63. What does the MIT report actually tell us about the AI industry?

0:00Welcome to The Difference Engine, the show for tech founders, investors and innovators.

0:09what have we got coming up today well we'll be backing our brothers and sisters in europe what's with all the negativity from across the pond and we'll be talking about our man of the moment the amazing herman hauser but first we are reviewing the reports and asking what's the social influence of ai

0:34I'm reviewing the situation. Can a fella be a villain all his life? All the trials and tribulation. Better settle down and get myself a wife. Well, probably one of the widest reported pieces of research, certainly this year maybe ever is this mit report about ai and saying that 85 or some say 95 of ai pilots result in failure just everywhere um there's also interestingly has come out a time when another set of researchers has talked about um forgetfulness and uh the fact that ai might be bad for all memory is this as is reported in many papers shamefully the middle class uh graduate jobpocalypse.

1:25What can we learn societally and what is AI's influence on society, sir? You know, let's face it, AI is genuinely scaring many, many people in developed industrialized societies. And I think more than industrialized societies, I think we're talking post-industrial societies here. You know, it has surfaced with concerns on a macroeconomic level and caused interesting debates which have long thought to be simply the domain of politics like the idea of a universal basic income you know that's actually an idea from ancient egypt and the philosopher thomas moore a man for all seasons in the uk in about the 16th century we've been on this for like half a millennium yeah but you know is this all very so is it sensible to be so worried about it And what are the latest thoughts on how we cope with this tech, which will be game-changing?

2:27This will be as game-changing as the application of machines to production. We're looking at the application of a machine to the production of intellectual property. Right. And just to go back to our sources and deconstruct a little bit, this MIT report, for me, is perhaps the most hyped piece of software law since Marc Andreessen penned the Software is Eating the World blog. And I had to look this up. That's actually a quarter of a century ago. He did that in 2011. Now, that was his opinion. That has been cited now, widely cited, as a truism. Software is eating the world. As with my accountant yesterday, he was bemoaning the fact that he can't go and get a coffee without people asking for it, you know, as he signed up to the app.

3:13Or they want feedback, which they then ignore. They want your data. Now, we call both this MIT report talking about all the AI pilots failing and Mark Andreessen's software is eating the world. We call those missives, open letters. And in the category world, those are points of view. So sure, Andreessen is shooting from the hip with his personal beliefs when he wrote that. The MIT report is somewhat different. It's based on a survey uh albeit small but it's been taken to be this massive point of view uh and i just haven't seen anything that's been as widely cited but if they if they are just povs you know points of view um and and you know fine they're points of view the issue is the points of view can become very powerful for sure as the integration of andreessen's you know comments into popular culture prove and points of view do define categories so they're certainly worth thinking about so we have to ask ourselves you know what is the correct response to how ai is affecting not that not just the tech world which we're all massively aware of course in tech but society in general yeah and there's so much to untangle and the issue with ai as i see it is it's all happening in real time you know there's a multi-billion dollar deal this week there's another one next week People are buying chips from each other.

4:36People are taking stakes in telco companies like Nokia. It's really running very fast. Therefore, when another once robust publication, not going to talk to its journalistic credentials at the moment, but Fortune. So Fortune talked about another report which says AI is ruining our attention spans. Sorry, Paul, what did you just say? Exactly. And just like humans, it seems that when humans get older, they suffer what could be regarded as brain rot. But this study claims that AI rots the brain in real time, producing low-quality text. And according to the study, the effects of this brain rot, forgetfulness that we're joking about here, lingers.

5:25But there's two gems, I think, in this report, aren't there? when we looked at it that just jumped out. It says, I quote, I quote, The gap implies that the brain rot effect has been deeply internalized and the existing instruction tuning cannot fit the issue. Stronger mitigation methods are demanded in the future, wrote the researchers. And there's another lovely one here, which is the gap implies that the brain rot effect has been deeply internalized. Yeah, right. And the existing structured tuning cannot fix the issue. Oh my God, I've got brain wrath. Exactly. You just said it twice. It's a peer-reviewed article, right?

6:07We should be fair. But with AI changing so quickly, there's such a dearth of clarity of what's actually going on. And the issue with such strong POVs, when you have a lack of people having a strong point of view, I don't think we lack a strong point of view, but lots of people do, is you become a bit of a bystander. And when that happens, what kicks in is the anchoring effect, as we know from Thinking Fast and Slow by Mr. Kahneman. Is that also something to do with something I've heard of, recency fallacy? Yeah. So yeah, the recency fallacy basically is that if you've heard something recently, and I can think of one particular national leader where this might apply to, the thing that you heard the most recently is the thing you believe the strongest.

6:54So you obviously need to take in a lot of inputs, which I'm sure this individual does. But the recency fallacy is that the thing you've heard the latest, and it could be the MIT report, it might be the Fortune article is what you believe. The anchoring bias, which works with this, is that the first person, the first time you read the POV, the first time you hear about the effects of AI rotting your brain, the first time you hear that software's eating the world, that's where you stay and you don't move on from there. But what is really interesting is that this report was mentioned many times on the stage at the recent Sifted Summit in London and also at BoardWave's recent excellent event also in London.

7:40I mean, frankly, it's everywhere. It's been used by naysayers to imply AI is a giant waste of money, a distraction used by those with second-generation AI tools and services to say, you don't get it yet, so let us sell you some more of our stuff. There's certainly a lot of AI washing going on. Now, the MIT report is the more interesting, I feel, of these two reports. Even though it's potentially less well-researched, it's certainly been influential, as you say, BoardWave and Sifter Summit. It was embarrassing how many times it was mentioned. And it states, despite the rush to integrate powerful new models, about 5 % of AI pilot programs achieve rapid revenue acceleration.

8:20The vast majority stall, delivering little or no measurable impact on profit and loss. This research is based on 150 interviews with leaders and a survey of 350 employees and analysis of 300 public AI deployments. So you could argue it paints a pretty clear divide between success and failure. Again, got an hour history here. There is a counter argument on this. This is a very new technology being deployed at an incredibly rapid rate and throwing sticky stuff at a wall comes to mind here. Is 5 % such a tragic success rate? I'm sure the early developers of steam boilers would have been very happy if in the first couple of years of experimentation they were actually getting 5 % of their boilers which didn't explode.

9:11But what that did do was actually encourage more experimentation. So the question we ask here, is this data at this stage so damning that it's going to encourage less experimentation? Yeah, so we're basically saying 5 % ain't that bad, guys. You know, it's pretty good. Just, you know, concluding where we're going and how society is going to deal with AI, I think it's super early, right? So we don't know what we don't know. And despite all of the crazy market hype about valuations, the effects on jobs, etc., there has to be more of a focus on how society copes with this information and whether we need a universal basic income, etc.

9:53And it also has to cope with the amount of slop, AI slop that's coming out there. And probably the most obvious example is AI fakes. But there's a lot of slop. And we need to, as a society, stop being observers here and get involved. And I think we should have a lot more of these points of view. There's less reason to fear AI than most people think. And we have to accept that AI changes the way we work. It may change the way we think, as per the Fortune article. And ultimately, it's going to change the way that we live our lives. So it's truly a society-level experiment. But it's also our experiment.

10:31Yeah, so if we continue to use AI, we will lose essential traits. You know, once you needed to be as strong as an ox to be a farmer or a truck driver, and that's just no longer the case because of technological development and automation. You know, before we all started to use calculators in the 70s, mental arithmetic mattered in exams. Now you are freed to do high-level mathematics with just the basic calculations being taken place by a silicon chip at your hands and on the desk. So, you know, what do we think of this? I think, you know, let's make sure we use AI to accentuate human intelligence, not replace it.

11:22are you still developing separate marketing plans for each of your different channels to market can you really sanction messages reaching the same customers from unconnected targets listen it's now the ai age things need to change positive is a different sort of ai powered agency from our very beginning we've integrated marketing communications from pr to events social to paid ads working smarter means working integrated let us show you what a truly joined up strategy could do for your tech business contact positive today with an email to hello at positive marketing.com so paul i hear you had an encounter with with a hero of mine foreign blonde eloquent who is she has made a lot of money for themselves and others who is that more fast and swift though Ah, not Taylor Swift, Herman Hauser.

12:15Well, you know everyone in tech is searching for the prophecy, that singular glimpse into the future. And there are plenty of noisy VCs out there looking at you, Mr. Stebbings. But I actually had the pleasure of listening to the real deal. Honestly, if anybody has earned the right to make predictions, and any European has earned the right to make predictions, it's Mr. Herman Hauser. Herr Hauser. I completely agree. I mean, Austrian by birth, but for me, the pivotal person in UK tech over the last 40 plus years, this guy just doesn't just talk about the future. He builds it from founding Acorn with Chris Curry to sharing Arm, which let's face it, is inside nearly every smartphone, to now backing over 60 deep tech startups with his Amadeus Capital.

13:04His track record is immense. Yeah, you could say I was a little bit gobsmacked and a little bit freaked out. Starstruck, even? Starstruck, even. I actually bumped into him briefly at the Sifted Summit in London. Total, total tech legend. But very cool, right? No fan mobs or screaming groupies, thank God. His legacy is going to outlast 100 times the flashier pop stars of this world. Yeah, and that's what makes him different. While some people are out there chasing hype and headlines, Herman's laying the actual foundations. He's like the anti-pop star of tech. And I mean that as a compliment. Yeah.

13:44And he's still pushing the edge, right? He was on stage talking about a$20 billion fund he's involved with. He's deeply involved in quantum. I think he's made six or seven investments already there. He looks for deep tech, looks for it to be super, super hard. that's not just ambition and hype that's putting your money where your mouth is that's betting on the future of europe leading the next wave of ai and quantum innovation yeah and what struck me is is just how grounded his his argument was and or is you know it's about better returns from european vc than us vc that's not widely known that's counterintuitive but it's bang on right Yeah, sure.

14:28And also equivalent IP being produced. But the problem is there's just a fraction of the capital chasing it. It's really classic investor logic, high potential, under-invested market, fill your boots. Yeah. And as he said, I think this is a direct quote, we do not have a startup problem. We have a scale-up problem. And whilst that can be fixed, he's also got a warning for the individual regions. So for the UK specifically, he's talking about munich catching up fast it's minted over 100 startups last year um and germany is not waiting around uh for everyone else to get their act together yes it certainly isn't and and you know they've got good reason uh to need a tech boom in germany given how the current economy appears to be stalling um you know as it is so over leveraged on on the auto industry but i mean if you think go back to Herman, you know, he's also very much into quantum.

15:26He's the most lucid person I've heard talk about quantum. Yeah, he just makes sense, you know, but he's backed seven startups already in that space. You know, he had a really, really interesting point about 100 qubits being the tipping point. Yeah, that's the magic number, right? Apparently, who knew? It's, you know, it's not sci-fi anymore. You know, we're talking five to 10 years to to massive disruption across cybersecurity, finance, AI, everything you can frankly possibly think of. Right. Yeah, and because with great thinkers like that, they join the dots for you. So he's joining the dots between, as you say, cybersecurity, finance, and AI.

15:59And what struck me was the way that he joined the dots on Quantum because he said that Quantum will turbocharge AI. That's not a soundbite. That's the roadmap in his head that we're charging towards. Yeah, and it's not just some, you know, spotty obsession with tech. He also just explained the structure of the new fund, where he's mixing institutional with industry money. It's clever. It's just smart. It's really clever. To me, it just reflects how the market's finally waking up to deep tech after nearly three decades of him being out there, doing it largely alone, it strikes me. And he made another amazing point, which was, if they just chucked 20 billion euro at deep tech, there'd be a bunch of people who'd soak it up, probably fail, and that would be the 20 billion spaffed.

16:49But they're only going to cornerstone the investment. So they will just put in one-fifth, and they'll want four-fifths of the money to come from private enterprise. So skin in the game, real commercial VCs, and the smarts that he brings and his team brings. Well, that's what happens is what he sort of joked about, which is, you know, after 27 years, the market's finally coming around to us. But I did sense he's still frustrated. The scale-up problem in Europe is real. U.S. startups get 70 times more seed funding. We're watching our IP walk overseas, and with it, the economic upside. Yeah, not just our IP, our founders, right?

17:31We talked about Revolut. and he compares what we do in deep tech to China's strategy, the Belt and Road Initiative. So basically, they'll go into a country that needs investment, needs some money, and they'll swap it. But of course, once they're in, you are a vassal state for their tech and you are just another market. It's a whole different way of going about things, but it's a strategy. Yeah, no hesitation there, isn't it? No gap between innovation and investment. and Herman's trying to fix that. And I love his investment framework. Simple, but really sharp. Market, people, then tech. Yeah, and what was shocking to me, and I think actually to his interviewer, was normally the order in which you put those three things is market, then tech, then people.

18:18And what he did was he said, first of all, it's the market, then it's the people that are going to ex-earn it. He looks for one superstar, he says, and builds the rest of the team around it. And then finally, the tech. Now, you might think him being a deep tech guy, he would start with the tech. That's the last thing he considers. And he justifies that because he says, an A-team with a C-tech idea beats an A-tech idea every time. And apparently we'll get even more insight in his upcoming book. Yeah. So he's eight chapters in of 12. Hurry up, Herman, get that out. I'm guessing it's going to be required reading for any serious investor in Europe and any category leader, for sure.

18:58Yeah, I mean, you know, no doubt. And good to hear he ended his talk on a high note. Quantum prizes going to European scientists and the EIC making real moves. Yeah, exactly. That's where he was representing the European innovation side of things. And the two Nobel Prizes recently, one went to an Austrian, which, of course, he's very proud of. Another one partially went to a Brit. So we have got the smarts. There's no doubt about it. just need to scale up yes i mean herman is showing that european tech can absolutely take on the world not just with ideas but with real deep scalable innovation it's not just about shaking off the height it's about doubling down on what actually matters yeah and apologies to all the swifties out there um as far as we're concerned deep tech is still deeply cool

19:51look there's captain hindsight right welcome back to captain hindsight where we talk about the tech the shape of the future and potentially breaking it based on the evidence of the past of the history of the tech business yeah so today we're asking the question i think a lot of people are actually thinking but most are not saying out loud are we already at peak chat gpt or should we call it chat gpt for reasons that will become apparent well i think you're just pronouncing the french word for cat there so that's fine let's call it what it is um chat gpt you know your taxi driver your granny everybody uh agrees it's uh dominating the llm space it's almost a generic it's become the category leader it's the default ai interface for millions of people um one of the fastest growing if not the fastest growing software product ever and open air keeps doing what a leader should it keeps you know huge rounds hardware deals innovation who knows what next and it seems um you know a massive uh program of acquisition and alliances and i think the latest one which is really interesting is um around agentic commerce protocol right so that's an instant checkout inside chat gpt where it acts as a shopping agent you say find me something or other and boom it shows you product results and you buy without leaving the app i can't help feeling that this is just all about convenience i also can't help feeling that amazon who has a similar sort of agentic offering called rufus seems to have you know missed the point here um you can see exactly where this is headed relevance based rankings that are very very appropriate to the person doing the prompting which allegedly gives you answers based on what the LLM discovers.

21:35We know that there's bias and you can own that. But what's going to happen is brands, specifically consumer brands, are going to start competing for visibility. And just like that, potentially, ChatGPT becomes the new Google, potentially the new Amazon. Yeah, and it's a classic platform arc. I mentioned this before, an American guy called Cory Doctorow actually coined the perfect general term for this phenomenon, inshittification. Something scatological we've mentioned before, as I said. In this case, though, because we understand the need for differentiation, we prefer to call it category crapification.

22:20Well, let's go down that dark and dingy alleyway. What happens here is, first, you get a product that's fabulous for users. Everybody loves it. like amazon like chat gpt and and often these are built out um these great products at the expense of any profit and that's how the platform scales um until a certain point it pivots and this is where the um crapification comes in um because it pivots um and it becomes uh something that's monetizable eventually it's just great for shareholders not for the people that originally loved the product. Everyone else is left navigating a cluttered, compromised, and, well, crappy experience.

23:03That's the story of Skype with Microsoft. That's the story of Google's search engine. Yeah, and, you know, that's the worry here. ChatGP2 starts out as this magical, frictionless, really useful tool and slowly morphs into a horrible marketplace where its only reason to exist is to buy and sell attention. And there's a track record. Google's buying in of the AdWords product took a couple of years. Facebook seemed magical and free for a long time, and now plastered with ads. And now, please subscribe if you want to keep your Instagram. So think about it. OpenAI launched ChatGPT publicly in late 2022.

23:42But guess what? Hot on the heels of the development of the agentic commerce protocol, PayPal has just announced their partnership with OpenAI, integrating its digital wallet directly into the LLM. So, starting next year, users will be able to shop and pay seamlessly within ChatGPT using BuyWithPayPal. Customers can browse, discover and purchase items securely, apparently, while verified PayPal merchants can list and sell products on the AI platform. Right? Great, great, great, great, great. But the logic goes, ChatGPT becomes a personal AI shopping assistant that can instantly check out. consumers trust paypal paypal handles all back-end payment validation and the merchant routing so sorry routing for american listeners um as ceo alice chris put it in the announcement it's a whole new paradigm for shopping um right so how many seconds before the brand ads appear now so chat gpt looks like it's setting a new record for ttcc you've heard that first here ttcc It is time to category crapification.

24:59The problem, it can't do anything else. Yeah, even ChatGPT has got to address the crazy valuation it has. All those people paying for its shares are expecting their money to be returned at some point. Last year's cash burn was a mere$8 billion. It's set to go up. And all that does is put off the day when the shareholders can get their money. and make them more anxious that they do get their money back. Of course, let's not forget that until very recently, OpenAI was structured as a for-profit entity controlled by a non-profit board. But only a few weeks ago, it restructured into a public benefit corporation.

25:42And that is, of course, known as OpenAI Group PBC. Now, why would you do that? Well, let's not be cynical, But a move like that, of course, paves the way for a potential IPO in the future. So what we're talking about here isn't crapification really, it's category. It's not about ads or cluttered UI. It's about what happens when a category leader needs or wants to shift from delivering value to the extraction phase, extracting it. There's a cycle, category innovation, category domination, category monetization, and sadly, category crapification. Absolutely right. But domination should not mean degradation.

26:26You can lead a market without eroding the product that got you there. The problem is the incentive structures, as it currently is in some of these companies, are in the other direction. if users don't leave and let's be honest most will not leave overcoming the inertia that limits change and stops innovation is just too hard the temptation is to extract the profits and just keep growing the bottom line retail banks have been doing this for years and the reality because i think what you've just said is the future unless there is pressure from from users regulators or maybe from inside the industry itself a cycle of the rawest form and most quickly enacted form of capitalism ensues.

27:09Now, it's the difference between value creation, the process of developing products or services that provide benefits and meet customer needs, very academic but true, and value capture, the process of converting that created value into profit for a company. And that's where all the wheels can actually fly off. The idea is that businesses actually must do both to succeed, create compelling offerings that attract the customers and then implement strategies to convert that customer value into revenue through pricing and market position. But in the age we have entered, that of tech totalitarianism, enabled by the pervasion of the internet and the speed of innovation possible in weightless industries the initial offerings are not what the customers originally signed up to and you know the the ultimate offerings are not what the customers originally signed up to and they have to get used to being force-fed a diet of crapified as we used to call it in enterprise software bloatware um and this is because there's a monopoly of user attention and and wallet share is sought for at warp speed.

28:22So ChatGBT is doing exactly this, right? It's becoming much more than a chatbot, very useful piece of software. It's positioning itself somehow and wants to be seen in a category of the next interface layer for commerce, for productivity, for search, for education. Yeah. So guess what? This makes this moment critical because if this turns into another ad-saturated engagement maximizing platform, we've learned nothing from the last 20 years of the internet and its history probably worthwhile at this stage just going into some of that history so that people really understand how crapification works yeah let's go to the end game of category crapification so um i'll take an example the uk tv sports they've maxed the um the amount of squeeze they can give on the with the market and sky built and some would say built the premier league but built a paid-for sports channel within a licensed regulated market and now is seeking to extract with more and more layers of service.

29:21I mean they're an monopoly essentially now because the competition BT exited the market and now their replacement TNT has also bailed out you know and I think you know according to our contacts no phone is currently ringing off the hook for the vacant competitor slot so you know it's so tightly has sky sealed it up also because of the layers and various um layers of subscription you can get the phones are literally ringing off the hook with their sales guys trying to persuade us to to go and pay for a an inch of idle crapified product and if you look at the catch-up um tv channels now prime in the uk paramount and sky box sets when you access them via now suddenly they've got up it's in them you want to get rid of the adverts pay us the money.

30:05They attracted an audience, they monetized it, and now they're crapifying it. Let's be clear about this from a fiscal point of view. These are price-gouging monopolies. Microsoft with Skype did it in the past. See how that actually ended up. CA did it with anything they acquired, anything they touched. Even Broadcom with VMware users totally up in arms. They don't care. It's like, we got you, you pay up. And these are examples of rentiers, which is a natural sort of end game for crapification because you basically give up on caring about innovation, you give up on caring about customer service, you just extract the rent rather than moving on and continuing to innovate.

30:49Yeah, and it is a powerful strategy for big tech. And just to be a bit European for a moment, it can hurt the most in terms of instances like Apple buying UK-based Shazam to Google taking over the entire team based in the UK at DeepMind. It's worst of all for an up-and-coming, some would say, region like Europe where we're trying to build innovation and we're being beaten down by big tech's cruppification. Indeed. But let's be positive for a moment. All is not lost. It's not lost. No, it isn't. Let's have a look at some counter examples. Look at Spotify, who have stuck to their guns in music distribution.

31:37And up until now, Strava has been focused on personal health. And there's amazing people at UiPath that are still RPAing out of Romania. you know, hold out and build, stay better people. Please do. We can't afford to repeat this cycle. We don't need another Google meta, et cetera, crapifying our lives. We need a new model for tech leadership. One where owning the category comes with some sort of responsibility to protect the user, not just shareholder value. I'm sounding like I'm pro-regulation, but you get what I mean. Yeah. So, you know, go back to the original thing, which we started this section on, which is, you know, are we at peak chat GTP, you know, chat to GPT as we know it, Jim, you know, maybe or, you know, maybe, maybe, maybe, or maybe, and I honestly, I have to believe this is going to be, is going to, is what's going to happen.