaker 2: 17:29

That's the core architecture for understanding and generating human-like text. They power the sophisticated chatbots.

Speaker 1: 17:35

They understand natural language, not just keywords.

Speaker 2: 17:38

Yes, and they can track context. You ask about Usain Bolt's record, then ask how does that compare to a cheetah?

Speaker 1: 17:44

It knows that refers to Bolt's record without you repeating it.

Speaker 2: 17:47

Precisely. They can translate, summarize complex documents, create tables from text. Very powerful language capabilities.

Speaker 1: 17:54

There's a big, but here we have to stress Hallucinations.

Speaker 2: 17:58

Absolutely crucial limitation. Llms can and do generate answers that sound totally plausible, very confident, but are just wrong.

Speaker 1: 18:06

Made up facts, incorrect statistics, false information presented as truth.

Speaker 2: 18:11

The WTTC report itself gives an example A chatbot inventing publication dates for reports. It looks convincing but it's fabricated.

Speaker 1: 18:19

So the message is verify, especially for anything important.

Speaker 2: 18:23

Essential For businesses safety critical applications. You must check the output, don't just trust it blindly. It's a major challenge for reliability.

Speaker 1: 18:34

Which is why we're seeing all these global discussions right the UK AI Safety Summit, the EU AI Act.

Speaker 2: 18:40

Yes, trying to figure out the opportunities but also manage the risks. The OECD identified seven common challenges nations face, emphasizing we need international alignment involving everyone, including sectors like travel and tourism.

Speaker 1: 18:53

And it's not just about the tech or the rules. It's about people. We mentioned the digital divide in access, but there's also a skills gap.

Speaker 2: 19:00

A huge one that, randstad survey, only 13% of workers offered AI training in 2023, despite job postings needing AI skills soaring.

Speaker 1: 19:09

And the training isn't always reaching the right people. Gitlab found lower-level staff often the most worried about jobs get less quality training.

Speaker 2: 19:15

It's a real mismatch. Ibm thinks 40% of the workforce will need AI-related reskilling in the next three years 40%, that's massive. It's a workforce transformation challenge for every country, every sector.

Speaker 1: 19:27

What are the solutions? How do people get these skills?

Speaker 2: 19:32

Well, there's a growing number of options. Big tech companies Microsoft, google, ibm offer free introductory courses, often via platforms like Coursera or LinkedIn Learning.

Speaker 1: 19:41

IBM is aiming to train 2 million people by 2026, focusing on underserved groups.

Speaker 2: 19:46

Yes, with courses, even on things like prompt writing, how to talk effectively to AI and people are learning informally. 2 billions of views for AI tool tutorials on YouTube.

Speaker 1: 19:58

Companies are using internal training, peer mentoring.

Speaker 2: 20:01

It all helps, but fundamentally AI education needs to be broader. It's not just business's job. Governments need to integrate digital literacy and AI skills into schools, support lifelong learning.

Speaker 1: 20:11

Especially in developing nations, to make sure AI closes gaps, not widens them.

Speaker 2: 20:15

It has to be a shared responsibility, absolutely.

Speaker 1: 20:18

Okay, thinking about how good AI is getting at generating content. To make sure AI closes gaps, not widens them, it has to be a shared responsibility. Absolutely Okay, thinking about how good AI is getting at generating content. That quiz in the WTTC report sounds fascinating Testing, if you can tell human from AI.

Speaker 2: 20:27

It's a great illustration and the punchline, without giving the specific examples away.

Speaker 1: 20:32

Is that all the examples they gave text and images were AI generated?

Speaker 2: 20:36

Exactly, it really hammers home the power, maybe the subtlety of this technology, how convincing it can be.

Speaker 1: 20:43

It really does, which brings us full circle back to Alan Turing's question from 1950.

Speaker 2: 20:48

From can machines think to where we are now With AI learning creating, powered by these incredible algorithms, data and compute. It's undeniably transformative.

Speaker 1: 20:59

So that engine analogy again if we steer it right, ai can help us build that greener, safer, more prosperous, more free future for travel and tourism for everyone.

Speaker 2: 21:07

But it requires conscious effort, careful steering. It won't happen automatically.

Speaker 1: 21:10

So let's leave our listeners with this thought If AI can learn, create, even hallucinate, almost indistinguishably from humans, if thinking is happening in machines, what does that do to our definition of intelligence itself?

Speaker 2: 21:24

And maybe more practically, as we embrace this powerful new era, how do we ensure AI aligns not just with efficiency or profit, but with our core human values, our ethics, globally?

Speaker 1: 21:35

How do we make sure this engine truly fuels us towards that destination of our choosing? Lots to think about.