(more AI stuff)
Existential risk, AI, and the inevitable turn in human history
- Hardly anyone you know, including yourself, is prepared to live in actual “moving” history. It will panic many of us, disorient the rest of us, and cause great upheavals in our fortunes, both good and bad. The good will considerably outweigh the bad, but the absolute quantity of the bad disruptions will be high.
- Since we are not used to living in moving history, and indeed most of us are psychologically unable to truly imagine living in moving history, all these new AI developments pose a great conundrum. We don’t know how to respond psychologically, or for that matter substantively. No matter how positive or negative the overall calculus of cost and benefit, AI is very likely to overturn most of our apple carts, most of all for the so-called chattering classes.
- The reality is that no one at the beginning of the printing press had any real idea of the changes it would bring. No one at the beginning of the fossil fuel era had much of an idea of the changes it would bring. No one is good at predicting the longer-term or even medium-term outcomes of these radical technological changes (we can do the short term, albeit imperfectly).
- When people predict a high degree of existential risk from AGI, “arguing back” on their chosen terms is not the correct response. Radical agnosticism is the correct response, where all specific scenarios are pretty unlikely.
- If someone is obsessively arguing about the details of AI technology today. Don’t be suckered into taking their bait. The longer a historical perspective you take, the more obvious this point will be. We should take the plunge. We have already taken the plunge. We designed/tolerated our decentralised society so we could take the plunge.
PS: I love this “moving history” phrase to think about what is happening with AI now
Malleable software in the age of LLMs
- The ability of normal people to fully harness the general power of computers without resorting to the complexity of normal programming. We might start to see some surprising changes in the way people use software.
- Normal computer users have their AI create and execute scripts dozens of times a day, to perform tasks like data analysis, video editing, or automating tedious tasks.
- People use AI to create entire GUI applications just for performing a single specific task—containing just the features they need, no bloat.
- Businesses develop more software in-house that meets their custom needs, rather than buying SaaS off the shelf, since it’s now cheaper to get software tailored to the use case.
- Consumers and businesses demand the ability to extend and mod their existing software, since it’s now easier to specify a new feature or a tweak to match a user’s workflow.
- Take the best parts of the different applications you like best, and create a new hybrid that composes them together.
- Incumbent Benefits: distribution, data (large + proprietary), capital, talent
- Startup Benefits: focus, speed, no existing business to defend, less repetitional risk
- Battle between every startup and incumbent comes down to whether the startup gets distribution before the incumbent gets innovation.
- Startups should aim to hone in on specific problems or vertical-use cases and build solutions for those. It’s less likely that incumbents will tackle these specifically and there may be more room to innovate by rethinking the workflows and processes for specific problems.
- It’s critical that startups build more than a simple wrapper around the foundation models. One way to do so is by leveraging additional data (through integrations, purchasing external data, standardising data into structured formats)
- Startups should be thinking about what the end-state workflows could be, and how to iteratively work towards that. It’s likely that they’ll have to insert themselves into existing workflows as opposed to completely reimagining them immediately, but it’s important to always think about the bigger picture.
- Leverage new demand channels Some of the incumbents may be more hesitant to build things like ChatGPT plugins or leverage those potential forms of demand. As a startup, it’s important to use any distribution benefits to your advantage.
- Innovate on user interfaces; The ideal user interfaces to use these AI products are likely not fully developed. Incumbents may be more hesitant to completely redesign their product especially early on, but startups have more of that luxury since they are starting from scratch.
The adoption of AI will change the software value chain
- We are in the early innings of a new platform shift, this time powered by AI. This platform shift is predicated on two things: first, within a decade, AI will be a core component of every piece of application software that’s built and shipped. Second, this wave will be powered by the widespread adoption of foundation models.
- Similar to how the rise of cloud computing disrupted software value chains and led to the emergence of new business models. We don’t exactly what these new business models will be, but we have some ideas about how the software value chain is changing and are starting to see early signs of new models emerging.
- Similar to the rise of cloud vendors a decade ago, we see a new class of companies emerging that will specialise in maintaining the latest up-to-date foundation models based on the most cutting-edge research.
- At the application layer, business models will shift over time to capture more customer-specific value unlocked by AI. Instead of simply paying for use à la SaaS, we envision a world where customers can pay based on the level of customization or personalization within the product, for instance, a model fine-tuned to the industry, the organization, or even the individual.
- This also creates an opportunity for a new type of managed infrastructure company, which helps software companies provide hyper-personalization on top of the foundational model providers. There is huge complexity in fine-tuning at scale, and lots of interesting questions around generating, managing, and administering large volumes of data and models.
The ugly economics behind Apple’s new Pay Later system
- Apple is getting into the “buy now, pay later” (BNPL) business with its new Pay Later service built into Apple Pay and Apple Wallet.
- On the surface, BNPL services seem harmless, as some come with no interest and allow for an easy way to pay back a big purchase in chunks. However, this kind of service becomes easy to abuse when used for nonessential purchases.
- BNPL services highlight its popularity among Generation Z; according to [SFGate](http://(https://www.sfgate.com/news/article/influencers-lead-Gen-Z-into-debt-17142294.php), 73 per cent of BNPL customers are part of this generation, and around 43 per cent of them report missing at least one payment. Another survey from DebtHammer shows that 30 per cent of users struggle to make their BNPL payments, and 32 per cent report skipping out on paying rent, utilities, or child support to prioritize their BNPL bills.
- According to data viewed by the outlet, the average Affirm customer spends $365 on a single purchase, as opposed to the $100 average cart size recorded in 2020.
- Missed and late payments, coupled with a volatile economy, have led Klarna’s valuation to reportedly tumble by a third — from $46 billion last year to $30 billion — and also has Affirm’s share price to drop. Last month, [Klarna laid off 10 per cent of its employees due to “a highly volatile stock market and a likely recession.”
- Apple’s Pay Later is on track to receive the same sort of scrutiny as it injects itself into an uncertain sector when inflation is spiking, and consumers are struggling to pay for everyday goods. But it also normalises the BNPL practice by building the concept straight into the iPhone, posing a risk to consumers and competing businesses.
Substack edition
Existential risk, AI, and the inevitable turn in human history
Hardly anyone you know, including yourself, is prepared to live in actual “moving” history. It will panic many of us, disorient the rest of us, and cause great upheavals in our fortunes, both good and bad. The good will considerably outweigh the bad, but the absolute quantity of the bad disruptions will be high.
Since we are not used to living in moving history, and indeed most of us are psychologically unable to truly imagine living in moving history, all these new AI developments pose a great conundrum. We don’t know how to respond psychologically, or for that matter substantively. No matter how positive or negative the overall calculus of cost and benefit, AI is very likely to overturn most of our apple carts, most of all for the so-called chattering classes.
The reality is that no one at the beginning of the printing press had any real idea of the changes it would bring. No one at the beginning of the fossil fuel era had much of an idea of the changes it would bring. No one is good at predicting the longer-term or even medium-term outcomes of these radical technological changes (we can do the short term, albeit imperfectly).
When people predict a high degree of existential risk from AGI, “arguing back” on their chosen terms is not the correct response. Radical agnosticism is the correct response, where all specific scenarios are pretty unlikely.
If someone is obsessively arguing about the details of AI technology today. Don’t be suckered into taking their bait. The longer a historical perspective you take, the more obvious this point will be. We should take the plunge. We have already taken the plunge. We designed/tolerated our decentralised society so we could take the plunge.
PS: I love this “moving history” phrase to think about what is happening with AI now
The ugly economics behind Apple’s new Pay Later system
Apple is getting into the “buy now, pay later” (BNPL) business with its new Pay Later service built into Apple Pay and Apple Wallet.
On the surface, BNPL services seem harmless, as some come with no interest and allow for an easy way to pay back a big purchase in chunks. However, this kind of service becomes easy to abuse when used for nonessential purchases.
BNPL services highlight its popularity among Generation Z; according to SFGate, 73 per cent of BNPL customers are part of this generation, and around 43 per cent of them report missing at least one payment. Another survey from DebtHammer shows that 30 per cent of users struggle to make their BNPL payments, and 32 per cent report skipping out on paying rent, utilities, or child support to prioritize their BNPL bills.
According to data viewed by the outlet, the average Affirm customer spends $365 on a single purchase, as opposed to the $100 average cart size recorded in 2020.
Missed and late payments, coupled with a volatile economy, have led Klarna’s valuation to reportedly tumble by a third — from $46 billion last year to $30 billion — and also has Affirm’s share price to drop. Last month, [Klarna laid off 10 per cent of its employees due to “a highly volatile stock market and a likely recession.”
Apple’s Pay Later is on track to receive the same sort of scrutiny as it injects itself into an uncertain sector when inflation is spiking, and consumers are struggling to pay for everyday goods. But it also normalises the BNPL practice by building the concept straight into the iPhone, posing a risk to consumers and competing businesses.
Malleable software in the age of LLMs
The ability of normal people to fully harness the general power of computers without resorting to the complexity of normal programming. We might start to see some surprising changes in the way people use software.
Normal computer users have their AI create and execute scripts dozens of times a day, to perform tasks like data analysis, video editing, or automating tedious tasks.
People use AI to create entire GUI applications just for performing a single specific task—containing just the features they need, no bloat.
Businesses develop more software in-house that meets their custom needs, rather than buying SaaS off the shelf, since it’s now cheaper to get software tailored to the use case.
Consumers and businesses demand the ability to extend and mod their existing software, since it’s now easier to specify a new feature or a tweak to match a user’s workflow.
Take the best parts of the different applications you like best, and create a new hybrid that composes them together.
Startups vs Incumbents in AI
Incumbent Benefits: distribution, data (large + proprietary), capital, talent
Startup Benefits: focus, speed, no existing business to defend, less repetitional risk
Battle between every startup and incumbent comes down to whether the startup gets distribution before the incumbent gets innovation.
Startups should aim to hone in on specific problems or vertical-use cases and build solutions for those. It’s less likely that incumbents will tackle these specifically and there may be more room to innovate by rethinking the workflows and processes for specific problems.
It’s critical that startups build more than a simple wrapper around the foundation models. One way to do so is by leveraging additional data (through integrations, purchasing external data, standardising data into structured formats)
Startups should be thinking about what the end-state workflows could be, and how to iteratively work towards that. It’s likely that they’ll have to insert themselves into existing workflows as opposed to completely reimagining them immediately, but it’s important to always think about the bigger picture.
Leverage new demand channels Some of the incumbents may be more hesitant to build things like ChatGPT plugins or leverage those potential forms of demand. As a startup, it’s important to use any distribution benefits to your advantage.
Innovate on user interfaces; The ideal user interfaces to use these AI products are likely not fully developed. Incumbents may be more hesitant to completely redesign their product especially early on, but startups have more of that luxury since they are starting from scratch.
The adoption of AI will change the software value chain
We are in the early innings of a new platform shift, this time powered by AI. This platform shift is predicated on two things: first, within a decade, AI will be a core component of every piece of application software that’s built and shipped. Second, this wave will be powered by the widespread adoption of foundation models.
Similar to how the rise of cloud computing disrupted software value chains and led to the emergence of new business models. We don’t exactly what these new business models will be, but we have some ideas about how the software value chain is changing and are starting to see early signs of new models emerging.
Similar to the rise of cloud vendors a decade ago, we see a new class of companies emerging that will specialise in maintaining the latest up-to-date foundation models based on the most cutting-edge research.
At the application layer, business models will shift over time to capture more customer-specific value unlocked by AI. Instead of simply paying for use à la SaaS, we envision a world where customers can pay based on the level of customization or personalization within the product, for instance, a model fine-tuned to the industry, the organization, or even the individual.
This also creates an opportunity for a new type of managed infrastructure company, which helps software companies provide hyper-personalization on top of the foundational model providers. There is huge complexity in fine-tuning at scale, and lots of interesting questions around generating, managing, and administering large volumes of data and models.