The Future of AI, Humanity and Work

Intro 0:01

0:01 Hi everyone, my name is Matthew Griffin. I'm the CEO and founder of the 311 Institute, a global futures and deep futures think tank looking up to 50 years out. And thank you for joining me for this special presentation on the future of humanity and artificial intelligence. Now I've got this lined up for you. Now we're going to be going through a variety of different topics. So firstly, we're going to be having a look at an artificial intelligence state of the union.

0:21 So what is artificial intelligence and what is it capable of today? So these are the systems that you're using today and there'll be some surprises here for you. No doubt we're going to be having a look at the future of artificial intelligence. So that's where we look really up to 20 years out and again we've got plenty of surprises in store for you there. We're going to be looking at how artificial intelligence is disrupting some quite fundamental things when we think about our world.

0:46 So that's everything from artificial intelligence inventions all the way through to how it's starting to disrupt health care and other industries. Then we're going to be having a look at artificial intelligence for personal development. Now this is kind of a bit of a twist on the future of work, jobs and skills. And really what I'm going to be doing here is I'm going to be taking an individual's bird's-eye view of how you ensure you are not left behind by the future even as the future itself becomes increasingly chaotic, complex and fast-paced.

1:26 Now when we actually have a look at the future, I kind of often joke that as humans we only really need to get our heads around one thing. Just one thing, right? Yeah, unfortunately the thing that we need to get our heads around the only thing that is actually changing is everything and that pace of change is getting faster. Now we kind of call this the VUCA world.

1:41 So volatile, uncertain, chaotic and ambiguous. However, the Max acronym really came in a little while ago because increasingly the changes that we're seeing around the world are massive. They are accelerating and they're increasingly exponential. Now exponential doesn't just relate to technology. It also relates to the speed of change. And this is what we're going to be showing you throughout this particular presentation. Now when we think about AI, Demis Hassabis quoted this about a couple of years ago.

2:21 He said solve intelligence and then use that to solve everything else. And this has really been his guiding mission since he was a boy. And in some respects, we're starting to get closer to his vision of solving intelligence. And this is what we're going to be going through. Yeah. As well as other things. Now, one of the things I get asked very regularly as I travel around the world is as we actually look forward to a future of artificial general intelligence, but also artificial super intelligence where ASI potentially has an IQ of 1.6 million circa 2040 2045.

3:08 What's the place of humans in humanity? So being human first and human centered and a human I've given this a lot of thought over the past year and this is what I think it is to be human. So I think being human in a world of super intelligences because there'll be more than one is to live with purpose. It's to belong that's belong within a community. Is to support so to support our family, our friends with our communities and one another as good global citizens should we say and then to scale progress.

3:35 Now the scale progress I think is particularly important because there are plenty of people who are on the one hand yes they're scaling progress. If you look at artificial intelligence we are scaling it to billions of users. So that's one scale. However, there are a lot of people that are still missing out. So when I talk about scaling process or progress rather I talk about scaling progress to everybody everywhere so that everybody can benefit.

4:12 Now when we think about exponential technologies Peter Diamandis but he is sort of relatively famous for saying when we look at artificial intelligence AI from an intelligence perspective at least is improving at an exponential rate. However, as humans, we haven't had a software upgrade for over 20,000 years. Our brains are the same as they were in the Neanderthal days. To which my reply is, education is our upgrade.

4:41 So, if you want to upgrade your mind, it's fairly obvious. You go and learn about new things. You explore new things, you get curious, you experiment, you play around and so on. So now when we think about artificial intelligence, this is very much where we are today. Now, Satya Nadella, the CEO of Microsoft, says that he's seen over 30 years of artificial intelligence progress in just three.

AI benchmarks and performance 5:06

5:13 Now, we did go through a 30-year AI winter. That's it. So, between sort of the 1980s and 2010s, but now we're really starting to cook on gas, as we say. And recently, we've seen over $2.7 trillion pouring into the development of these artificial intelligences with AIs like GPT-5 costing over $1 billion in compute to train. And that training cost doesn't actually include the cost of the hyperscale data centres, the GPUs and everything else that we've actually had to use.

5:44 It's just the raw computing cost. Now this is an interesting quote as well. We've seen this we saw this originally with McKinsey and then we saw this coming through from Microsoft. We saw it a little bit with Google as well and now we've seen it with Silicon Valley. All of these individual organizations as well as regions basically feel that artificial intelligence is an existential threat to their business.

6:14 Now we do talk about AI being an existential threat to humanity and yes as a self-developing technology now it can develop its own bioweapons and we've seen a couple of those. However, in this particular case we're talking about the role of artificial intelligence as an existential threat to a company like McKinsey. Now McKinsey, it's fairly obvious. You know, McKinsey are famous for writing large consulting reports for tens or hundreds of millions of dollars and artificial intelligence is able to do that very nicely within a couple of minutes.

6:46 Now, and that's in their own words as well as mine because I've actually shown companies like KPMG and a lot of my clients how we actually do that. Now, Microsoft is a very different animal. You would have thought that artificial intelligence would not be an existential threat to Microsoft, but there's kind of two things here. So firstly, artificial intelligence changes the business model of software and it so we go from SaaS where we pay for a subscription to agentic artificial intelligence where essentially we're paying essentially agents when they complete tasks.

7:18 So we go from a subscription-based model to a task-based model. So that's kind of one area. Secondly, really when we talk about an existential threat to Microsoft, increasingly AI is able to develop software and clone Microsoft software, which I'll talk about in a little while at speed. So today Microsoft's moat is very much the software and services that it takes to market if AI can clone all of those very rapidly then what's the moat and recently we saw Anthropic's Claude code for 30 hours continuously and clone Slack and clone Microsoft Teams.

8:19 Now, as Jamie Dimon has sort of said, you know, so I get asked a lot, you know, what are the main use cases of artificial intelligence? Well, Jamie Dimon here had a great answer. It's everything. That's it. Now, when we think about scorecards for artificial intelligence, as humans, we love anthropomorphizing technology. So, we also love trying to compare ourselves with technology. And with artificial intelligence, while the comparisons aren't actually direct, we're using traditional human metrics and human scoring systems to try to rate whether or not AI is better than us or worse than us in particular things.

8:54 This is the scorecard that we have at the moment. So the average human IQ globally is 100. Gemini 2.5 has a measured IQ of 136 and GPT-4 had a measured verbal IQ of 155. Now unlike humans, AI's IQ has increased by 51% in 3 years. Now when we start fast forwarding over the next couple of years, there are quite a lot of people in the research community that expect AI to have an IQ of between two and 300.

9:34 And then we start scaling from there. Now, at the moment, artificial intelligence does not actually have any EQ. So, about a third of humans actually are very strong when it comes to EQ. And you all know who they are. They're the charismatic types. They're keynote speakers on the stage who everyone loves. Anyway, when we think about memory as humans, basically we can typically remember seven things in our short-term memory and we've got about a petabyte equivalent of storage for long-term memories.

10:00 But increasingly when we think about context windows, artificial intelligence's memory, partly thanks to compute, memory, RAM, storage, etc., etc., is increasingly unlimited. Now this means that AI in Sam Altman's words at some point will be able to remember your entire life. Now this becomes interesting from a retail perspective as well as a getting to know you perspective because artificial intelligence today is able to gather over 165,000 separate data points on you.

10:32 Bring those together and essentially combine those into a digital twin of you. I said that we can then use to predict behaviors, different psychological responses and so on and so forth. So when we think about knowledge, if we have a knowledge of one as a human, then ChatGPT has a thousand times more general knowledge than any human. It also has a learning speed according to Geoffrey Hinton who's the godfather of artificial intelligence that is 300 million times faster than any human.

11:10 So these numbers are already getting quite bonkers and we're only in 2025 and this is really before the $2.7 trillion worth of artificial intelligence, data centres, energy, compute, nuclear power stations and everything else comes online. Now when we actually think about processing speed as humans we can process information at a rate of about 50 bits per second. AI it's 100,000 times more than that. And again this is only going to go one way.

11:38 However when we talk about new skills this is actually where the ball is still in the human court. Artificial intelligences are still really quite bad at picking up new skills. And if they do pick up a new skill, they've got to have significant data sets in order to be able to do that. But as an industry, especially Silicon Valley and China, we're trying to get around this problem by developing few-shot learning, which essentially takes us towards things like general purpose robots that you simply show them once what they need to do and they mimic it and then they learn from there.

12:14 And also artificial general intelligence. So increasingly what we're trying to do with especially AI is create more general-purpose kind of artificial intelligences that don't need anywhere near as much training data as they have in the past. Now when we think about idea quality artificial intelligence still isn't that creative but it is actually getting better. In fact its idea quality according to Stanford is eight times better than the average human.

12:45 When we think about the novelty of those ideas, yes, it beats humans again, 5.6 out of 10 compared to 4.8 out of 10. However, what's most interesting, and this is really sort of what I lean into later on in this presentation, is when you start using these tools, AI is still very much a tool to amplify human skills, the combination of humans with artificial intelligence actually produces a score of 5.8. And then when we think about Humanity's Last Exam, which essentially is the exam that has been put together by around 400 of the world's top experts to try to figure out when do we reach AGI, then humans, we're still beating AI in the Humanity's Last Exam, 26.6% compared to 26.5.

13:44 So we're pretty much neck and neck. But again, look at the increase. So artificial intelligence has increased its score in the past four months by 476%. Now it doesn't necessarily tell all of the story because some of these large foundation labs are actually trying to teach their artificial intelligences to answer the exam questions when they already know what the exam questions are. It's a little bit like giving your kids the exam paper and also the answer sheet to the exam paper.

14:14 So when we think about baselines, we do need to be very careful here. Now one of the biggest innovations really that isn't talked about you know we talk about you know ChatGPT for example having two trillion parameters and so on and so forth and DeepSeek having about 600 billion parameters etc.

Evolution of AI agents 14:23

14:30 In my estimation, the biggest disruption has been the introduction of true behavioral interfaces. And these are your friend. And I'm going to tell you again why later. And I'm going to give you a guide on how to use these to remain relevant. So it's now possible to talk to essentially the technologies around us using natural language. You don't have to be a programmer, a coder. You don't have to be fluent in Python.

15:07 You don't have to be really a fluent product engineer etc. So this is really the world changer. Now when we think about Grok 4 and when we think about GPT-5 both of these platforms reached PhD levels of knowledge. Dario, who's the CEO of Anthropic, which produces Claude, said at Davos at the start of this year, he said very soon, everyone will have access to a data centre full of PhD experts and now we seem to be there.

15:41 Now again you've got to be quite skeptical about some of the reports coming out of these large labs but generally these systems are very near PhD level and in some cases post-PhD level so they are getting much better and in fact Sam Altman's main ambition ultimately is to get to the point where he's developed a robo scientist that can just innovate new things autonomously and automatically. We've also seen from Google DeepMind again under Demis we see that artificial intelligence is able to create its own knowledge. So information plus information as humans we take that combine that together and say I have new knowledge.

16:24 Artificial intelligence has done very much the same thing. Albeit in the chess space and the board game space where it took two sets of information, combined them together and generated, in DeepMind's own words, new knowledge. So what happens from a human perspective when we have AI generated knowledge and human generated knowledge and then after knowledge comes wisdom and after wisdom arguably comes sentience. Although when we think about sentience and artificial intelligence it would still be synthetic sentience.

17:05 We wouldn't be talking about the development of conscious machines. But from an artificial intelligence perspective, it's highly unlikely we are ever going to get conscious or sentient machines in the way that we humans think about that. It will typically be a mimic or it will be a synthetic or different version. Now when we think about OpenAI again this is GPT-4 especially we've actually seen artificial intelligence bootstrapping its own intelligence in its own IQ.

17:53 So in this particular case and there are two examples, OpenAI's artificial intelligence learned from a data set. Once it had run out of that data set, once it had completed training from that data set, it then generated a bunch of synthetic data and started learning new things from the new data that had been created. So we have that. However, we also had the MIT SEAL program. So SEAL as in the marine mammals.

18:18 Now that particular artificial intelligence was able to change its own model weights as it wanted. So by changing model weights by developing synthetic data, we've seen two AIs that in the words of the researchers behind these AIs have now bootstrapped their own intelligence. So now think about that. You know what happens when AI's intelligence and AI's IQ, for better or worse, is not linked to human activity.

18:59 AI is also the world's first self-developing technology. Now, it's been designing itself. We've seen that with Google DeepMind. We've seen that with Meta and Llama and so on and so forth. It's been able to optimize its own algorithms. We've seen that, as I've said, with MIT SEAL. We've seen it with Meta, we've seen it with Microsoft, we've seen it with Google. But it's also been able to replicate itself.

19:16 And again, we've seen this, this was about two years ago with Google where an AI created a new AI and the new AI was an image was an image generation AI and it was 30% better than anything that the humans had actually put together. So when you think about the consequences of this and Google actually now have a dedicated artificial intelligence called AlphaEvolve. What happens to anything?

19:56 When artificial intelligence is able to evolve itself as it already is. Now when we think about state of the union at core of course we have agentic artificial intelligence. Now you can think of agentic artificial intelligence as lots of individual bots being given a task basically and they complete that task. They go from one end to the other. They decide what actions to take but then they also are able to take actions themselves.

20:20 So agentic AI is not the same as traditional deep learning or machine learning artificial intelligence. And what we've actually found is up to a thousand there's a little story here, but up to a thousand artificial intelligence agents can collaborate with one another to get consensus. However, the smaller the artificial intelligence, the fewer AI agents can collaborate to get consensus. So a small AI will maybe only have 10 agents that can collaborate together to get a consensus.

21:03 But the larger artificial intelligences again like the ones we see from OpenAI predominantly because of computer resources, the way that they've been designed and built and so on so forth can go up to a thousand. Once we get past a thousand agents, then consensus starts falling apart. Now when you think about this in human terms once we get past 100 humans increasingly it's more and more difficult to get consensus.

21:27 We see this with the European Union in Brussels all the time. So when we think about agents Mark Zuckerberg talks about a world full of billions of artificial intelligence agents on the one hand that's an opportunity because these agents can be given a task and they just go off and complete it and say I've finished that. Here it is. But it could also be a problem because at some point we're going to end up seeing some degradation.

21:52 And the result of that degradation could be problematic. We won't say catastrophic, but it depends what they're actually do. They can get things wrong. And then it's down to us to identify whether or not they've actually got things wrong. And that's where we need humans back in the loop. Now when we think about the future of artificial intelligence, so this is really looking potentially up to 30 years out, but it's looking at some of the developments that we have now.

Future AI architectures 22:22

22:22 We have neuro-symbolic artificial intelligence. So again, OpenAI are increasingly embedding EQ into their GPT-5 model. So this is where we've seen this a little bit with AI being sycophantic but EQ is really the use of training data to try to help artificial intelligence understand better understand human relationships human behaviors and so on and so forth but then reply appropriately according to the emotional state of the person who's talking to them or interacting with them.

22:59 But this is also the combination of intuition. So deep learning systems with symbolic systems. So essentially we end up with computer systems that are both able to see patterns in the data but then also actually logically think and reason about those patterns in the data. So neuro-symbolic artificial intelligence is essentially the attempt to create machines that truly do think for themselves as opposed to fantastic automated pattern recognition systems.

23:35 Now next comes swarm artificial intelligence. Now when we think about swarm artificial intelligence, this is both the use of a single AI to coordinate for example lots of different drones. However swarm AI is also the ability to coordinate AI to AI collaboration and communication because increasingly platforms like ChatGPT, they aren't just going to be talking to themselves. They're going to be talking to lots and lots of different artificial intelligence models all over the planet.

24:07 So swarm artificial intelligence is the research that's dedicated to trying to ensure that those AIs are able to functionally talk to one another without should we say twisting their minds. Next we have embodied artificial intelligence. Now Mark Zuckerberg famously said about a year ago that he doesn't think that an AI can be truly intelligent if it does not have a body and can't should we say participate in the world.

24:37 So embodied artificial intelligence is really the embodiment of artificial intelligence in physical things. Now that could ironically be something like a Roomba, but it could actually also be a humanoid robot. It could also be a self-driving vehicle and so on so forth. But embodied artificial intelligence when we have a look at things like those you know like systems from Toyota they're increasingly being trained on audio models visual models behavioral and relationship models and all sorts of other things.

25:16 And then the next hot thing so we've seen this with Nvidia we've seen this with Google basically this is the attempt to create what we call world models. Now artificial intelligence is very good in specific domains for example text or multimodal domains. But if you ask artificial intelligence to create a simulation of a city, if you ask it to create a simulation of a product that it's just developed, maybe it's a Formula 1 race car and we're talking about computational fluid dynamics, then really this research here is the attempt to develop AIs that are able to create digital simulations of the world.

25:58 Now that's either the entire world as we're actually seeing especially with Nvidia or that's the biomolecular world the cellular world if we're talking about healthcare or biotech space. So these world models are generally seen as the absolute next frontier for large foundational models and the amount of training that they need is huge. Yeah, we are typically going from exascale to yottascale data centres. Bearing in mind that AI with companies like Oracle, we already have a couple of zettascale data centres that are now serving artificial intelligence models, especially in the United States.

26:44 We're also trying to, as I said earlier, develop artificial intelligences that can learn from nothing. So zero-shot learning. Now, if you think about this from a human perspective or an animal perspective, when a baby deer is born, how does it learn to walk? It's not taught by its mother. It's kind of an intuitive instinct. So, zero-shot learning is the attempt by companies like Alibaba, for example, as well as companies like Meta, Google, again, etc., etc., of the usual suspects to create artificial intelligences that just simply learn for themselves and actually more than that learn in the potential absence of data.

27:25 So when we think about data it's said that by 2026 we're going to run out of data if we don't need data. So just think about the implication. If we don't actually need data to train the vast majority of new artificial intelligence models, then does the traditional status quo of the country with the most data wins still hold up? And then as we start having a look at some of the other artificial intelligence models, we've seen AI made out of glass.

27:57 We've seen 3D printed artificial intelligences called diffractive artificial intelligences, but we've also seen the development of chemical AIs, AIs and neural networks that are made out of DNA as well as liquid AIs. Now, these liquid AIs are AIs that are based in ionic computers. So within ionic computers, especially in some of the research labs in the United States, researchers have been able to create ionic neural networks.

28:30 So these ionic neural networks are quite literally liquid. So when we actually think about the future of artificial intelligence, we've also been able to turn human cells using CRISPR technology into dual core computers which we can then turn into very primitive DNA AI we'll say computers but DNA AIs or neural networks. So about three years ago we saw the beginning of the development of essentially what we could call the first human supercomputer.

29:06 Now we are a very long way off from being able to turn the human like myself into a fully fledged biological supercomputer but we have the first couple of Lego building blocks for that. Now when we think about DNA neural networks that have actually gone on to do some quite complex maths calculations as well as DNA computers from companies from universities like University of Manchester with Ross King. Consider this.

29:36 A DNA computer that takes up half a test tube potentially according to calculations and everything else will have more computing power than every computer on the planet combined. And that's before we start talking about quantum computers, quantum algorithms which you know again change the AI game, neuromorphic computers by which are self-learning and all kinds of other artificial intelligences. So when we think about the AIs that we have today, the multi-trillion parameter GPT-5 models compared to what's coming, compared to what's already starting to edge onto the lab bench, they are very rudimentary.

30:25 We have a huge runway ahead of us. And then when we have a look at artificial general intelligence, which again is pegged for about 2028 to 2035 depending how we define it. Then we think that an artificial general intelligence will have an IQ of 1600 and will have a billion times more cross-domain knowledge than any human. So the change here is we're moving from a thousand times more general knowledge with platforms like ChatGPT and GPT-5 to an AGI that has a billion times more knowledge than humans but that is cross domain.

31:12 That changes how you think about the future because from a world of work perspective, how do you start competing with an artificial intelligence with that kind of power and capability and knowledge? So with a lot of the work that I do in the future of work, jobs and skills space as well as education, it's not in my opinion just enough to be thinking about how do we change the workforce in the next couple of years. It's how do we keep the workforce evolving so that people can be gainfully employed not just by the end of the decade but by 2040, 2050, 2060, 2070.

31:52 And that's difficult and that's what we're going to be covering a little bit later in this presentation when we start coming on to some of the human aspects. Now when we think about artificial general intelligence, Sam Altman actually and the OpenAI board quite a little while ago now defined AGI as the point in time when AI outperforms all humans at all economically valuable work. Think about that.

32:26 If you are in a job that is economically valuable, whether it's wealth advice, whether it's software development, you know, whatever it happens to be, this is what OpenAI ultimately are trying to automate. And that's interesting. I won't say it's an issue, but it's interesting. Now because Microsoft invested so many billions of dollars in OpenAI, they needed a commercial definition of AGI. So in this particular case, the commercial definition of AGI is an artificial intelligence that can generate $100 billion of profit by itself.

33:09 Just think about that. Which takes us on to this. Now, a little while ago, Sam Altman was at a JP Morgan event, and I've been talking about this for years. Yeah. In 2019, I talked about AI building companies with companies like Accenture in Paris, and that's on the YouTube channel. So, I can put a little link in that for you if you like. Increasingly, Sam Altman sees a point in time where artificial intelligence can build companies, specifically multi-billion dollar companies with either one person or less.

33:44 Now, when we think about OpenAI and when we think about Google DeepMind, both organizations have roundly decided that the new Turing test should be whether or not AI can create, operate, and scale a successful business. Bearing in mind that we blew past the Turing test and no one has really talked about it. So what happens in a world where artificial intelligence is able to build its own company?

34:09 Now think of that as a digital company. Maybe it's going to build a software company for example. And again I've got some little snippets and nuggets for you in a bit. And again, when we think about Sam Altman's vision of the future, ultimately what he wants to try to achieve is he wants to be able to put AGI, all of that power onto a single computer chip.

34:31 Now, when we think about computer chips, when we have a look at some of the Moore's law, when we have a look at things like GPUs over the next sort of five-ish years, we think computer chips basically will gain over a million times in performance. And that's coming from people like Jensen Huang, who is the Nvidia CEO and increasingly when we think about AGI on a chip increasingly we're seeing sort of the Nvidia DGX platform which is now being sort of rolled out which can essentially help you train your own artificial intelligence models basically right on your desktop. So there's a lot of work in the semiconductor space to try to create semiconductors that allow us to do artificial intelligence at the edge.

35:19 And there's lots of reasons why we'd like to do that, especially from a healthcare perspective, travel and transportation perspective, environmental and monitoring perspective, and so on and so forth. But again, this is another game-changing advance. And we can probably start expecting something like this to emerge in the next 10 to 15 years because it's difficult. And this is sort of where today we're putting shallow neural networks basically onto chips.

35:35 An example of this is MIT's neural network which was able to drive a car but it only had 16 neural network nodes as opposed to the usual number of thousands of nodes that we would actually see in a traditional autonomous vehicle like a Waymo.

Industry applications 36:04

36:04 Now, when we actually start thinking about some of the game-changing advances that artificial intelligence has already racked up, there are some impressive ones. Now, we're going to start here. So Demis Hassabis recently said that he thinks in the next 10 years AI will cure or be able to cure all diseases. All diseases. Now when we think about Jeff Bezos and Mark Zuckerberg, both Jeff Bezos and Mark Zuckerberg with the Chan Zuckerberg Initiative and Altos Labs have been pumping literally billions and billions of dollars into research groups that are trying to on the one hand cure all disease by the year 2100.

37:17 Now on that one hand if we're going to go off on slight tangent here then if we cure all disease within 10 years now it doesn't necessarily mean that you have access to these technology these technologies and drugs and treatments but we can see a way that this happens and I'm going to show you that in a minute but when we think about immortality yes we can 3D print human organs and so on and so forth. And what happens when we all start living beyond the age of 100? That's a different topic for another day.

38:03 Now when we think about DeepMind so AlphaFold which is one of DeepMind's legendary artificial intelligences now took six months to simulate over 200 million proteins. Now, part to put this into perspective, in the whole of human history, we've only ever simulated around 2,000 proteins. And all of a sudden, within the space of really a three-year research project, DeepMind went from being rubbish at it to Demis getting the Nobel Prize for the achievements in this space.

38:43 Now, proteins are essentially the building blocks of all life. If you can understand every protein on the planet, you can create new drugs, new treatments. You can literally start thinking about curing disease. But it doesn't stop there. Artificial intelligence a little while ago found over 12,000 new antibiotic chemicals which when combined actually led to the discovery of eight new novel antibiotics. That's impressive. When we think about an artificial intelligence house out of Australia, they use their artificial intelligence to create 30,000 new drugs in 21 days.

39:39 And again, when we look at things like vaccines, we are now at the point where we can use artificial intelligence because we've done it to create new vaccines in seven minutes. So the way that we do this is we model a pathogen for example in a supercomputer or a quantum state and then these artificial intelligences simply try to assess which of these 25 million compounds in the databases they can combine together in which formats to kill that pathogen. So what we're really doing is we're doing computational medicine at scale and at speed.

40:15 But when you get to the point where you can create 30,000 new drugs in 21 days and bearing in mind that the US military along with the FDA are now trying to use AI not only to develop drugs but also approve drugs and then get them through into the general population so that we can all benefit from those drugs. Hopefully this starts being a very bonkers crazy world from a healthcare perspective. Now, Google a little while ago discovered 380,000 new materials.

40:49 Again, putting this into perspective in the whole of human history in the open-source US Department of Energy materials database as humans we had only ever found 40,000 materials. So within the space of two weeks we went from 40,000 materials to 420,000 materials. Now again think about the relevance of that. You know if you want to create next generation semiconductors or electronics it's all about materials. If you want to create hypersonic weapons systems it's about materials etc.

41:23 Materials are literally what makes the world work. So now that we have over 420,000 materials in an open-source database, you can change everything. Communications networks, computing, all sorts of stuff. We've also used artificial intelligence and combined with rapid gene sequencing to sequence the genome of 16 humans in 5 hours. So when we think about the development of personalized medicine, I can now sequence your genome for $100 in five hours.

42:02 And if you have cancer, I can take that and then I can use CRISPR technology combined with artificial intelligence to create what we call a human hybrid T-cell immune system, and that modified T-cell in many cases has been able to kill 100% of stage four cancers. So when we think about the use of artificial intelligence with rapid gene sequencing with personalized medicine with very powerful genetic engineering technologies to create new hybrid immune systems for ourselves which identify disease within our body and then specifically target it like a sniper and then kill it. We've done it.

42:53 We're there and then we've got to ramp up and scale. Now artificial intelligence is also becoming an inventor. The only place where AI is able to own its own inventions is in South Africa with the South African patent office. Anywhere else around the world like the USPTO, they don't rate AI as the inventor because the inventor has to be a human. However, with Toyota, artificial intelligence in two weeks developed a completely new kind of electric vehicle lithium ion battery.

43:23 Now, normally it takes two years to develop EV batteries including testing. In Germany a company called Hyperganic developed a brand new rocket using their artificial intelligence or rocket engine but they did it in six hours and this rocket engine was a single piece 3D printable rocket engine. And then LEAP 71 in the UAE in Dubai also did the same. So while AI is not necessarily creative or particularly novel, it is getting to the point where it's managed to create and dare we say invent some very radical new things and those things include algorithms.

44:07 So with Google's AlphaEvolve that I mentioned earlier, AlphaEvolve has been evolving and developing and literally inventing new algorithms. But these algorithms are very good, but also human experts don't really understand how they work. And here's this. Now the IE consortium also used a convolutional neural network. And this CNN was tasked with trying to develop a new fast computer chip and it did a wonderful job. It created one of the fastest computer chips on the planet.

44:43 But when the experts actually looked at this computer chip, again they didn't understand how the AI had built it, designed it, and they didn't understand how it operated. And the reason for that is because as humans, we have designed for example computer chips in this way. We've always done it this way and we iterate and iterate and iterate and we build on knowledge that we've already got from previous generations.

45:11 But with artificial intelligence, AI is saying, well, if I want to create this output, what are the million different ways that I can get it? And in many cases, what we're seeing is artificial intelligence is saying, well, the way that humans designed computer chips is actually quite inefficient. I found a better way to do it. So AI is not necessarily creative. It's not necessarily an inventor as we would think of a traditional inventor.

45:37 We can't quite give it the inventor's moniker at the moment for a variety of reasons. But it's technically discovering and/or inventing or certainly creating some amazing new products that surpass anything that any humans have actually been able to put together so far. Now, one of the reasons I said that Microsoft sees an existential threat with artificial intelligence is AI is coding. Now, if you think about Microsoft, Microsoft is a software house.

46:11 They provide software and they provide services to go along with that. Consulting advisory services, cloud services and orchestration services and everything else. Now, as I mentioned earlier, Anthropic's artificial intelligence Claude managed to code continuously for 30 hours and it cloned Microsoft Teams and Slack. So, Eric Schmidt a little while ago said that in the future he sees a point in time where artificial intelligence will be able to clone TikTok, its algorithms, interfaces, and everything else in 30 seconds.

46:44 Now using a natural language interface, I as a non-programmer eventually will be able to say I would like you to create this kind of program, this kind of software. The AI will understand the context. It will understand what I'm asking for. It will understand what that software is designed to do and the features that it needs and then it will just code it. And then using my natural language voice, I have a natural language voice.

47:09 Use my natural language voice ultimately I can then just say tweak this tweak that how does that work etc so what happens to human potential when everybody on the planet is able to develop software what happens to human potential when the people who think that they might be able to design a better rocket engine but haven't had the skills the resources the money the labs the infrastructure etc are just able to talk to an artificial intelligence that pretty much costs them nothing to use and just say create a new rocket engine, create a new training shoe, you know, whatever it happens to be. We're going to come on to the implications of all of these things in a moment. Fusion.

48:03 When we think about platforms like Google DeepMind, those AIs are so good that they are actually containing the plasma rings within fusion reactors. Because one of the problems that we have with fusion reactors is as a 300 million degree Celsius plasma ring goes around the tokamak, it starts hitting the sides. You've got to be able to keep magnetic containment and you've got to be able to keep perfect magnetic containment.

48:33 So AI is helping accelerate the fusion revolution which is still decades away but AI is playing a significant part on the development of fusion and AI has even theorized that cold fusion. Yeah. So this is fusion at room temperature is possible and there's a company that's actually developing a cold fusion reactor. Now when we think about food artificial intelligence is able to understand the genetics of a particular plant and then it's able to take weather data and soil data.

49:04 It's able to analyze all of that and it's able to within the space of two months predict crop failures. So now think about the impact of being able to predict crop failures two months in advance on the human food supply. Bearing in mind that 40% of our food is actually lost to bad weather. And then about another 20% to pests and pestilence and so on and so forth.

49:24 So AI is actually proving very useful in the agricultural space. When we think about vertical farms, companies have been using robotics and artificial intelligence combined with machine vision to produce fully autonomous vertical farms. Now, these vertical farms use 99% less potable water. And if they extract water from the air, then they can use 100% less potable water because they just recirculate it inside. They use 100% less pesticides and chemicals and herbicides.

50:03 The yields are guaranteed and can be up to 350 times greater than we would see in a traditional farm. And then of course because there's no herbicides or pesticides, the vegetables that are grown in a vertical farm are organic and they can be grown next door in your local city. So AI when we think about the use of artificial intelligence in precision farming but also in some of these new farming techniques is significant.

50:32 And if you want a little snippet, if you wanted to flip today's agricultural system to a cellular meat production system, which is growing meat in a lab as well as vertical farms, we estimate it would cost about $350 trillion, which is almost but not quite four times more than the global energy transition, which we estimate would cost about $92 trillion. Where in that space AI is actually creating new solar panels and actually helping improve the efficiency of grids by at least 20 to 40%. But that's a different presentation.

51:02 You can check that one out somewhere else. We also see artificial intelligence being used to create and generate new food. So we've got new strawberries that artificial intelligence has designed and we've also got new cows that are resistant to tuberculosis because artificial intelligence increasingly understands genomics. We've also seen artificial intelligence creating at least two fully synthetic artificial human genomes. So when we think about the introduction of artificial intelligence and its potential to create synthetic life, we have an E. coli bacteria called Syn57 because 57% of the E. coli bacteria is fully synthetic.

51:53 And increasingly we're getting to the point where it looks highly likely that we will be able to create the first fully synthetic bacteria. And then we move from bacteria. And then if we look at MIT, Harvard and UCL by 2037 they want to be able to use technology to create the world's first artificial synthetic human. You can read about that on the website.

Personal development and work 52:23

52:23 I'll put links below. Now, all of this sounds magical, science fiction-like, etc., etc., but in a world that is full of science fiction-like technologies that are getting exponentially better, exponentially faster. How do we as humble humans manage in this world? Are we all going to be automated by machines? Or is something else going to happen? Is it going to be something like the singularity? Let's dig in.

53:01 Now, when you think about the future of work, jobs, and skills, ultimately, you're either working for yourself like I do as an entrepreneur. Congratulations. It's a hard job. I know. You can cry on my shoulder later. You're either working for yourself or you're working for someone else. Now, if you're working for someone else, companies ultimately want to be able to scale as efficiently as possible. So, the reason you have a job is because there is no other way.

53:36 No, there's no better, cheaper way to automate. Now, that starts changing. The economics of automation start changing as we see agents and AIs developing and so on so forth and getting easier to integrate into enterprise infrastructure. So as an individual if you want to remain employed for the next 10 20 30 40 years and I've been talking to Google the Google about this very topic in both Dublin and New York recently then ultimately you want to stay relevant to that company or to the jobs market.

54:01 Now the way that you do this is this. Firstly, you need to be able to align yourself with the mission, values, and direction that the company that you work for is going in. You need to be able to anticipate what the future company that you work for will need. Now, one way to anticipate what your company is going to look like, what it's going to produce and sell, and how it's going to sell it, and who it's going to sell it to in the future is a very easy thing.

54:36 Follow the money. If you see teams within your organization that are being made redundant or cast aside, and you see the organizations hard-earned cash flowing into particular business units, then guess basically where the growth area, guess where the leaders think the growth areas are. And then you need to be able to adapt. So align with the future trajectory of your company. Anticipate what that is. So follow the money and then adapt at speed.

55:07 And this is how you adapt at speed. We'll go into that. Now when we actually look at the top artificial intelligence use cases for 2025, Harvard pulled this together. The first is companionship. The second is life organization. The third is finding purpose. Now on the one hand, this is kind of sad. Over 50% of people on the planet basically report being lonely quite a lot of the time.

55:27 So, but also a lot of people are starting to use artificial intelligence for help with for example financial advice. So, 40% of people are using AI for financial advice and if you're a high net worth individual that rises to 80%. But we're also starting to ask AI to help us with life organization and finding purpose because it's not judging. And so I can talk to my artificial intelligence about anything that I like.

56:00 And I know that the only thing that it's going to actually share is everything with Sam Altman if I've got my settings set wrong. But that's a conversation on privacy and security, which we can have another chat on that another time. Now when we think about how we use these increasingly powerful artificial intelligences whether it's in the education space or whether it's in the work space Microsoft got 400 of their software developers and they told them for a period of three months exclusively use artificial intelligence to do software development don't use anything else and what they found is at the end of the and you can read about it at the end of that three-month period they'd lost 40% of their critical thinking skills.

56:40 That's a lot. So, the lesson here is the more you rely on artificial intelligence to do your job and your thinking for you, the worse your critical thinking skills are. And as everyone will tell you, critical thinking is needed in the future. It's a core soft skill today, 10 years, 20, 30, 40, 50 years from now. Now in addition to that metacognition is people who think about how they're thinking and what we actually saw with for example Anthropic when they did similar experiments was in 50% of cases when people were using artificial intelligence to do their work for them they just stopped thinking for themselves.

57:33 So that was an issue and then when we actually think about learning loss basically what we actually see is people who are relying on artificial intelligence and this is from research papers people's ability to learn new things reduced by 17%. So the lesson here with this is if you're using artificial intelligence as your crutch, not as an aid. If you are using AI to do all of your thinking and work for you, then as we saw with MIT's MRI research, you are literally getting stupider and you're losing some of the key skills that you need to remain relevant in the future.

58:19 So you've got to remember as powerful as it is and as powerful as it's going to get artificial intelligence is a tool at the moment like any other self-developing self- evolving bit of a funky tool but it is still a tool. So when people have used artificial intelligence to help them in the right way. So I do my work in the way that I would normally do it.

58:45 I do my homework in the way that I would normally do it and then I use artificial intelligence after the fact to assist me helping make helping make that work better. A little bit like you know a mentor or a coach relationship. Then what we actually saw is when individuals who used AI not to do their work but to help support them do their work in the right way.

59:09 So you've got to think about the level of effort that you put into your work. That's how we measure this proof of effort. Then the amount of stress associated with doing that job and that task actually reduced by 30%. So your cognitive load reduces by 30%. You become a little bit freer which is nice. But we've also then seen a 60% increase in productivity. We saw this with research out of MIT.

59:43 So if you rely on artificial intelligence to do all of your work for you, you lose critical skills and you become dumber and stupider and you become less relevant and less employable. If you know how to use artificial intelligence in the right way, actually you become more relevant and you become more employable. And this is what I'm now going to dig into. So think about this. AI has an IQ of 136 to 155, a thousand times more general knowledge than any human.

1:00:12 Increasingly, for zero or $20 a month, in the case of ChatGPT, we are democratizing global access to intelligence. We are also democratizing access not just to information like Google did with traditional Google search but we are starting to democratize access to knowledge and even to expertise in some cases. But here's the real rub. As artificial intelligence automates different jobs, tasks and skills, as an individual who has maybe never ever been trained in those skills or learned those skills or those jobs or those tasks or those particular domains, I now have access to those.

1:00:58 So, I'm not a programmer, but increasingly I'm using artificial intelligence to develop software for the 311 Institute. I'm not a lawyer, but I can use artificial intelligence not only to put a legal contract together, but to actually review legal contracts to come into me and then recommend adjustments. So, what happens to you as an individual? What happens to your potential when you have access to free intelligence, free knowledge, and all of the world's skills for free?

1:01:40 Ultimately, we see the end of T-shaped individuals. I now make software. I write legal contracts. I use artificial intelligence to perhaps do my marketing, to do my research, my analysis, all this kind of stuff. What happens when you are no longer T-shaped individual and we haven't really got a name for the replacement. I'll call you a W-shaped individual because you have multiple skill domains. Now, what happens in the world of work when people who used to be employed within specific lines of business because they were T-shaped individuals now have access to all skills that have been automated.

1:02:31 Now on the one hand this changes how company structures work. We kind of go from pyramid structures to barbell structures to obelisk structures if we're McKinsey to diamond structures if we are Tata, HCL, Infosys and so on and so forth. And there's lots of sort of conversations about the future structure of companies. Ultimately do we end up in a swarm company structure where I now have access to lots and lots of different skills.

1:03:03 The company need that I work for needs something done and they have a way to tap into me to access that skill for that particular task and in the meantime they're also working with others as well. So from a future companies perspective, which again is something I've been working with companies like Walmart and Google on, do we end up seeing companies that are built around, should we say, collaborative swarms of people and agents because obviously AI agents will not only be managing people, but they'll also be managed by people and in many cases they might actually be orchestrating the work that needs to be done and it's the AI agent that will say this is the work that needs to be done. I'm going to reach out to Matt because he's got that skill.

1:03:52 I'm going to reach out to Damon because he's got that skill. And I'm going to reach out to Sarah because she's got that skill. And together, we're going to coordinate with one another to get that particular piece of work done. Now, ultimately, as artificial intelligence automates hard skills, you need very solid soft skills, but your hard skills need to be switchable. So, today for example, I'm a data scientist.

1:04:17 Tomorrow I might be a quantum data scientist. Today I might be a software developer. But tomorrow because AI is writing software that we might not actually understand what it's done, how it's done it, how reliable or secure it is, I might end up being a software governance or software forensic inspector. So increasingly we need to be able to swap in new hard skills and swap out hard skills that are no longer relevant at speed.

1:04:47 Now increasingly you need to become AI literate. I see that is an absolute must. There are so many people I meet around the world they say there will be two kinds of companies, two kinds of people in the future. Those that use AI and those that don't. And the people who use AI basically will be relevant. The people who don't use AI, they'll be kicked to the curb.

1:05:08 We also need to be able to adapt and learn at speed. And this is really what I want you to dig into. Now we're in the final leg of this presentation. I know it's been a long one, but hopefully you've enjoyed it. Now, on the one hand, being able to adapt at speed is understanding and anticipating change, understanding the impact that those changes are going to have on the things that you care about and then you being able to proactively put a plan together so that you can navigate those changes for your own benefit.

1:05:39 Now in terms of learning at speed, I along with the students that I teach and I teach eight-year-olds up to 25-year-olds as well as senior executives up to 80 years old really. By using artificial intelligence in the right way, I can accelerate how fast you learn new topics and new things by at least 12-fold. So I can use artificial intelligence increasingly as what we call an adaptive tutor and it can teach me for example the fundamentals of cyber security and it can teach me to a 101 level as I dig into the rabbit hole it can then teach me to a 201 level.

1:06:19 So now let's say I've acquired new knowledge that I didn't have before, new expertise and skills that I didn't have before maybe in the cyber security space and I take those level 201 skills go and work for a company that then provides me with the 301 skills the experience the firsthand experience of in this case being a cyber security analyst maybe because it's all very well let's say actually learning something new, but you only really fully learn it, people say, when you apply it. So, AIs can get you from a 101 to a 201 level at the moment, but you need to then work for a company that takes you to a 301 level. So, this is how you remain relevant.

1:07:04 So, you know that for example, cyber is going to be an increasingly competitive field in the future. Your company's investing lots of money in it or they're not, but you know, they're investing lots of money in it. Use artificial intelligence to train yourself up. Now the way that you do this is very simple. Get something like ChatGPT or Copilot talk to it. Say I think that in the future my company is going to be more AI-centric.

1:07:33 For example, if it's Google, I think that AI is going to be more heavily embedded into their retail products and as a consequence I think basically that this is a future job role within Google with these future competences. So, I need to learn these new skills. Teach me these new skills. And you will find, especially with ChatGPT, because it's very good as well as Gemini, it will start teaching you those new skills.

1:07:51 But it's not necessarily about the AI that you use. It's about how you use these artificial intelligences to help you adapt to the world that we now find ourselves in as well as the future world. So as you realize that the future direction of travel for your individual job is that way. So I'm going from a data scientist to a quantum data scientist. Again, you can ask artificial intelligence to assess your quantum data science skills and it will ask you a set of questions and it will at the end of it like an exam it will say right when it comes to quantum data science your level of expertise is down here.

1:08:42 Now what you can do is you can then start asking the artificial intelligence to put courses together for you. Okay, well, if this is my skills gap, teach me about those things. Help me help me master those particular skills. And then once you understand what those gaps are, you keep going. And then ultimately once you have learned enough, bearing in mind that we talk about continuous learning, try to use these AIs that have an IQ of 155 and in the future a billion times more general knowledge or domain specific knowledge than any human.

1:09:28 Use these AIs to master whatever the thing that you care about is. So if I want to master being a quantum data scientist, these AIs have read all of the information, all of the research papers on a planet. When you use these technologies in the right way, they don't make you stupider, thicker, or help you lose your skills. They can turbocharge your skills and most importantly they can amplify your skills.

1:09:57 Now if you think what would happen if you had access to a thousand Albert Einstein because that's where we're kind of going. That's kind of where we are at the moment. How would you tap into that? You have access increasingly to free intelligence, free knowledge, and free skills. Use these AIs to master whatever it is that you care about. A very simple example of this is while you're in the car going from A to B, try to identify gaps in your skill set and then try to master them.

1:10:52 You will very quickly see these artificial intelligences in a very different light and you will see your future in a very different light. And with that, I hope you enjoyed that. My name is Matthew Griffin. If you want to download my books and codexes, if you want to check out other YouTube videos, basically then you can do that for free. Just scan the QR code. And if you have any questions, put them into the comments below. Ask me directly. You can find me on LinkedIn. All of the usual suspects. And have a fantastic future. Great. And goodbye.