Introduction

Thank you, Adrian. Thank you, Henry. Thank you, Rob. Hard acts to follow.

Thank you all very much for coming this evening. Thank you, Mindstone, for hosting. Adrian, thank you, Rob and Origin.

I'm actually an Origin member in the lounge. I can highly recommend Origin if anyone's looking for a workspace. space.

Who’s Speaking and Why This Topic

My name is Mark Riley. I am the founder and CEO of a small boutique AI consultancy in Bristol, in Clifton here, called Matheson.

And there is a small prize of my own book called The AI Profit by Mark Riley, available on Amazon and all good booksellers. Signed copy for anyone that knows or can figure out where the name Matheson comes from by the

the end of the talk and you're not allowed to take part and um so my background i spent uh probably been in the ai trenches about 10 years now so i was initially involved with ai in new york when i was working for dow jones and the boss willis asked me to figure out what is ai and how can we use it in a large media organization so i spent five six years wandering around the u .s went to see a lot of AI startups in Silicon Valley, introduced AI into Dow Jones, which is the Wall Street Journal, came back during COVID, set up Matheson, went back to school, studied AI at Oxford Said Business School, which I highly recommend, and London Business School.

And I now mainly service media companies, but we're also looking at kind of mid -tier risk and insurance businesses.

Why AI Job-Loss Fear Is Peaking Right Now

There's a lot going on in terms of AI fear at the moment, and I could have picked many Many topics from existential risk through to biochemical attacks, through to China, through to job displacement. And I picked this one because this one seems to be the topic du jour at the moment. This one seems to have percolated up quite recently as the topic that is engaging a lot of people.

A few caveats. I'm conflicted giving this talk. I'm not a natural optimist. But I'm going to try and be optimistic in the next 10 minutes and see if I can bring you with me. That's not to say I don't have my own concerns.

I sometimes wake up as the ultimate Luddite and I want to smash the looms. And I think you're allowed to have a smash the looms moment or nuke the data centres or however you want to do it.

But I think I'm not trying to belittle the danger that is out there and every job lost to AI is a travesty in its own right. So I'm not belittling that, but I just want to try and be as positive and optimistic and see if I can bring you with me.

And what we'll just do is go through the current news cycle, which is I think we're in peak doom at the moment. And then we'll look at why this fear -mongering is coming and bubbling up.

I'm going to lean into some Scott Galloway thoughts. Is there any Scott Galloway fans here? Prof G, I recommend him.

And then we'll look at Jevons Paradox and then how we can apply that to the future and really aim the talk at this room and see if I can help you figure out how to navigate the next five or six years of certain disruption and where we land up.

So this is really the message of the talk. If they take one thing away, it's probably it.

Pessimists sound smart. Optimists get rich.

And then you don't have to go far.

The Current Doom Loop: Who’s “In the Firing Line”

Just go spend the weekend on YouTube and you'll find yourself going down a YouTube doom hole of despair spare about how everyone's going to lose their job but the people that are most in the firing line at the moment seem to be developers creatives writers journalists that's my my area designers

anyone that's involved in analysts looking at data generating reports people in support roles and especially and i think this is pertinent to anyone that's just finishing university or just finished university is entry -level roles across legal accountancy and consultancy and there's data to support those jobs aren't coming through.

As Henry just mentioned, this is how not to handle an AI rationing role. If you're a CEO, this is probably not the best comms to say that you're going to get rid of lower value human capital, which is in this morning's Times. Bill Winters has since walked back those comments.

But there's a lot of companies, Duolingo included, who announced their cuts and then had to walk back their announcements because their customers were like, hang on, I'm not going to use Duolingo if you're going to fire half your support team for machines.

Meta, famously, are now cutting 15 ,000 workers either because of AI or reassigning them. So that's about 10 % of their workforce. Apparently, it's miserable working for Meta at the moment. They're all just being tracked on their usage of technology and people that aren't measuring up are getting forced out.

And this was yesterday's Guardian, a third of university students in Great Britain and think AI job losses will cause social unrest. So this is the environment that I'm trying to make you optimistic about.

So there's a website, if you're very interested, called jobloss .ai. You can track AI, which they think jobs that have been replaced by AI. And the current tally is about 170 ,000 people they reckon have lost their jobs as a result of AI. So that's the doomsters. That's the agenda for the doomsters.

When the Model Builders Amplify the Panic

And then it's not being helped by the people that created these models telling us how awful they are. And it really winds me up when you see these interviews with these mega rich oligarchs who are being interviewed by the media telling us we should be so afraid of these monsters that they've created.

And I particularly get annoyed by Mustafa Suleiman, who is now the head of AI at Microsoft, who's written his book, The Coming Wave. Has anyone read The Coming Wave? So his argument, he spends 14 chapters telling us how awful it is and then two chapters on how to fix it.

And by the way, he's a billionaire on the back of it, but he's not. Thank you for telling us this. Anyway, I'll get that off my chest.

But his argument is that in 18 months, for all white -collar work to be automated by AI, Daria Amodai, who's trying to be the good guy in all of this, said 50 % of entry -level tech legal consulting will be wiped out in five years. Unemployment can spike to 20%.

Deedley Das, who's a venture capitalist at Menlo Ventures, tweeted at the weekend about 10 ,000 co -fronted labs sitting on generational wealth.

Scott Galloway’s Take: Fear as IPO Marketing

Now, if you lean into the Scott Galloway argument, his thesis is that they're talking up the dangers in order to increase their capex prior to an IPA. And that is rational.

They are these gods of the universe. They've created these machines. I would call it the Oppenheimer complex. I think Sam Altman has an Oppenheimer complex, but what they're doing is saying these machines are so dangerous They're so disruptive that my company's worth the trillion dollars, and I'm about to float on a stock market IPO this year Thank you very much. That's Scott Galloway's argument

You might be that they're just saying when this does happen at least we warned you and we gave you some For for warning that this eventuality would happen, but what I'm trying to say to you is take these Terrible predictions with a pinch of salt And I think they're being talked up in order to market fear to drive their valuations.

This is quite entertaining. I haven't got time to play it now, but you can look it up on Google, on YouTube.

Eric Schmidt was giving a commencement address at a university last week, and he got booed after telling these graduates that the future was AI, and they all sought to get on board.

Same happened the same week with Gloria Caulfield, a real estate executive, executive, called it the next industrial revolution, and all the graduates there booed her out of the stadium. That's where we are now.

A More Grounded View: Transition, Not Instant Apocalypse

Now, I'm not going to sugarcoat it. There are going to be job losses.

There's going to be a transition phase, and I think it will be pretty brutal. It could be two years, it could be five, it could be ten.

There will be a large number of people that have to retrain, reskill, transition.

But we've got to take the world as we find it, and Joseph Schempeter is back in fashion 1930s economist who came up with the term creative destruction the idea that any new technology is going to cause a lot of disruption along the way but we have to look beyond that at the upside but in the data if you really do follow the data there

is weakness in graduate employment for sure especially in job postings but we're not seeing radical increase in unemployment yet, still about five, just over 5 % in the UK.

And a lot of these layoffs are actually covering up for excess hiring during the COVID pandemic, suggesting that they overhired in 21, 22, and they're now cutting back.

They're using AI as a smokescreen, or it's called AI washing, to try and cover up their bad management skills.

And I lean into this graph whenever i'm feeling really depressed so this is gartner suggesting that yes there will be a lot of jobs lost in the in the millions you know four and a half million jobs will be lost but if we

hold on long enough 20 30 plus we're going to see these jobs up to nine million come back so it's all about the shape and size of the v there is a v coming and it's all depending how sharp it's going to be how long it's going to last how fast the new jobs will come in and how badly that's going to affect individuals along the way.

The Optimistic Framework: Jevons Paradox

So Jevons paradox, very simply put, suggests that when a resource or a material becomes significantly cheaper, the demand actually goes up because people find new use cases for it. It's the simplest way.

It came out of the Industrial Revolution when coal was incredibly cheap, and everyone thought, well, that's going to kill the coal industry, but actually they found new ways of deploying it across steam engines, and the demand for coal went up.

so if you apply that to today intelligence is effectively going to zero what does that mean for us it actually means we're going to find new use cases this is a positive argument we're going to find new use cases for intelligence we're going to find things we can't even imagine of at the

moment and the interesting and reassuring case study is radiologists they found that screening screening made it incredibly cheap now to do mammograms and everyone said radiologists are out of work what actually happened is in u .s mammograms became far more accessible far more people could use them so the demand for radiologists actually went up so new jobs are emerging at the

What New Work Could Emerge

moment we're seeing implementation consultants coming through something we do at matheson there's lots of roles online for forward deployed ai engineers product managers managers integration specialists ai government leads and future roles there's this very positive

argument there's going to be so much data created by ai especially from things like wearables there's going to be a whole new layer of employment around people interpreting that data and turning it into human speak so for medical purposes you might find there's a whole new category of employment of people who will then interpret your health data and then make recommendations

recommendations and prescriptions. They're not doctors, they're data interpreters.

In Henry's world, you're going to have so much data about your financial life that you'll then need not someone as clever as Henry, but someone intermediary who's going to interpret that data and give you financial advice on the back of that.

Resources for Finding ‘Good’ AI Work

If you want to do something worthy, there's a site called 80 ,000 Hours, and they've listed all the jobs that will help protect against the downside of AI.

There's lots of adverts for courses for 175K AI leadership roles. And there's a website called artificialintelligencejobs .co .uk,

which is the best resource for finding specialist AI roles in the UK.

Sorry, I'm going really fast at the moment.

Where the Money and Demand Will Be

Corporates. It's going to take corporates five years to catch up.

Deloitte, McKinsey, PwC, the big four, they're all making money now on AI transformation. It's probably about the only thing they are making money on at the moment.

So if you want to get involved, there's going to be an enormous amount of demand for AI specialists to go into corporates. Or if you're feeling very brave, you could build an application to sit on top of the model. Some real -life use cases that are generating revenue at the moment.

Practical Use Cases and the App Layer Opportunity

For my clients and media clients, they can build co -pilots on top of their archives. There's lots of use cases within contract review, insurance document review, healthcare.

care and then matheson we've built six seven prototypes which have been now been deployed to the enterprise we farm them out between 20 and 90k a pot so there is demand for for experiment

Startup Ideas: Transcription and Healthcare Workflows

and r &d in the space i won't dwell on this but here's just some uh ai startup ideas you might consider i think i'm very hot on transcription at the moment i think a transcription has come a long way in the last six months.

We're seeing a lot of use cases in health. So your doctor might well ask permission now if he can use or she can use a transcription tool during the consultation to save note writing.

Same goes for the police when they're interviewing suspects.

Same goes for care workers when they onboard an elderly client into a care home.

New Career Paths: Solo Work, Governance, and Avoiding Scams

Now, if you want to go it alone, good luck. I think a lot of people are looking at this now.

If they do find that they're sadly no longer wanted in their corporate or their current job a lot of people are looking at going solo they're not necessarily willingly going solo but we're seeing a lot more solo entrepreneurs bubble up and again it's the

same roles that i've been talking about you can go into a company and advise them how to integrate claude uh which is all about context connections and skills or you might go in as a governance and

ethics advisor i'll give them a time one minute okay um avoid these guys no one's here because they've been on tiktok but these guys are trying to plug you you know you're going to make 100k this month if you follow this course and guess what you've got to buy their course for 40 dollars 40 pounds a month um scams they never work they're all in dubai they're all taking selfies

From Execution to Judgment—and Why Humans Stay Valuable

it's a load of bollocks um so where i see this going is we're going to shift away from necessarily execution tasks as human to judgment tasks there's a very good essay here written by uh alex imas

goes to electricity which is an argument about scarcity and we'll pay for scarcity and in the age of ai scarcity is going to be relationships we're going to pay for humans starbucks tried

to automate a lot of their out their coffee shops recently three years of declining revenue and they realized people actually pay four pounds of coffee guess what because they want a human barista that they don't want to just pay for a machine.

So scarcity will be the human significance of an interaction. That's where they're heading.

And then there's all these other articles coming, bubbling up now from the New York Times, why AI job apocalypse probably won't happen, and a nice piece on Puck about AI job apocalypse revisited.

So the mood is starting to turn, and people are starting to be a bit more optimistic.

I haven't got time for this.

A Reality Check: Token Costs and the ‘Uber Discount’ Phase

Does anyone know what this means? Does anyone explain this very quickly? Yeah? Correct. Rob, do you want to?

So last week, Anthropic put a throttle on any API calls on Claude. So if you're building Claude in their harness on their software, that's fine. They'll charge you £200 a month. You can use it all you want.

People realised the amount they were using it, they're actually paying a £4 ,000 discount on the actual usage of the token. So what Claude said, OK, if you're going to build an external plug -in for an API call on Claude, we're going to charge you market rates. And all these startups have been plugged into Claude on £200 a month, suddenly paying £4 ,000 or £5 ,000 a month for one user.

So what's happened is they've been discounted in the same way that Uber was discounted for years by the venture capital community to get you all using Uber. And you take VCs away, and they're now going to market rates. Now what's actually happened and if you if you apply the math maths is that token usage is making Machines more expensive than the equivalent human

Closing Message and Community Links

Which is interesting going forward, but depends how long the VCS will back it up anyway Augment replace it's up to you guys. Really? You've got to navigate this stay human and enjoy the transition. That's me

I've got a newsletter. I've got a newsletter. We've used a I to build a newsletter for Bristol I've got a book on Amazon and we started a supper club called JAIL the jovial AI

league invite only we meet once a month in the pub if anyone's interested to ask me about that sorry I'm over thank you very much give it up for mark okay it

Audience Q&A

Accountability: Who’s Responsible When AI Goes Wrong?

feels like you know all of our sort of human systems run on kind of accountability and it feels like accountability is kind of the backstop for a lot of ai adoption and i was just wondering do you see that changing can that change how would

that change well that gets into the ethics and the philosophy of accountability which i think is still to be figured out so if a self -driving tesla takes itself for a joyride and kills someone at night we haven't yet figured out who's accountable for that is that what you're getting

it yeah so yeah in most processes where we're introducing ai the ai can't itself be accountable correct and then is it the software developer is it the person that owns the software is it the company behind it uh or if you're being really anthropomorphic can you put an algorithm in court punish it tell it not to do it again so it teaches all the other algorithms not to miss

but how do you punish an algorithm? Thank you. A quick question from me is,

‘Picks and Shovels’: Data Centers, GPUs, and the Infrastructure Bet

in respect of your points around commentary on the markets and, I guess, AI -adjacent industries like consultancy, you mentioned McKinsey, EY Deloitte, et cetera. We're seeing a lot of movement in the data center space at the moment.

Obviously, it's a booming industry around the country. The point about tokens is interesting. I guess my question here is,

in a gold rush don't you want to be the person selling the shovels that's what we're seeing at the moment I guess in data centre space and just your views on that would be interesting

yeah so someone in this room I won't point to him he bought AMD shares

a year ago they've doubled in the last year so if you're heavily invested in AMD

and Nvidia and all the picks and shovels you've done very nicely um demand is still not catching up sorry supply is still behind demand

so i don't see that ending soon but absolutely uh the ai application layer which is all the startups trying to build on top of open ai and claude are highly vulnerable they could get wiped out but the infrastructure layer so the stack below open ai and claude which is all the the gpus and the data centers are all fairly good bets i reckon for the medium term with one caveat

which is taiwan and if taiwan gets blockaded then the picks and shovels won't turn up anymore which is an interesting geopolitical discussion thanks mark as always um quick question on the

Decision Fatigue and the Need for Human Judgment

judgment and execution um i obviously i personally experience a lot of decision fatigue so making a a lot of decisions about the outputs i'm receiving from different lms do you think that maybe inspiring some doom and gloom conversation but do you think that humans are geared up to adapt to that judgment as opposed to execution that's hopefully what i was trying to convey is that

we need that human layer to make the judgment calls and if i if i have one other very strong belief system is that we should never ever assign consciousness to these machines because as soon as we do that they've suddenly got judgment and we need as humans to have superior judgment

and it comes back to everyday use of validation and trust but verified people use llms and take their answers as gospel they still make it up so human judgment is so important in this space and and we have to reserve the right to turn them off,

which some people find controversial, but I think it's quite sensible.

Wrap-Up and the ‘Matheson’ Reveal

Does anyone know where Matheson comes from? No? Anyone? Rob?

Oh, someone want a book? It's Alan Turing's middle name, Alan Matheson Turing. There you go.

Conclusion

Brilliant. Thank you very much, Mark. Everyone give it up for Mark, and I'm sure there's plenty more conversations to come.