- Fintech Brainfood
- Posts
- Fintech 🧠 Food - The Future of Trust
Fintech 🧠 Food - The Future of Trust
Plus, Apple savings launches, Twitter does stock trading & why AI hallucination is a feature not a bug
Hey everyone 👋, thanks for coming back to Brainfood, where I take the week's biggest events and try to get under the skin of what's happening in Fintech. If you're reading this and haven't signed up, join the 30,089 others by clicking below, and to the regular readers, thank you. 🙏
Hey Fintech Nerds 👋
What do Wrexham, Cricket, and a Fintech company called Neuvi have in common?
Feels like a great quiz question.
The answer, of course, is Ryan Reynolds, who's becoming a great marketer. He finds a story he can believe in and makes it work for the brand.
There are so many great stories to tell in Fintech, many of which didn't get told when the story was valuation. The story today is trust. Consumers are looking for brands they trust and need unique marketing approaches. The key to marketing is authenticity and storytelling. Ryan Reynolds has mastered both.
Trust is everything.
With the banking crisis, the fall of FTX, and scar tissue from 2008 relevant in many minds, trust is also in short supply.
Meanwhile, Apple, another master storyteller, has entered the chat.
Apple has focussed on user privacy and placing identity, healthcare, and finance on your device. They command a premium from suppliers and consumers because they build beautiful experiences and embody trust.
Apple has a shot at owning identity, winning trust, and building an ecosystem. This is the single biggest threat to traditional financial institutions.
Finance needs better storytellers.
Because there are so many stories to tell.
PS. Trying out a new segment this week 🐟 Tales from the Fish Tank, would love feedback 🙏
PPS. Thanks to everyone who came to the first ever Fintech Brainfood x This Week in Fintech Happy Hour. Over 150 of you showed up. London Fintech is BACK. Watch this space for more.
Here's this week's Brainfood in summary
🐟 Tales from the fish tank: Money Laundering in 3 Steps
📣 Rant: The future of trust in finance
💸 4 Fintech Companies:
Lillyfunds - Retirement benefits as a Service
Renew - Renter renewals as a service
Bunce - Payment Operations as a Service (Nigeria)
Cyrus - The personal digital security company
👀 Things to Know:
📚 Good Read:
Chat GPT Gets a Computer (please read this section even if you usually skip it 🙏)
Tales from the fish tank 🐟
I joined Sardine because I believed believe fraud & compliance would become the new user growth in Fintech. But more than that, it's a blindspot for me, and I wanted to learn, and wow, talk about drinking from the firehose. I wanted to share something that Krisan, our head of compliance, taught me this week.
Canonically, money laundering has 3 phases (placement, layering, and integration.)
Placement often looks like fraud (e.g., depositing drug money in ATMs or claiming a chargeback for delivered items).
Layering is a game of hiding the ball, moving money around the system quickly and through multiple steps to make it harder to track.
Integration is making money look legal, for example, by using it to purchase from a company operated by a network of criminals.
Fraud for one bank is money laundering for another and appears like a simple payment for an SMB focussed bank that suddenly has a ton of chargebacks or ACH fraud issues.
The kicker?
No one bank or Fintech has visibility of the whole thing. Internal bank divisions are often split between fraud and compliance, and communication isn't always great. The more the industry collaborates on this stuff the better.
Weekly Rant 📣
The Future of Trust
What do you trust?
That's a very personal question. Family? Friends? Your gut?
Trust is a promise made and kept.
The psychology of trust is critical to understand in finance and money because money is vital to our basic motivators and requirements. Shelter, safety, food, and water have a cost associated in many places. To survive and thrive, we need these things.
The ultimate rug pull is when trust is shattered about money. We place our money in banks and take for granted that it will still be there when needed. But behind the curtain, money moves around the banking system like ocean currents. It seems permanent, but it's also a complex system that impacts everything on the planet.
Banking crises, bipartisan politics, inflation, and instant 24/7 social media have changed the nature of trust in ways we're only beginning to grasp.
Today trust isn't just about large institutions with marble halls sounding official. It's not just about the opposite of that and only trusting a TikTok influencer.
It's halfway between the two.
This is critical to understand if you're building a financial services brand. Different customer segments expect different things. Moving fast and breaking things won't work for all customers and stakeholders, especially after a banking crisis.
I think Apple, and maybe even Elon have a shot at making a real dent in the financial services industry because feature velocity is a power law. Companies that ship products faster grow faster (assuming they're good products).
So how do you balance that if you're a large incumbent, startup, or somewhere in-between? In a regulated industry.
When complete failure is catastrophic, and no failure is a gradual slide into irrelevance. What is trust?
Let's dive in 🏊♀️
What trust used to mean.
Stability
Consistency
Professionalism
The problems with the traditional approach to trust
The lack of innovation creates gradual erosion of a market position
De-risking is exclusionary
The lack of transparency means some issues are never resolved
The customer smells BS
The new approach to trust
Real-time
Transparent or Private
Storytelling
The problems with the new approach
More harm
More panic
Leaving a generation behind
Messaging around inclusion can quickly become inclusion-washing.
Lessons for Incumbents
Get better innovation through consistency
A data-driven approach to de-risking
Finding the right place for transparency
There is zero downside to being a better storyteller
Lessons for Growth Companies
Attack the fraud & compliance problem with data-science:
Find the balance of reassurance through transparency:
Understand older or corporate audiences:
Prove inclusion and risk management with transparency:
How do we get there?
1. What trust used to mean 🤝👴
Every time I speak to an executive team at a bank, they tell me about the importance of the "trust" they have from their customers.
If trust is a promise, what did traditional financial institutions promise?
a) Stability. The bank will remain solvent, and your deposits will still be there. You can rely on us to store money and move it wherever needed.
b) Consistency. We'll be around for a long time; you will be made whole if something goes wrong.
c) Professionalism. "We take this seriously because you matter to us."
These promises appear with a very early 20th-century interpretation, like large, sturdy buildings with vaults (branches), human relationships, and a don't-rock-the-boat approach to risk management.
When a crisis does hit, a sense of sounding official, professional, and buttoned up is projected—official press releases, scant on detail, heavy on reassurances. Imagine your bank has an outage. You might get a notification in your app, and the website will have an official release. But it's unlikely there'd be an engineering blog or company CEO tweeting about the precise technical cause and how they fixed it.
Seeing behind the curtain isn't always pretty. Celebrities take the service entrance; doctors mess up your surgery, and during an outage at a bank, people scramble to figure out WTF happened and fix it ASAP.
But the aim is to be like a swan. Hide all of that work and present the graceful front.
Diligent professionals operate things at scale and can't fail.
Institutions would rather appear professional and move slowly than expose the messy chaos.
When the risk you're managing is avoiding failure at all costs, often the optimal thing is to change nothing. Don't take on the risky client sectors because things might go wrong.
I remember speaking to one executive team who playfully but pridefully suggested they're "not the fastest land animal, but that's ok."
Older people, the vulnerable, and large businesses who also can't afford to have failures want this. Stability and professionalism are features, not a bug to many in the economy.
Considering recent failures, the management team had a point. That bank is still here, still big, and profitable.
But it's also not growing.
2. The problem with the old form of trust 🤝🤔
The world is changing.
a) The lack of innovation creates gradual erosion of market position. Moving slowly is playing not to lose, which is a good strategy in zero-sum games. If, as an incumbent, you assume your competitors are other organizations like you, the long-term game makes the most sense.
The problem is the economy is not a zero-sum game. New businesses rarely attack head-on. Clayton Christensen's innovator's dilemma describes innovation starting at the edges, serving the clients and segments that are underserved today and overcharged.
When your job is to deliver quarterly results and avoid outages, serving customers who aren't profitable in your existing business model makes no sense.
That is a rational decision.
It is also an opportunity cost.
Brick by brick, that cost adds up.
b) De-risking is exclusionary. De-risking is deciding that a customer segment or industry is "high risk" and that the organizational policy is not to provide services to those customers. One example might be low credit scores; another might be migrants or businesses involved in cross-border payments.
Incumbents can overcome these risks with enhanced due diligence (EDD) and must "treat customers fairly" and give equal access under various laws. But opening accounts and lending isn't just about equal access; it is about preventing fraud, money laundering, and terrorist financing.
And unfortunately, segments of the population that are low-income or industries like cross-border payments are high risk for lending or account opening. They have more fraud and are less likely to pay back borrowing.
When your job is managing risk and maintaining regulatory compliance, de-risking is a rational decision.
It's also opportunity cost.
Brick by brick that adds up.
c) The lack of transparency means issues are never resolved. How often I hear stories of incumbents not wanting to open a can of worms and expose the giant risks hiding in plain sight blows my mind. How many customers would fail KYC/AML today if you had to re-KYC everyone? Let's leave that alone, shall we?
How many accounts are dormant? Let's not share that too widely.
This quiet avoiding embarrassment is a corporate response by middle managers acting rationally. Getting anything live in these large organizations is incredibly difficult. Meeting annual KPIs to hit targets and not messing up is enough of a problem without creating new ones.
If your goal is stability, it is rational not to create new problems.
More opportunity cost.
d) The customer smells BS. Digital changed the game. Information now moves faster than a bank press release. A bank run can happen just as fast as the news breaks.
From the financial crisis through occupy Wall st to today, the gradual erosion of the "official" position is true regardless of where you look on the political spectrum. If you aim to calm markets and reassure investors and vulnerable customers, being official and professional is a rational choice.
But more opportunity cost. Brick by brick.
3. What trust started to mean 🤝🧒
The customer growing up with digital is pushing 40, making up most of the workforce. Digital brands make three promises.
a) Real-time. It will happen right now. We're responsive and available.
b) Transparent or Private. If something goes wrong, we'll tell you instantly, and increasingly we'll differentiate on our ability to protect your data
c) Storytelling. We align with your values. We're authentic.
These promises seem fluffy to the older generations. For real-time, what is the value added in something happening so fast anyway? You kids today need to slow down.
Well, the problem is they can't.
The internet made the world move faster, and competing means moving every bit as fast. If your website is slow, you'll get fewer sales. You're missing clicks if you miss that big news story as a creator.
Transparency and privacy are nuanced topics. But notice how Apple, in particular, is pushing the privacy narrative and how Whatsapp and others "protect the right to end-to-end encryption." Privacy is now a luxury good. We see this show up in Crypto, where all transactions are public records (transparent), but your identity is private.
The most interesting example is how Elon Musk tweets the precise cause of a Twitter outage. Far from aiming for professionalism, he's appealing to engineers and an audience who find knowing why something builds trust, even if something went wrong.
Elon is also a storyteller. Like Ryan Reynolds and Apple, the storytellers tap into their customer's beliefs and values and build trust that way.
Wanting the data and transparency seemingly contradict the desire for storytelling, but they're related. If you share the data, you build trust and consistently align with the values of your customer base.
4. The problem with the new form of trust 🤝🤔
or every celebrated founder, there's a fall from grace. This Forbes image has floated around my Twitter feed as a classic example and cautionary tale of what happens when we buy into a story and narrative too easily
a) More harm comes from more risk-taking. Without risk, there is no innovation; with risk, there is more potential for harm.
Moving fast and breaking things is terrible when dealing with essential human needs like safety. Fintech companies that have rightly identified the opportunities left on the table by incumbents also create more risk, and more risk can create more harm.
Great storytellers can also be frauds. This Forbes image has floated around my Twitter feed recently, and with FTX, Theranos, and Frank we saw cautionary tales of buying too deeply into the narrative and the consequences of not being professional and doing due diligence.
This is why regulators have to push back, and there is no smoke without fire. Crypto and Fintech have massive fraud issues.
b) More panic. Your transparency and speed of change are someone else's panic. Some people don't want chaos; they want stability. Most people don't want chaos regarding their entire life savings. If the conversations between SVB and the FDIC were live-streamed, I'm sure the panic would have multiplied. And yet, here we are, the problem got solved.
Speed and transparency can create more panic than it solves in the wrong context.
c) Leaving a generation behind. Often the best customer of an incumbent is older, values the human relationship, and may worry that digital is less secure. We can't assume that digital is default inclusive because it takes more risks. Incumbents play a critical role in serving populations that need something else.
Moreover, in finance, older customers tend to have better credit scores, own assets, and present far lower lending risks in many cases. The rise of real-time payments is a net risk to this generation from scams.
d) Messaging around inclusion can quickly become inclusion-washing. How many Fintech companies are "democratizing access to finance?" It's something they nearly all say. On the one hand, I doubt we would have cashflow-based underwriting as an industry norm if it weren't for Fintech companies. Conversely, Fintech companies also top the league table for high-risk of fraud and delinquencies.
5. Insights for Incumbents 💡
Large organizations receive criticism even though just staying in business is objectively challenging. An incumbent is too regulated and has too many vulnerable or large clients that rely on it to adopt startup approaches. They'd likely lose businesses if they started failing fast or dropped their professional aura for storytelling.
Taking each problem, we can contrast the current approach, the startup approach, and perhaps something in-between.
Note: I've tried to label this "tech" default; it's not the default for every tech company. But it's designed to illustrate what both sides can learn from each other.
a) Better Innovation through consistency: Instead of accepting eroding market share or making a sweeping decision to default-real-time, incumbents could use their consistency to target consistently moving faster as a KPI. This will take years to get right. But it's a powerful north star. Many have already headed this way in the larger banks, but how many governance processes are measured in months?
At scale, that makes sense, but intentionally focussing on pace is a north start could be powerful. Remember how during the pandemic clinical trials happened fast? The pharmaceutical sector still took precautions, but stuff got done.
The same was true in banks; technology roadmaps shunted forward half a decade in weeks. Across much of the industry, that has reverted back to normal but doesn't have to. There is a middle ground.
In March 2015, the UK Government Office for Science published a report suggesting the finance industry adopt a "clinical trial" methodology for higher-risk Fintech companies and businesses.
Why can't incumbents do the same internally? Today things either happen in a lab to avoid worries about compliance or come with all of the heavy machinery of a full live production launch. Most institutions still lack a middle ground. Innovation needs a way to graduate from daycare to school to its first job. And providing that path requires consistency.
b) A more data driven approach to de-risking: Instead of the blunt instruments of labeling entire sectors "high-risk," intentionally target risk teams to become better data scientists. Incumbents can be truly exceptional at data science. EWS is a modern miracle of privacy and fraud prevention through cutting-edge data science (although coincidentally, exclusionary because it is only available to the largest banks).
Again this requires consistency and a culture and mindset shift. The high-risk sectors and segments get that label for a good reason because they are high-risk. But the middle ground is consistently pushing the barriers of data science and not accepting the same old provider, same old data, same old outcome cycle.
c) Finding the right place for transparency: New technology here may help. Regtech is a fantastic development, and I know companies like Hummingbird, Alloy, Unit21, and Sardine* are now live with banks.
The entire Fintech infrastructure ecosystem is now in play for banks. Large core providers like FIS, Finastra, and Temenos actively partner with Fintech companies, but often, there's a procurement or priority mismatch.
Perhaps the best business case I'd advise any incumbent CEO to focus on is getting better at procurement. If institutions can get better at procurement, they'd get much better at everything else.
d) There is zero downside to being a better storyteller: I doubt Jamie Dimon will start tweeting about recent outages, but his annual letter is a model to think about. Consistent leadership can create consistent narratives and results. It's a different way to be a storyteller, presenting the positive and occasionally addressing the negative.
Addressing negatives in a more head-on way demonstrates authenticity without lacking professionalism. Another model here is Microsoft under Satya Nadella. Here is an older organization that has broken some of its historical taboos (like supporting Linux in Azure) and maintained, if not even grown, its market share with enterprise customers.
Does the world need another advertisement or corporate brochure straight out of the 90s? Nope. It would be weird if a bank tried to partner with Mr. Beast or a Kardashian, but there's a middle ground here again.
6. Insights for Fintech Companies 💡
The last 12 months have taught every Fintech company that VC cash for growth doesn't last forever and that the weird, counterintuitive nuances of finance infrastructure and regulation are utterly critical. To survive and thrive.
But perhaps there's a lot to learn too from incumbents. They've been around for a long time for reasons. The goal is not to copy what they do but why they do it.
a) Attack the fraud & risk problem with data-science: Some Fintech companies are amazing at this, but sadly others were either naive to the issue or didn't focus on it as a priority. The CEO of Sardine* Soups once said to me, "All fraud problems are data science problems." You solve them with more data and link that data together better.
Historically fraud & compliance are silos. Break the silo, break the black boxes, and build better models (again, some do this but not everyone can).
There's also much to be said about incumbent approaches to high-risk sectors and de-risking. Everyone is a genius lender in good market conditions. This market correction will show us who did innovate.
b) Find the balance of reassurance through storytelling: many Fintech CEOs do this well today, but it relates to the next point as you grow your audience changes. This impacts how you communicate through the product, marketing, and what you can assume a customer will understand.
c) Understand older or corporate audiences: Fintech companies that target these audiences have a head start. But when you think about at-scale companies with customers pressured into signing up by a scammer or signed up during the pandemic. How do you treat that customer differently?
Segmenting audiences is powerful.
Some incumbent companies do this in marketing and the products they offer. Not all products have to be uniform. Some consumer Fintech companies need to find new revenue; product line and segment extension are worth considering.
d) Prove inclusion and risk management with transparency: let's ban "democratizing access to finance" and put it in the same bin as "omnichannel" and "exchanging synergies." Trying to improve financial health is not exclusive to Fintech companies, and it undermines the great work of teams at incumbents who do incredible things.
However, there is still a ton of work to do.
I'd love to see Fintech companies share more data about risk levels, delinquencies, and high-risk populations they can address. The state of financial health for most consumers, regardless of income, is still awful. How can we better benchmark that over time in a way that's not just regulators and agencies publishing data?
7. Working together for a better financial system 🏧💸
Have you ever seen a Bison herd crossing a river with steep gauges at either side? They'll thrash across the river and struggle to find the way back up the other side. Individuals keep trying, failing, and repeating until one Bison finds a path out of the river.
Then within seconds, the whole herd follows that path, and within minutes, they're across the river. In startup land, we call this blitzscaling. They keep moving incredibly fast until something works.
Humanity is a herd too. We are more effective when urgently facing a crisis and forced to work together. It's not pretty; often, it's ugly.
But that's progress.
In times of instability, progress can be a dirty word.
Failing faster works if you limit the blast radius. The Falcon Superheavy rocket exploded shortly after take-off and experienced several failures this week. Thank goodness nobody was on board at the time. The innovator's approach to try things and fail improves the learning speed vs. the rigor and planning of NASA suppliers like Boeing traditionally use.
In times of crisis, like the pandemic, progress gets shoved forward by decades in weeks.
What if more transparent incumbents figured out risk-taking and limited the blast radius?
What if we had tech companies who knew when to be professional and focus more on risk?
What if incumbents and Fintech companies found the right ways to collaborate and raise the bar of risk management, transparency, and financial health for the economy?
I think we'll get there.
But the process is messy.
And there's so much opportunity in cleaning up a mess.
ST.
4 Fintech Companies 💸
1. Lillyfunds - Retirement benefits as a Service
Lillyfunds is a retirement platform for modern businesses competing with traditional 401k offerings. The platform is aimed to scale from teams of 5 to 500,000 and covers both W2 (full-time) and 1099 (contractor) staff. It provides a dashboard for staff to view their current savings and projected income at retirement. The platform also features spend tracking and rewards with merchants.
🤔 Nobody wants more admin as employers or employees. Lillyfunds can bake in many of the traditional benefits of a 401k for employees (like employer matching and auto deduction from payroll). An IRA also has greater savings options and lower fees (which Lilly passes on to employers in this case). It's interesting to see someone productize the IRA this way and turn it into the low-cost, more flexible 401k. Imagine if Lillyfunds had a partnership with Deel or Remote.com?
2. Renew - Renter renewals as a service
Renew provides property management companies with a simple interface to help renters renew an existing property or move to a new one in their portfolio. Renters can renew in a few clicks, quickly move to another property in their community or find new properties. Property managers quickly decide whether renters stay or leave to reduce losses through vacancies, forecast occupancy, and minimize manual teamwork.
🤔 This "property management as a service" space seems to be a theme lately. It's a neat vertical-saas, and Renew has executed it perfectly. The consistent theme is founding teams from property backgrounds attacking a slice of inefficiency in the market. The founder market fit for Renew is strong, and this is a well-crafted, market-ready solution with existing partners. Their FAQ pages are goals.
3. Bunce - Payment Operations as a Service (Nigeria)
Bunce helps subscription, Fintech, marketplace, and creator businesses manage payment operations through a single dashboard. It provides failed payment retries, loan recovery, partial payments, invoicing, reconciliation, automation, and reporting. Users can connect to tools like Flutterwave, Zoho, Kora Pay, and Mono.
🤔 Payment operations for rapidly growing digital markets like Nigeria make sense. The rise of subscription businesses and digital business models makes the market ripe for this product. Bunce is relatively young but has customers and some incredibly thoughtful content around ROI and customer segmentation.
4. Cyrus - The personal digital security company
Cyrus provides dark web scanning, social media, and email monitoring to protect individuals and small businesses against hacks and account takeovers. Cyrus also sells this service as an API for companies to use in their KYC procedures.
🤔 This is a fairly standard capability for identity fraud providers. Cyrus is early and looks to have several go-to-market plays (direct-to-consumer, small business, and Fintech infrastructure). The world needs more people focussing on identity fraud, so let's hope these guys get a ton of traction.
Things to know 👀
Apple has launched high-yield savings (at 4.15%) in partnership with Goldman. The account has a maximum of $250,000 in deposits (the FDIC-insured limit), and the Rate is slightly higher than the one offered directly via Marcus by Goldman.
🤔 The experience Apple offers is unmatched. Apple can now store a driver's license and payment cards and offer a credit card and savings with zero compromises on experience. Combine this with healthcare and health tracking, and you have a puzzle that's slowly becoming clear. They're building an ecosystem where they are the center of a consumer's digital financial life. Money is just one aspect of that. The more data you have, the more you can connect the dots to build digital experiences. The best place to do that is via the device and with privacy.
🤔 It's not about the deposits for Apple. This is about payments and brand in the short term. Apple is second only to PayPal for branded payment types used online and via mobile. By gradually becoming the dashboard for a consumer's financial life, Apple creams off its cut of all commerce. The premium experience it offers consumer is a price worth paying.
🤔 Why a better rate than Goldman? Jason Mikula speculated that this will have been a negotiation tactic by Apple to ensure they have a marketable product. Apple absolutely has leverage over its financial services providers. But as we've seen Goldman's card business report massive losses historically, could Apple's focus on experience and price come at a cost for its partner Goldman? The instant sign-up immediately makes me think there has to be fraud risk lurking here.
🤔 Brand and Trust used to be the advantage banks had; now it's Apple's. In his shareholder letter, Jamie Dimon called out Apple by name as one of their biggest competitors. In a world of deposit flight and bank collapses, trust matters most. That's why Goldman is willing to pay over the odds.
Twitter will partner with online and mobile stock & crypto brokerage eToro to offer its users a similar service. This builds on its partnership to share market data through the cashtags feature. Elon has stated that he wants Twitter to get into debit, credit, loans, and checking services over time.
🤔 Why not Yahoo or Bloomberg? With all due respect to eToro, my first thought was, why them? They already had the partnership and could trade stocks, so it's an easy move, but Elon is a noted fan of vertical integration (owning the stack). Could there be M&A in the future? Or is this short-term opportunism to capitalize on the fact Crypto and Stock Twitter is a great place for breaking context and market data?
🤔 What's the right entry point for Twitter into Fintech? There are obvious starting points like creator payments and to copy+paste TikTok and Meta. But as we see with Apple, the keen to own finance is creating the best UX ecosystem and starting to own identity. Twitter is a long way from that today.
🤔 Super Apps and "The X app vision" in Elon's head? Twitter has a similar customer demographic to Tesla. The middle-class, tech-leaning Twitter-chattering elite probably have a good overlap with Tesla owners on a Venn diagram. Tesla has already started to own more of its car financing stack (although that's quite standard in the car industry). But with Starlink, energy production, and transport, Elon is going after physical infrastructure and turning it into software. Finance would bring all of that together, and remember, the original name for PayPal was x.com. This is some wild-ass 100-year vision; when it's complete, it won't look as we imagine today.
🤔 What is a Super App anyway? When we hear "Super App" we think about WeChat or Ali where insurance, healthcare, and financial services all work in one app and ecosystem. That worked for the Chinese context, but it doesn't copy+paste into the West. We have too much legacy infrastructure, regulation, and incumbency. Apple has the best shot with their wallet and identity today. Google is nowhere; Meta is hopeless in the West (sorry, but it's true). Elon has a completely different set of assets to try and build an ecosystem with. Interesting times.
Quick hits
🥊 UK Fintech Yonder got its series A. 🤔 It feels like UK Fintech is getting its mojo back. We could be in great shape if the government gets the new financial services bill right. We also don’t have enough challenger credit card products here.
Good Reads 📚
AI Hallucination is a feature, not a bug. You need to read this incredible piece by Ben Thompson in full. The mere fact that large language models (LLMs) can get something wrong is a suggestion they are creating. A working definition of intelligence is the ability to make predictions, even if they are wrong. Ben takes on a tour from the creation of "logic" as a concept from Aristotle to George Boole and contrasts this with how humans understand and interpret the world. It's eerily similar to LLMs, which are scary because they're like us.
🤔 LLMs are the labradors of AI. They're eager to please humans. Because human feedback loops train them, they're trying to predict what we want, not what is true about the world.
🤔 There are other AI models. How LLMs work with other models and data sources will become a crucial area of research and development. Hugging Face is starting to use LLMs to figure out what other models it should play with, but expect to see more here.
🤔 There are cases where 95% isn't good enough, but many more where it is meaningfully better. No matter how well you prompt, you can outwit an LLM meaningfully in your chosen specialist subject. But what about all subjects in existence where you're not an expert? LLMs are not only a default time saver when used well for admin. They're a default 75% to 95% pass mark in almost any subject. (Like 83% as a brain surgeon)
🤔 What does this mean for Fintech? 1. Stop banning this tool. 2. Use it for what it is great at.
I have lost count of how many banks announced bans on ChatGPT, and I get that urge on some level. Banks also banned social media in the early days, and corporate laptops are still a nightmare to use. Microsoft will no doubt package ChatGPT for corporates (beyond Office 365, which it is already piloting), but a default ban is not the right posture.
The compliance research assistant is a use case I hear of once or twice weekly. There is significant manual work in due diligence companies or individuals. Quickly summarizing that at a level better than most humans can is a cost saver and likely an upgrade to performance.
Tweets of the week 🕊
That's all, folks. 👋
Remember, if you're enjoying this content, please do tell all your fintech friends to check it out and hit the subscribe button :)
Disclosures: (1) All content and views expressed here are the authors' personal opinions and do not reflect the views of any of their employers or employees. (2) All companies or assets mentioned by the author in which the author has a personal and/or financial interest are denoted with a * (3) Any companies mentioned in Rants are top of mind and used for illustrative purposes only. (4) I'm not an expert at everything you read here. Some of it is me thinking out loud and learning as I go; please don't take it as gospel—strong opinions, weakly held.