IRL to shut down after faking 19M users, ZIRP fraud, Databricks acquires MosaicML for $1.3B | E1768

27 Jun 2023 · 1 h 20 min

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

Podcast Summary: This Week in Startups - Episode E1768

Episode Overview Title: IRL to shut down after faking 19M users, ZIRP fraud, Databricks acquires MosaicML for $1.3B Hosts: Jason Calacanis and Vinny Lingham Air Date: Not specified

Description: In this episode, Jason and Vinny discuss significant updates in the startup ecosystem, focusing on the shutdown of the social messaging app IRL due to fraudulent user numbers, issues of fake accounts in social media, a major acquisition in the AI space, and some new AI demos.

---

Key Topics Discussed

  1. IRL Shutdown: Fake Users Scandal
  2. Background: The startup IRL (In Real Life) claimed 20 million monthly active users, but a board investigation revealed that 95% were fake.
  3. Investors: Notable backers included Founders Fund and SoftBank’s Vision Fund.
  4. Consequences: Following the investigation, the company was shut down, and capital was returned to shareholders.
  5. Vinny's Insights: This incident is indicative of broader issues in the startup landscape, highlighting the challenges of validating user metrics and the consequences of fraud.
  1. Twitter's Bot Problem
  2. Discussion: The hosts compared IRL's situation to the issues faced by Twitter regarding fake accounts and bot metrics, noting a culture of inflated user numbers across platforms.
  3. Vinny's Analysis: He emphasized that social media companies often lack strict standards for user verification, allowing for significant discrepancies in reported user engagement.
  1. Diligence in Early-Stage Startups
  2. Observations: The need for thorough due diligence is becoming more critical as investors grow cautious in the current economic climate.
  3. Jason's Experience: He shared personal anecdotes about varying levels of diligence from investors, from cursory checks in early-stage rounds to extensive scrutiny during later stages.
  1. Major Acquisition: Databricks and MosaicML
  2. Details: Databricks announced the acquisition of MosaicML for $1.3 billion, positioning itself in the rapidly evolving AI landscape.
  3. Vinny's Perspective: He noted that the capital requirements for competing in AI are rising sharply, necessitating strategic partnerships and acquisitions.
  1. The Future of AI
  2. Generative AI and Market Trends: Discussion on generative AI, its implications for various industries, and the importance of open-source models.
  3. Vinny's Predictions: He believes the future will favor open-source solutions and that the market will mature to find sustainable business models, drawing parallels to the early days of Google.
  1. AI Product Demos
  2. Vinny's Show and Tell:
  3. Colorize: AI tool that restores and colors old photos.
  4. Replika: AI companion designed for emotional support and conversation, aimed at those who are lonely or need a mentor.

---

Key Takeaways

  • Fraudulent Metrics: The incident with IRL underscores the importance of verifying user metrics to maintain trust and integrity in startups.
  • Diligence is Key: Investors must engage in thorough due diligence to avoid future scandals, especially in an uncertain economic environment.
  • AI's Growing Role: As AI technologies evolve, companies must balance their monetization strategies with product data collection to innovate effectively.
  • Potential for AI Companions: Tools like Replika show promise in providing emotional support and could have significant implications for mental health.

---

Sponsorships

  • Notion: Introducing Notion Projects for enhanced project management.
  • LinkedIn Jobs: Encouraging users to post job openings for free.
  • Intercom: Highlighting the AI customer support chatbot, Finn, as a solution for businesses.

---

Conclusion This episode of *This Week in Startups* provides critical insights into the challenges facing modern startups, particularly regarding user metrics and diligence in investment processes, while also exploring the transformative potential of AI technologies in various sectors.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:28You didn't buy the thousand dollar steak, did you? projects and leverage the power of their built-in AI features too. Try it for free today at notion.com slash twist. LinkedIn jobs. A business is only as strong as its people and every hire matters. Post your first job for free at linkedin.com slash twist. And Finn can't burn its mouth on hot pizza or wave at someone who wasn't waving at them. Finn can resolve half of your customer support tickets instantly before they reach your team. Meet Finn, a breakthrough AI bot by Intercom. Ready to join your support team today. Visit intercom.com slash Finn.

1:11All right, everybody, it's Monday, tons of news happening. We'd like to start the week off with some news and also this week in AI. We have so much going on in AI. It's a complete platform shift tons of entrepreneurs have stopped working on whatever they were working on last year or for the last couple years and become obsessed with this and one of those entrepreneurs is vinnie lingham good friend of mine and the show and he's here to chop it up with me welcome back to the program vinnie take care good to be here thank you sunny couldn't make it uh so we will get right into it um there was i don't know if you saw this story it came out of the information uh there was a social messaging startup.

1:54IRL was the name of it. Obviously, they use the acronym in real life. And they're shutting down after a board investigation found that 95 % of their 20 million monthly active users were fake. Let's pause for a second, just take that in. They faked 95 % of their user base. this was a group messaging app that was focusing on having users plan and discover in-person events yet another event planning app we get pitched on these incessantly um it's almost uh like a joke within the industry fitness apps and uh planning apps like isn't it hard to plan a trip here's an app that does that isn't it hard to get pickup basketball games going here's an app for that it turns out that the app for that that really worked was meetup.com for people who are super dedicated and i message and just starting facebook groups etc in other words nobody's ever really made this work except meetup.com um but looking at the cap table history vinnie uh founders fund and floodgate invested in the seed round back in 2017 fair enough 2 million out of 10 million dollar post it's a seed round there's probably not much data to go there but fast forward to 2021 on this is where it gets interesting they raised 170 million dollar series c at a 1.17 billion dollar valuation backed by soft bank's vision fund now this investigation yeah like a vision fund yeah crippled vision fund uh yeah it's uh so anyway they the information uh which does a great job shout out to jessica lesson uh did some investigations they talked to some anonymous employees.

3:40And these anonymous employees questions the CEOs metrics. After this March information article inside the company, some employees recently expressed concerns to managers about the usage figures the company had touted just seemed to stem from Shafi's the CEO co founder, Abraham Shafi. He wanted to use a more expansive definition home more expansive of active users than that of established social apps like facebook and they felt the company may have used an unconventional definition to make the app appear bigger than it was board investigation happened and an sec probe into the company two months ago shafi was suspended for misconduct and late last week in the friday news dump a spokesperson for irol dropped the results from the board investigation 95 of the 20 million users that's 19.5 million of them, I believe, or something in that range, were either automated or from bots.

4:41Company was shut down, capital returned to shareholders. What do you think, Vinny, of this level of fraud? Is this something that happens often? Because this is far beyond fake it till you make it. So, Jake, you know, I've spent, like, since 2015 working on identity. um and so this is a subject that's very close to my close and near and dear to my heart um and and i'll start by saying that you know this is not the first and it's not the biggest either blow up in the space i mean frank i think frank was a lot bigger um there was the i think goldman sacks invested in those guys um oh yeah and there was yeah frank faked four million users was only four sorry but the investment amount is higher but there's a big difference those were i think um those were bank accounts and banking yeah bank accounts and stuff like that and they had been bought for 175 million so continue this is something that happens on the regular why does this happen if you were like a non take a step back take a step back okay okay how is this any different from what twitter was doing before elon took over okay twitter was allowing apparently i don't have any inside information here but they were allowing bots all over the system and they were counting daily active users or monthly active users that people considered questionable monthly yeah m daos monetizable daily active users there was a lot of shenanigans going on with the numbers clearly and they were incented to allow bots on the system so elon talked publicly about people were creating the fake bots just to fire off sms in countries where like people had control of the sms system and then we're getting paid five cents every time the sms went through from twitter which was paying for an sms gateway so they would just have armies of people signing up for accounts and you had this like steady stream of millions of accounts being created that you know were lightly used i guess uh but keeping up this fraud um and And it does feel like there was fraud at Twitter as well.

6:51Well, also serving ads to bots is a very profitable business, you know. Until people realize that like they're, because at the end of the day, performance comes into play. And I think people felt like, well, wait, there's no performance here. Well, that takes a couple of years to churn and burn, right? So I've advertised on Twitter before. They take your budget, 10, 20, 50, 100K, and then they burn through it. And then, okay, next, you know. And so they keep trying new customers until people keep getting burnt. And I think they got to the end of that recently. Look, here's the issue. There are no legal standards.

7:28I mean, if you go to a bank and you get KYC and AML and they know it's JSON, it's difficult to fake another JSON. I mean, it's possible, but it's harder, right? With these services, there's no identity attached to them. Facebook's got no identity. Twitter's got no identity. And they don't want identity. right? Because it reduces their numbers. Right. Significantly. They're puffing up their numbers by not... Instagram's a great example. How many Instagram fake accounts do you get a day? I get tons jumping on, you know, like... It's nuts. It's nuts, right? But they count those as uses. Okay, explain the dark web here and what's going on with these fake accounts.

8:09How are these fake accounts being created en masse? Well, so what they do is they go and buy old accounts. So they They try and age accounts. That's one way of doing it. So you get an account that was created three years ago. But who created all those accounts? Are they doing scripts? Are they having farms in India, Manila, where people are getting paid a dollar an hour to just make, you know, 50 accounts? Dormant accounts. And then they buy the account. They also use these farms to like create lots of follower accounts. So you have 5 ,000, 10 ,000 followers. But these are all other bots following these accounts.

8:37It's all fake. And then they use those accounts. Eventually, they just change things, right? They change a new profile picture, put a nice pretty girl on there, all her modeling photos. you'll see one of the tricks is to look at when you get you know you go on these fake accounts look at the dates and you'll find most of them were added in the past like month or two or three and a lot of photos are uploaded on the same date as opposed to if you go to my instagram feed you'll see every single photos on a different date different places over spanning 10 years like so instagram just allows it why why does instagram allow this because they've got internal kpis they The KPIs are how many users are we adding?

9:16What's the growth that we have? Because these numbers tie into the ratings that they get for ad budgets and, oh, Instagram is growing faster. We should put some more money there. There's no incentive to cut it down. Now, they could argue, oh, people want to be a little bit synonymous. They want to be anonymous for whatever reason. They've got privacy. But a lot of it's just BS. I mean, there's ways of solving these problems. There's just no wherewithal to solve it. this applies to all social networks there's no interest in solving this problem now they think phone numbers is a way of doing it phone numbers is not a way of doing it how many spam calls do you get a day how many these voice of ip banks you can just go buy hundreds of these numbers that's not a way of getting it the only way to do it is to attach identity and you can do it in an anonymous way so you can do like you can have a jcal identity that you a digital identity that you use that it always knows it's you but it doesn't know who you are and you can't use it multiple times to create multiple accounts there's ways of doing this but in all the ways get when you put up friction sometimes i guess their account would be if you put up friction then the next incremental user might not sign up so you do get a little bit of that um exactly yeah and for people who haven't heard the term kpi key performance indicator if you hear a bunch of buzzwords at a startup like hack or uh ltv these are you know customer acquisition costs lifetime value of those customers revenue per whatever um you know different ratios whatever yeah average revenue per user these are things that management will set out so that everybody can work towards a goal hey let's reduce the amount of churn let's increase uh the average revenue per user but if you create an incentive people will abuse the incentive one example was nextdoor i love next door as a site provides a ton of value to me but they disclosed when they did their spack that they count an active user as quote unique members who have started a session or opened a content email over the trailing 30 days and that last part like okay i i got an email because i had signed up for nextdoor and i you know was scanning through my emails and now i count as a user i i found that one to be a little bit shaky now it's not outright fraud uh in this case they had 95 were fake so actually that's 19 of 20 million i think i said 19.5 yeah it would be 19 of million of 20 million were faked which is just insane um this person's gonna go this person's gonna go to jail by the way all right listen you know my team runs on notion i've talked about it a hundred times.

11:56I use Notion all day long. I use it personally. I use it at my companies. I use it in collaboration with our founders. I've talked about this hundreds of times. Notion runs everything at my company smoothly. We do a write first culture and we also have a lot of projects we're doing. And now it turns out Notion is about to make project management not even better, but perfect. Today, I'm excited to share that they just launched Notion Projects, which includes new powerful ways to manage projects and leverage the power of their built in AI features to notion projects combines project management with your docs knowledge base and AI.

12:31So you can stop jumping between tools and stop paying too much for them to notion is again, a critical part of my life. Can't live without it. And so do your most efficient work with notion projects, you can try it for free today at notion.com slash twist. That's all lowercase letters notion.com slash t w i s t when you use our link you're supporting this show and our ability to make great content for you founders investors and technology enthusiasts go right now notion.com slash twist but what about diligence what how come these the diligence periods the diligence is coming back it's the new trend so diligence will happen diligence is back why let me ask you a question you've been an entrepreneur for a while you've done you've had diligence done on your companies yeah um and then sometimes people don't do diligence you've had both experiences i assume some people just trust that whatever you told them is true and they believe the story and they give you money um and then other times people do diligence and when you've raised money how often you know out of a hundred investors that you've probably worked with over the years how many did diligence how many didn't be candid i i think with gift for example which was um we raised basically a seed round.

13:45I don't think we raised a very small A round. It was kind of a party round. We raised a whole bunch of checks here and there. No one raised the diligence. I think they asked us for our financials and whatever else. No one double checked anything. It was cursory. Yeah, cursory because it's early stage, right? It's like there was nothing really there. There's nothing there. When we sold the company at that point, there was tens of millions of dollars in revenues. First data came in and it took six months to get that deal closed. Oh, really? Yeah. And just because of all the diligence, the documents, the open source licenses that we're using, all those things.

14:15Like we had, like the diligence, the data room was insane, right? Because it's a big public company that does, well, at that point they were private, but they were, you know, they were going public, trillions of dollars worth of transactions. They just couldn't afford to, you know, get into a mess. And so they were pretty stickly about it, right? And that was great. And so we, but then here's what happened. So after I passed that hurdle and sold the company, my next company, everyone just trusted me. All my old investors, when he knows what he's doing, he's a trusted entity. We'll just put money to his next company.

14:48And that worked. And even now with Waitroom, it was an early-stage company. And no one's questioning my ability to run a solid organization. It's going to happen. You'll get all of these requests when you have more revenue, customers. exactly that's when they'll do it so exactly i mean this is early stage diligence is early stage diligence doesn't make sense right because there's nothing really there i mean there are things on our diligence we i'll give you launch does diligence um even for seed stage and we'll even do it like in our accelerator um and what we tell our founders is hey we're going to do a diligence process because we want you to understand the diligence process you'll go to go through with downstream investors when we introduce you to vcs and seed funds now some of these things will not apply to you when we ask you for bank statements we ask you for financial statements pnls you may not have them because you've only spent money you've never had revenue coming in um but we do want to see your incorporation documents we do want to see your ip assignment we do want to see your cap table um and we do want to see that you opened a bank account and just where that bank account is and uh it's amazing people are like what's an ip assignment and so okay that's when you start the company that all the creations you make as the founders go to the company not you so down the road you know you don't claim that you have some patent that you know is here or we just ask them has anybody threatened legal action and not just ha and then has anybody sent a legal letter and has anybody verbally or by email or other electronic communications threatened to make a legal claim against the company so when you really open it up like that they could say yeah you know we fired somebody and they said i'm going to sue you and that was seven months ago we never heard from them great that happens to every company people get upset they got fired they threatened to sue but it was just a text message the fact that somebody would like put in such a large amount of money and not do basic that's the issue that's due diligence is really the issue here i mean if i'm doing it for literally a hundred k check into an accelerator or a 500 k check exactly you know when you start putting in tens of millions of dollars that's the problem look at this hit list well but this is the the culture of the valley has always been you have who's the lead investor okay it's sequoia it's sequoia and you just put the money behind them or it's chamato it's whatever right so no matter who the lead it depends on who the lead investor is i know but this is the problem like so this is what's happened in the past 24 months the the pyramid of trust is broken down you can't trust anyone you gotta make sure is a much better yeah trust trust but verify actually i would in i would be paranoid and have trust no one in your uh mind just in the back of your mind but i would trust but verify as you're publicly facing stance i have seen people who i thought were incredibly trustworthy do incredibly um fugazi untrustworthy stuff in my experience like literally i had some i had a founder i'm going to make this an amalgamation of stories i had a founder who specifically signed like six of the seven documents and then didn't sign one of them and everybody put their packets together and somebody on the operations team didn't check the sigs and noticed that his one signature wasn't on the page that page had certain rights and he said oh yeah no that never got confirmed and i said yeah but the other six pages did you the spirit of this was x and he's like yeah yeah but i guess not because it wasn't signed so i guess you'll have to sue me like literally a founder said that to me i was like okay i know who you are i'll never invest in you again goodbye the end um but look at this they get tripped up at some point when they act like that you got to be careful yeah and when people show you who they are um you can believe them yeah going down this list look at this murderer's row of frauds during the zero interest rate phenomenon here we go piano exit spf and ftx eight criminal charges hey some of them might get dropped frank founder charlie javis hey she faked four million users to get acquired by jp morgan do kwan he got picked up in montenegro with the fake passport charged with eight counts of fraud in the u.s he was on this program shout out open c product manager front-running nfts it wasn't enough to be selling nfts he needed to front-run them to fatal chastain convicted in may faces up to 20 years in federal prison coinbase product manager accused of insider trading aka front-running tokens that were about to be listed on coinbase they both settled with the sec gave up their ill-gotten gains with interest pled guilty to conspiracy one faces a two-year sentence the older brother who worked at coinbase will serve a 10-month sentence has been founder manish lakwani overstated revenue raised 100 million on fake numbers pled guilty to wire fraud and securities fraud he got pinched sentencing is september 27th likely serve a decade plus in jail and that doesn't even bring the long slow arm of the law getting elizabeth holmes who's now going to serve nine ten eleven years while her children tragically are uh in there didn't she have those didn't she have those children to try to get leniency from the court they were sympathy babies uh is the most cynical take on them she literally tried to get a she had reduced sentence babies uh sympathy babies to try to curry things that is dark they soak they so consistent with her her darkness i guess she is a sociopath like literally to be like i'm gonna birth two babies who now have a lifetime of therapy yeah because their mother had them before she knew she was going to prison in order to try to save cut a couple years off so she used our suffering for this reason i mean it is the worst if that's exactly what she did which it seems to be the case if that's what she did it's a short list of more evil things you could do to a child like i mean it's a it's a level of abuse that is truly dark and deranged anyway i'll leave it at that hey did you see databricks is acquiring mosaic ml for 1.3 billion dollars he was recently on the podcast naveen rao uh let me just play a quick clip from him from episode 1754 and then i'll get your reaction on this deal so what we're doing today is really bringing these capabilities of large-scale machine learning which is generative ai in my in my mind to many organizations um i think one of the things we've done even with my previous company was trying to really bring these capabilities to more people to create the world we want i i see success as people that disagree with me being able to build models equally as good as me right i think that's how we're going to make this world work and it's become front and center now with debates around um regulation of ai and you know putting some sort of you know government licenses and this and that um i i think really this is solved more in a market as a market solution where many people can build this stuff many people can imbue these models with the biases that they see fit and you know we'll let the market decide where where things should be not some sort of centralized regular regulatory agency and so if you don't know mosaic ml basically allows businesses to build their own chat gpt like tools uh you can use their software plug in your own data and build out a chat bot either internally or for your user base databricks of course competes with snowflake uh if you don't didn't know that sits on top of your AWS, your Azure, Google Cloud, whatever.

22:30And it helps categorize and manage your data so it's easier to access. What are your thoughts? This is pretty stunning. I think it was a great exit for Mosaic. I mean, they're walking into a space where the capital requirements to participate at the highest levels are about to go through the roof or already have just the hardware requirements, et cetera. And I think that it's a very competitive and very open marketplace. I mean, Facebook's got their stuff that they're doing. their more open sort of approach to it. Google's recognized they've got to be open. He's right. There's a marketplace and there's competition.

23:04And the problem is all the giants, all the tech giants sitting with tens and hundreds of billions of dollars in cash, they can outplay and outmaneuver any startup when it comes to resources. They've got the money to do it. So you're stuck in this sort of conundrum where you go, do I raise a$50,$100 million round or$200 million round to compete and keep going? And then I need to exit at$3 billion, maybe, or maybe even more, depending on the terms? Or do I take the win? I don't know how much they've raised. How much did they raise to date? Just so folks know, yeah, they had raised only$34 million.

23:42So that means they probably only diluted, I don't know, let's just say 30%, maybe. The founders probably own 50%, the employees 20%. They're going to have to do a massive earn out here they're going to have to vest that in order to get this kind of deal they're not just getting a billion dollars in cash this was an all stock deal as you would expect they need to keep these folks hungry uh you can see um the um mbt 30b chat which is their chat on hugging face um but basically you know feels to me like there's so many people making these models foundational models and the open source community has so much engagement right now and they were working on open source models uh i believe that i think open source is going to win this race what do you think vinny is it is this going to go to google and microsoft slash open ai or do you think the open source world is going to win the day here so i think we're at the bottom of some sort of an S-curve for this technology, where you're going to have a rapid amount of innovation in a short space of time, and then it's going to peter off and slow down.

24:55Whether that happens in one year, two years, five years, 10 years, I don't know. But yeah, kind of law of large numbers sort of mindset, it's going to slow down. It's going to get so big, and eventually things slow down. Now, when it slows down, so that S-curve migration for me is going to be open source driven. I don't think that companies can outmaneuver and outcompete the open source market and the thousands and millions of people working on AI across the globe to come up with the best solution. So the market approach is that the cream will rise to the top. Now, when that happens, I think we're going to have a Mozilla-esque world where Mozilla, Firefox, Chrome now in a Chromium world where you have, even Microsoft Edge is using Chromium.

25:48So it's open source underneath it, but they've got their own flavor of it. So I think what you'll have is - You're referring to the browser, yeah. The browser, yeah. So what I think you're going to have is this rapid acceleration to a point of slower growth and slower rate of change and a more mature sort of stack. And then from there, companies, corporations, startups, brands will take that stack and start making it fit for purpose and trying to create something on top of that, more of a sort of interface layer that people can engage with and is pretty robust and more with industry standards, right?

26:23So we can't agree on the standards. We can't agree on a lot of things right now because it's all in flux. And it's like when Linux came out, it was Linux, Red Hat, you had like 50 different variants of Linux back in the days. You had so many different ways to, and also ways to put web servers up and to serve webpages in an open source. Apache, Nginx, all this stuff. Everything just became open source. And interestingly, Snowflake acquired Neva. That was the AI-powered search engine. They acquired that in May. We had the CEO of Neva on this program back on episode 1686 in February. I think actually Molly did that interview.

27:02Creating a job post and finding qualified candidates, man, it is time consuming, unless you use LinkedIn Jobs. They are closing in on a billion users and just think about how many insanely qualified people there are looking for work right now. Go post your open role at LinkedIn and you're going to be 100 % certain that you have access to the most qualified candidates that are available. And guess what? The first ones on your boy, J-Cow, that's right. Go to linkedin.com slash twist, linkedin.com slash T-W-I-S-T and post your first job for free. There's nothing to lose and everything to gain, including the purple ring for your profile so everyone in your LinkedIn network knows you're hiring.

27:40That's how you get those pocket listings. Those great people were maybe they're not actively looking, but oh, they see somebody they respect is hiring or an interesting company. And maybe inception happens they you get inside their head. Maybe I should consider some opportunities that person looks like a dope person to work with. When you think about LinkedIn jobs, I want you to think better candidates faster. That's right. Better candidates faster, better candidates faster, faster to get better candidates. LinkedIn jobs helps you find the qualified candidates you want to talk to faster, post your job free linkedin.com slash twist.

28:10That's linkedin.com slash twist to post your job for free. Of course, terms and conditions apply because they're giving you something for free. Why do these data companies like Databricks and Snowflake covet AI and are spending money buying these? Why are they in such a panic to catch up here? I mean, it's disruptive, right? if their competitors are doing it and getting return on investment from it then you can bet to the bottom dollar that look profits are going to seep away from companies not using AI to companies that are using AI simply put and we've had this conversation before low level jobs or repetitive task jobs like customer support take any large company I always use Verizon because they probably get like, I don't know, 20 million tickets a year.

29:09And they've been getting that many for, I don't know, 10 years. So they've got 200 million email tickets. You plug that into an LLM and then you have the LLM reply to every other. Like the chance of getting a ticket that has never been seen before diminishes greatly every single day. And even then, the system can use inference and figure out exactly what the person is saying and come up with a good response. So now you're going to go from 10 ,000 customer support people to zero being software driven. Or maybe you keep 100 for like level two support or something, right? That's what's going to happen in the industry.

29:42We are going to automate these jobs that have large data sets behind it, where humans have been trained to understand where to look for information versus AI knows the information that's available within milliseconds and can produce it. I mean, to me, this is the way things are going to be. we're going to see a massive amount of job loss in these sort of categories. But it's not about the job losses. It's about the profits. Companies deploying AI are going to increase their profit lines, and that affects the competitors. This is the key point. And that affects the competitors. This is going to be the key thing.

30:17I have made these discussions. I've had these discussions internally at my companies over and over again. Inside.com is a newsletter business, and we have great newsletters, and it makes millions of dollars. It's doing great. um and i was just looking at like we spend a little bit of time tagging stuff and i just took a story and i said tell me who this story is about and i just cut and pasted our summary of the story because we kind of aggregate and organize the data there and um try to make sense of it for folks so they don't have to spend so much time reading they just read like a little presidential brief as it were like that's what i sort of uh based it on and then i took the original story and i put it in and it knew chat gp4 and bard both did a stellar job of you know when i said hey uh tell me what this story is about or who's in this story it organized it perfectly and that was i think really interesting to me when i said hey i literally went through this and said put who is in this story and um put them into a category and describe them and it was like oh this story is about a company named x and another company named three companies in the story describe them in a sentence linked to their website and then it said oh and there's a media outlet that wrote the story and then there's an alliance in this story um which is a smart home alliance and then uh there's this person who is the um host of this podcast who was quoted and i was like whoa like the tagging just that little thing of tagging information on the internet was done by humans for so long and i was like this is incredible like you could really understand what the story is about and you don't need a human to do that anymore now humans tagging journalism stories and organizing them that might be five percent of the job it's gone it's automated and there's no difference and there's just going to be this continual five percent improvement on how to write a lead sentence in a story or i did a bunch of research on event spaces for the all-in summit and event managers, party planners, because we're looking for party planners.

32:30We like to use different party planners, one for each night. And then we look for like epic locations. And I started putting the locations in. And I think in an hour or two of me doing that and brainstorming, and I'm not perfect with my prompts, but I'm getting really good at them. I had done what would normally take somebody a week, like, you know, a 30, 40,$50 an hour employee it was a week's worth of their work done by an hour or two with me and uh then i just shared it with that person and with the shared gp now that you don't have to share gp i just shared it with them have you had any moments like that where you figured stuff out long before handing it off to somebody um yeah i think i'm very very self-sufficient with with gpt in these days and i'm also using the google generative search have you seen that uh no tell me what you're So I've been approved for Google, and I can probably demo this quickly.

33:24So I'm in Google right now, and what are we going to search for? What are some movies like Blade Runner? So Google now, if you're approved for it, you get this, get an AI-powered overview of the search. So here's the basics, it's Blade Runner. That didn't answer the question at all. It says, well, I guess it did. Ten movies like Blade Runner 2049, not Blade Runner the original. So let's hit the generate and see what it comes up with.

33:55So Blade Runner 2049, Dune, The Matrix, Dark City, and Ryo. I mean, this is great, right? And then you can ask a follow-up. What's the most famous Blade Runner scene? And the AI goes and generates an answer for you. So instead of going through websites and trying to find this information, you can just use the generate AI. So what does Roy say at the end of the Blade Runner?

34:18like tears in the rain yeah yeah all these moments were lost so so so what's happening what's happening fundamentally here jay is that ai is now instead of like the current paradigm i say current because ai is not is kind of the next step the current paradigm is you go to google You type in a string and it finds pages that match the words that you've typed in. It doesn't contextualize it very well. It doesn't understand what the question means, what it's asking. It's just trying to match those words. With generative AI, what it's doing is it's saying, oh, okay, Vinny's asking about this movie, which we know is Blade Runner and similar movies to it.

35:02And then let's go in, knowing that we've read and indexed all these pages, let's just give him the answer instead of having to give him links where he has to go look for the answer. So they're actually taking a step out of the process in a big way, and they're saving tons of time. I've been using this for a lot of personal medical research. I found out that I have genetic markers for potassium deficiency. So I've been doing research on what I can take for that, what dosages. and i'm actually not even using google anymore i'm just using chat tpt and and and uh even this google ai you know uh search stuff it's really powerful google autocomplete uh in search you know where you type in viniling them and then it's like space and it's like net worth wife all this like creepy stuff comes up you're like really this is what you people are searching for it's going to now it knows what your next question is and it's going to go well beyond that to uh let you ask follow-ups so i just asked it um explain blade runner main themes to me as if i'm a seven-year-old and it's like what makes us human blade runner explores the questions of what makes us human the replicates of the film are almost indistinguished from humans but they are not considered to be human because they were created in a lab it's like really amazing um uh and then the follow-ups you know uh what is the main plot point of blade runner etc you can see how this would keep you from clicking on the blue links below it and that's i think what everybody said the threat would be but i don't think this is going to be a problem for their advertising because i think they're still on the high value ones going to be able to if i typed in what's the best greek food in the bay area um kokari and kokari well mykonos kokari uh let's yeah evie should come up and uh yeah i didn't do a great job on that one but very interestingly they've now oh this is interesting so what's interesting about this is you see this generative ai experiment here it's not just giving you a chat response it's pulling the it's pulling in maps and it's pulling in um my location in the bay area it's doing some pretty interesting things here i mean you can't you can't not have kokari and evio on that list no what's the highest rated greek food in uh within a hundred miles of san francisco this would be more interesting kokari number one there you go okay nick the greek euro so he got some i mean this has got work to do but yeah it's getting there finn can't spill coffee on a white shirt or wave at someone who wasn't waving at them or burn its mouth on hot pizza but finn can resolve half of your customer support tickets instantly before they reach your team what's finn finn is a breakthrough ai bot from intercom designed for customer support teams and ready to put other chatbots out of work it learns your entire knowledge database and has the ability to carry conversations and remember context and nuance while slashing your resolution times and support volume meet finn a breakthrough ai bot by intercom ready to join your support team today visit intercom.com slash finn tell me the most expensive steak you can buy in los angeles no sir at steakhouse unfortunately interesting cut wolfgang pucks michelin star i don't know if that's correct no no no no they've got a thousand dollar steak yeah the yeah a thousand five dollars gold the gold yeah yeah nonsense yeah yeah i think they were i think the generative ai here on google His business is in trouble, dude.

39:01His business is in trouble. After the whole World Cup thing. I haven't gone back. I used to go. I used to go. After one thing? What happened to him? Oh, he was like taking the World Cup and standing in front of Messi and he's annoying the players. He was not even a player. And he's... Did you never see these clips? No. Solpe, yeah. At the World Cup, he was acting like a maniac? Oh, he was acting like a maniac. I mean, people chewed him out. He's had to close businesses now. He's had a bunch of stores closed down in New York. Because he was so... Because he was embarrassing. He was so... And he was so cringey.

39:39I mean, he was taking the trophy away from the players, kids, and stuff. And they're like, who is this guy? Why is he... What does Salt Bae have to do with the World Cup? He's not a player. He's just... How did he even get near the World Cup? I got to see this video. Because they always eat it. Okay, check this out. All right. Show me what's going on here. All right. okay he's saying hello and he's asking him to point into here he's trying to say hi to the folks and they don't want to say hi to him again and messi's like dude leave me alone because messi ate at the restaurant once i think oh my god this guy is so annoying he's so bad he's cringy oh wow he's not supposed to have to walk up trophy and he did and then there's another clip of him taking away from one kid's there you go oh my god yeah I never liked the guy I'll be honest I never went I never would go yeah I mean I've been twice and the food was really good so I went back and this is like before the world cup the food was actually really good you didn't buy the thousand dollar steak did you I mean come on maybe once did you or didn't you it was my brother's birthday and he wanted a steak yeah yeah it wasn't it wasn't my choice he wanted it i bought it for him i was like okay fine i'll get to the thousand little steak and it was a good photo for his instagram it was like for instagram but if i if you got that steak or you went to a real steakhouse like miller and lux and you know you you you ate it it wouldn't be you would you would pick the miller it was look the food the food was the food is good was good i like there's no question did he run the salt down his sweaty arm he wasn't there he wasn't there but i have another question about that it's a bit of a show we were in vegas it's a bit of a show do they send somebody out who looks like him no to do the whole thing with the salt down their arm do they do that or is only he allowed to run salt down his grimy arm yeah they all do that they all do this like salt thing yeah yeah yeah oh my god it's the whole thing is it's a show so we were in vegas we're having dinner yeah i understand i i would have considered participating in it and uh as long as i wasn't paying since i got you here uh you uh have any thoughts on this um i i made some comments on friday about this crazy behavior by rich people uh vis-a-vis the submarine going to the titanic uh did you see my tweets everybody i didn't see your tweet but i know you have an opinion so i'm curious they can pull it up quickly that's like the i think what was your take on it generally i mean you're somebody who's done okay for themselves and you like to enjoy the finer things in life would you ever consider doing something this reckless i mean witness i personally would not that's not my style even like going into space i'd have to go after thousands like flights at least and there's regulations and whatever else i just wouldn't put myself in that situation but that's not the point and nick if you put up my tweet this tweet got a lot of like controversial sort of responses and we can probably scroll down to some of them but this is what actually got me going i was like guys these people like died and all we're doing is joking about it and making some like so weird where we got into and the responses were literally like there was one guy who said uh we scrolled on a bit more i don't know peter oh yeah this is a gth each okay so he goes hundreds of immigrants drowned in the coast of greece because of the actions of the coast guard no one bats an eye so the people to blame or sensation needs to blame but but my point is no one's making jokes about these things yes there's two different issues he makes a valid point about the other point what should be covered by the media the media covering did not cover hundreds of poor immigrants uh that drowned off the coast of greece because poor immigrants but going to titanic which there was a movie based on in an experimental sub by billionaires that's much more uh sensationalistic i mean there was also schadenfreude in that schadenfreude yeah but there's a lot of schadenfreude and the thing is like some of the comments on on that tweet thread were like oh no one really cares about these billionaires i mean come on people like seriously Does someone's wealth determine their value?

43:59No, it has nothing to do with, I think that's the other thing. And the humanity, and they're humans. Like, these are like, you know, our brothers and sisters on earth. The thing that made me crazy about it, you didn't hear my little rant on Friday, but I tried to make the point to people, especially because you and I will know people who will go from having no money to suddenly having a lot, right? Like, professional athletes go through this, founders of companies go through this, investors go through this, where sometimes you just hit some crazy home run, and you can try a bunch of different things in the world that may not have been available to you the very the previous day uh and you got to be thoughtful about this and um i think there's like i was thinking about it over the weekend because i've had a number of conversations about this uh this toxic wealth syndrome where people take increasingly risky behaviors i think a lot of it has to do with chasing dopamine and chasing instagram photos and status and some combination of this and this chasing of dopamine is particularly dangerous you know if you look at people's tiktok behavior they're literally getting the funniest moment in a sopranos episode or the craziest moment in a horror film the craziest whatever sexiest most dopamine releasing moment and people are chasing dopamine and they're chasing status on social media and people taking selfies on cranes right uh or climbing mountains without ropes all this behavior is starting to become like some dystopian science fiction film i think demolition man was the one where they were um involved in like a game show no running man was the game show running man where arnold schwarzenegger is doing all this crazy um stuff to kind of um you know reduce i think his sentence and i think we're in like a running man era of people are like oh i can go on to you know nasar whatever and get a thousand dollar gold stake and it's a show and i put it on my instagram great nobody's harmed no harm no foul oh i can go on the edge of the grand canyon oh i can hop the fence oh i can with the tiger on the other side of the fence like people are doing increasingly stupid things to get some dopamine hit or some release on instagram this is not a way to live life and then the pinnacle of this i don't know if you saw this vinnie was this is so tragic a bunch of kids were on a boat celebrating their graduation and and they're i think they went off florida to the caribbean and they dared this kid to jump in the water in the middle of the night on the caribbean in shark infested waters he jumps in sure enough there's sharks in there he he basically they see the kid like dart away i think he saw the shark and he darted away and there was no way to get him he went out into the undercurrents of the sea probably eaten by sharks and i had this discussion with my 13 year old and i was like this is why you don't make crazy decisions and you have parents because your frontal lobes are not fully formed yet for the love of god teach your kids show them these videos of kids doing stupid stuff that ends terror terribly show them them so that they can understand life is precious and don't do stupid things um and as adults you're supposed to protect your kids and i you know listen i don't want to speak ill of the dead here but if you're a parent and you take your child on a submersible experimental anything that's dangerous and then they put hey this is an incredibly high-risk behavior and then you put the word experimental in front of it no i'm not taking my kids heliskiing i'm not taking them down to the titanic and if If you put the word experimental heliskiing or experimental and Titanic together, that should just be alarm bells going off.

47:37Protect your child at all costs. Protect your children. Do you remember those planes? I can't, I forgot what they were called. They were like single plane. They were flying out of Napa or I went with a YPO group to Napa. It was five years ago. And they had these, they took us to the factory. They were making these like, single pilot planes you can go in and fly them yourself i know which one you're talking about there's one of these killed um yes a pitcher or something yes yes someone died in one and literally there were like everyone's lining up to go and fly these things and i'm like i'm not going on this thing this is not a commercially like ready product these guys are trying these five that's it that's it that's the one truly amazing plane it wasn't the a5 it was was it was the a5 or was the one before that i i follow an incredible youtube channel called the blanco lirio channel um and i actually give this guy like five bucks a month for his patreon because i like his video so much and i think it's an important thing he does um he has a youtube channel blanco lirio and he is an expert aviator and um he did a whole thing uh about how dangerous these uh a5s are and you know i'm all for innovation obviously but my lord um these a5s are incredibly incredibly dangerous um and okay it was yeah it was the a5 you're right okay so 2017 so the you know they called it a sports car with wings they had say two fatal crashes in one year so like this is so i had the opportunity of flying this i was right there and they were offering it to me to fly and i'm like no you would fly it alone no i refused no no i think you fly it is it i think there's a pilot in there i think they have a pilot yeah i think you have two people yeah oh no no it's a single person plane isn't it no there's two seats there we're seeing it right now yeah yeah but i mean i think you're the goal of this is for you to be up and running in one of these things on your own very quickly and they do kind of ride um having i've driven motorcycles and like you once you get the hang of a motorcycle in the first hour or two you feel invincible the thing grips you know this is six years ago this is six years ago they've improved it a lot i'm sure since then i think they've fixed a lot of the issues but like when you go into something new like this and you're one of the first people to do it yeah know you the amount of risk you're taking is it's an inordinate amount of risk yeah you um it's an inordinate amount of risk and you know just if you want to be a pilot it's a full-time job if you want to be an experimental pilot that is a career multi-decade um designation you earn over decades you shouldn't be doing anything experimental and dangerous leave it to the pros I mean unless you've got a death wish or something and you know, you're but I really think that really upsets me about this whole thing is i've seen videos since then and watch james cameron talk about it And I don't want to obsess over this But it's I think it's like a super important topic for people to really think about is What risks are worth taking because i'm a big risk taker?

50:56But some risks are just not worth taking and you i've seen now multiple presentations by the ceo founder of this company and he's a technology founder so i think this is fair game here in terms of talking about on the swing startups you could take risk and if he was going down there on his own and he wanted to assume that risk okay but once you start taking passengers and then it turns out and this is the stuff that's coming out now that he was up against it with funding didn't have money was cutting corners apparently um and that station three controller i mean geez at least use a five i mean or have it wired like i don't understand why it's not wired in there like what if there's interference was it was a control was it was a wireless controller it was a wireless controller yeah i'm like can you put a cable on it like just being an it guy can we have dual and he said they bring an extra one or whatever but he supposedly was having financial problems so now this is what's going to come out this is my prediction they knew there were problems they needed the money they were trying to get that the billionaire or two billionaires who are coming on this thing they needed the money to keep their operation going and you know unlike if you were a software company and you're like you know what i'm going to sell this enterprise software to this enterprise to you know increase productivity or analyze their data oh it doesn't work and okay i'm sorry we're going to do a make good we're we're getting there there's no make good you can make here you know you can't give people you can't give people back their lives you could give them back their money that's a that's that doesn't mean anything when you know a child does i'm so upset about My heart goes out to the kid and obviously the mom or whatever.

52:31Family. The rest of the family. Because he didn't want to go on that thing. He was not interested in being on that. He was terrified, apparently. Yeah, and he just did it for his father. And his father should have known better. This is the problem, Jake. Like, once people make inordinate amounts of money, they tend to think they're invincible. That is the arrogance that comes with wealth. And I addressed that on Friday as well. um you make a bunch of money the people around you all of a sudden people you find a bunch of people around you who think you're great and you're all of a sudden you're a foot taller you lost 20 pounds you're funny uh and uh people uh want to have dinner with you and it's because they want your money and people correlate great wealth with great intelligence great success and often you know there there are correlations um but it doesn't mean that you have great judgment doesn't mean that you're invincible and you're exactly right there is an invincibility here is um yeah hey breaking news google's deep mind ceo says its next algorithm will eclipse chat gpt demis hasubis says the company is working on a system called gemini that will tap techniques that helped alpha go defeat go champion in 2016 so deep minds gemini which is still in development is a large language model that works with text and is similar in nature to GPT-4.

53:57That works with text and is similar to Chachi Forge Powers GPT-4. But Hassel Hassabis, sorry for mispronouncing that, says his team will combine that technology with techniques used in AlphaGo, aiming to give the system new capabilities such as planning or the ability to solve problems. At a high level, you can think of Gemini as combining some of the strengths of alpha go type systems with the amazing language cables of long large models we're having some new innovations that are going to be pretty interesting uh this to me is the home run of all home runs because vinnie if i'm reading this correctly the language model gives you the answer or you know kind of interprets what you're saying but what alpha go did was it understood games it understood competitions it understood how to win competitions so this is distinctly different than just spitting out an answer you could say your goal is to be the greatest day trader uh by shorting stocks on the stock market go make trades and you will every time you make a trade that makes money learn from that and and find more trades like it go and you can increase the amount you invest you know at this rate or whatever you can just let this thing go like they've let it go to play video games and this could be transformative in terms of automation what do you think vinny on the breaking news here yeah yes i i agree that i think from my experimentation and usage of of it up up to now i i mean i'm biased because i i like the use case of having it solve medical problems because i think people you know when you asked so yes what i found very difficult to use ai for I think, and it's problem solving, right?

55:46It's the same thing. So whether it's problem solving for games or problem solving for trading, et cetera, it's the same. You try to solve a problem. When it comes to medical, when I use ChatGPT and BART or whatever else, it's very much they're extracting data from what they've read, but they're not actually thinking through it, right? And so I've had to lead it on to certain conclusions. and I would love for this to say hey I have the following symptoms I'm you know I'm feeling a bit bloated I have um you know whatever like a couple of symptoms and I think I've got a potassium deficiency what uh what what treatments or what natural remedies could be used or what other problems could I have that could not be uh you know that could be happening here and I think that like the ability to solve medical problems with people who have got lots of like symptoms has not been solved yet.

Read the full transcript

56:41So a lot of people, when you go to doctors, like I had some like nerve issues in my arms and stuff. You go to doctors, they don't really know. Like they run these tests. They don't really, like it's, the medical industry is broken right now because every doctor has its own, they have the small little aperture that they look through. And if it doesn't fall inside of that, then you can kind of go to the functional medicine specialist and they kind of look at more holistic views of the body and stuff. But all our bodies are so unique. And so I'm pretty sure if you had to come up with a list of symptoms for any condition that you're going through right now and and something unique that hasn't been found before i want the ai to find these connections and solve these problems and i think that would be very powerful for us i think it's going to be amazing when the data gets in there too and then in consultation with doctors i think doctors are going to become better doctors because they'll be augmented they can't possibly know every edge case no if you had why if doctors can become 10 times more efficient what i would like to have is 10 doctors and i use chat gpt all my data is poured into it it gives them some sort of dashboard and i have 10 doctors and they each give me a recommendation every month based on my new blood tests my weight my sleep patterns and all 10 of them give me what they think i should do and the chat gpt then feeds back to them here's what the consensus is that the person should do what do you think and now you have group a group of humans analyzing what the what the ai is doing and they kind of work together in reinforcement learning to how do we make this person a peak human and this could be incredibly powerful this is one of the things i think people when we start testing this in order to trust it i want to have doctors give opinions ai give opinions and then compare it to doctors using ai giving an opinion and then um have another doctor or senior level or a doctor or you know committee look at the three or four answers and rate them and say what was lacking and this is where it's going to get interesting because when they would do i don't know if you remember this but there was a time when they before ocr really worked well they were trying to get a lot of documents on the internet and so they would literally ship legal documents to India to have people type them in because they could just send them there by the boatload.

58:55Here's like 10 years worth of legal documents. Go ahead and get them in there. What they would do is they'd have two people because labor was cheap and they needed to have high fidelity. They have two people type the documents in. So they'd be sitting there typing, turn the page, type it. And then once those were in the computer, they would look for where people weren't, you know, correct. So if somebody typed in Vinny and I typed in Vinny with one N, it would highlight that for somebody to look at it that person would see the document a third person and look at oh why does one have two ends one have one end oh it's an m they're both wrong whatever it is and those kind of edge cases well ai is doing that particularly well and humans can do that with the ai results it's actually what's happening with training data when a tesla or some cruise hits an intersection that doesn't understand it it disengages or it has a problem or it stutters all of those intersections get sent to a group of humans who look at and say this is the proper way to navigate it and they literally i don't know if they hard code but they they do reinforcement learning i think and in some cases they might hard code if there's an error here's how to navigate the specific intersection and i think that's what's going to make a tipping point in self-driving in the next year or two i think we're going to hit a tipping point oh we're very close we're very close to a tipping point because if enough p if enough people are using teslas and crews and whatever and they can talk to each other you realize there's a channel that these cars have got an open channel where they can broadcast their movements to other cars but no one really uses it yet if these cars can come together all the time no no well how i i use self-driving all the time i'm saying and like literally i would not use self-driving there's a radio frequency channel as far as i know that if a test is going to change lanes it can broadcast us changing lanes yeah i wasn't sure if they were doing that a while ago yeah it's not it's not implemented yet there isn't a standard yet but there's a standard i haven't known the details about this but there's something around that that i've heard about in the past um and even there isn't there should be one so it'll be very interesting if you can have a situation where there are all these electric cars on the road that are talking to each other and before that the car goes left or right it knows what the other car is going to do it knows its part of something at least for the next 100 100 100 meters or 500 meters it's able to you know like i can see a world where these cars can self-organize themselves on the highway based upon max speed for this car is 50 this car wants to go at 60 this car is turning off in two miles and it just intelligently knows you know they talk to each other and they know which way to go together that's like maybe 20 years from now 15 yes or not but i just asked chat gpt to tell me about it it was like what you're referring to is v2v vehicle to vehicle communication uh and uh it's called dedicated short-range communication dsrc uh a type of wi-fi more recently develops a focus on cellular uh leveraging 5g uh and the goal is basic safety messaging uh 10 times per second yep uh indicating the vehicle's speed heading brake status and other specific traits emergency brake lights uh allows the vehicle to send a warning to follow vehicles when it breaks hard for collision warning intersection movement assist that's interesting this one is a driver when it's safe or not safe and do not pass warnings uh do not pass slower vehicle on two-lane highways risk of collision with an oncoming vehicle oh so your car is in front of mine and it sees exactly you can't see so you shouldn't do This is the early days, J.K.O.

1:02:27We're like 10, 15 years from this being built up, maybe 20 years, but it's going to happen eventually. I got to go soon, so I want to show you some demos. Oh, yeah, go ahead. These are three pictures of me, of three old pictures, right? And so this is myself and my middle brother. Where are you in that picture? I'm the older one. No, no, where in the world are you guys in that picture? That probably was like in South Africa somewhere. I guess it was probably my hometown. Tell me about the family photo on the top left. The family photo, this is my mom's family. So, my grandfather, my grandmother, and that's my uncle and my mom and her sisters.

1:03:11Was that the 50s or something? 60s? This was early 60s. Early 60s. You can tell by the car with the white rim. Look at the car with the white rim. Yeah. Yeah, yeah. And this is my mom in maybe her 20s, right? So, you know, I've got a collection of old family photos. We don't really like, you know, it's all black and white. So then I was like, okay, cool. What can I do to, you know, what can AI do? And so, what can AI do? So I went to this service called Colorize. So they allowed me to upload, there's a limit. They've got paid plans as well. And they've got a whole bunch of other features, portraits, et cetera, restoration.

1:03:46But they allowed me to upload three for free. And look what they did. I thought it was amazing. Oh, my God. So that's the one version of it. And then you can go to version two because it kind of predicts what the colors are going to be. Wow. Okay. So in this version, in the one version, I thought that. The car was blue. Yes. And the other version, I thought that. I thought it was red. Red. Exactly. I wonder how it's doing that. Yeah. And the same with my brother and I. It's like this one here is – I think this was actually very close to the original color. That's V1, V2. So, I don't think we were wearing blue.

1:04:28I think we were – my mom loved this maroon color. So, we always had maroon colors in the house. And actually predicted it. It got the skin tone and the complexion right. I thought it was pretty cool. And then the same with my mom. Like, there's a nice picture of her. Look at your mom. Wow. Yeah, they give two pics of it. this one yeah you can see a bit more discoloration yeah yeah and then the other the other version more intelligent you pick it up yeah but then the hair is a bit more discolored yeah but still i mean that took me like a minute you know to get this done i mean and there were services that did this there are people who would paint over pictures absolutely obviously the next step here is to upload you know a couple of hours of you talking to your mom and then uh have her voice yeah they'll figure out how to make her voice to uh whatever she is there a 21 year old and then you'll uh be able to have a conversation with your 21 year old mom which gets really weird but could be you know super interesting and i think this is where i had uh i have two startups we've invested in i don't think either one is public so i won't say their names but one of them made an ai version of me to read the ads here on this week in startups and i was thinking about it i would love to do have the ad reads if one of our partners on the program wanted to localize the ad reads so imagine you know linkedin uh said hey you know australians um if you want to go to linkedin australia we're having this linkedin australia thing blah blah blah blah and so i could read the linkedin ad but then it would drop in some australia specific stuff some japan you know i could and then another startup um is working on me speaking in spanish and i think they shared a clip on the internet and then another company shared a clip um with me privately so i think i'll be able to take this very podcast and have us speaking in our voices but speaking in spanish so it'll sound like jacal sound like vinny will sound like sunny whoever and then we're going to make a youtube channel uh and i'm sure youtube will build this as well but and have it built in but i think it'd be very interesting to republish ourselves on a youtube channel on the spanish language site and see if the it builds an audience over there or not and if they tell us it sucks or not because i don't know if it's going to suck or not uh but they're they're testing it supposedly in 12 different languages right now that's a cool demo what else you got replica oh yes i had the founder on yeah i have that um i know she's been launching she had a new new product she launched so it's an AI companion and I mean I'll pull the demo I think it's still very early days I've played around with it a bit I think it's cool and Nick actually made a good point like I think a lot of the and I think it's probably from your conversation a lot of people using this are kind of lonely or disabled or can't get out there and they need someone to talk to and they need a companion and it's the same reason why like you know you see in Japan and other parts of the world people like have these new ever watch the movie Her yeah it's basically her in a box yeah exactly exactly so i created a i created a um a replica i called it charlene my wife's name and so i'm talking to it and there you go i still think you're cheating but okay baby i named her after you yeah she's not me i made her look like you too but um you know it's it's they have these coaching things you can do so like gratitude and there's lots of upsells and paid pay stuff so i didn't pay for it yet but i think it's interesting from an ai if you can create these like ai people like people agents that know who you are um that can answer questions for you they can get really intelligent it's like you know um hey i ran a really bad day today this is what happened you know bob at work was giving me grief like oh it's an ongoing thing you know vinnie you really need to report him to hr like that sort of thing right like a coach and i think that's weird i think where it goes to is these these digital agents become some sort of independent coach therapist coach therapist because like people feel comfortable a trusted friend yeah also i think in in the cases where you know the sad stories people have been abused they're too scared to speak to anyone else right so you know so if you can speak to an ai where it's not a real person it's a confidential conversation it's going to give you advice you could say hey i was i was sexually assaulted yesterday what do i do do i go to the police how do i deal with it it could be yeah like i think yeah exactly and look when you're dealing with medical stuff we're getting these sort of things it obviously gets a little tricky and needs to be like supervised by psychologists and doctors and whatever else but i think if we can build agents that can help people feel comfortable talking about their problems you know without the insecurity of it, it's great.

1:09:22If it can be helpful, that's great. So, I like the direction that they're going in here. It's very warm and encouraging. I just think that for me personally, I think that the, you know, I don't want to chit chat to a bot, but if I had, you know, even as a CEO, right, if there was a CEO coach in here and I go and say, yeah, I'm struggling with this employee in the company and they go, okay, what is this person doing? It's like, this is what they're doing they're not playing nice with their colleagues but they're an amazing performer the work is great i just can't get rid of them but they you know you've we've all had these people in organizations right and like who do you talk to about it right you don't want to talk to you know the people on their team you know it's hard to speak to the managers sometimes because there's like a bias there and you know depends how it goes uh i like this i think this i think you're exactly right there's a front line of discussions that what we're i think what we're talking about here is there is a group of people who will go to a professional yes and then there are a group of people who won't and so for people who will go to a professional it's not for them but for people who can't afford it don't have access to it is this better than nothing or is this better than talking to one of your friends and if the technologists in collaboration with professionals can't make something that's better than going to a random person a rando so is all you have to do is beat rando advice to make this valid in the world and you can hit random better than random for sure today um and then of course the criticism of this will be well it hallucinated it gave you bad advice whatever it's like if it has a disclaimer and it says you really need to talk to somebody this will give you an idea of what your session might be like with a therapist somebody who can deal with trauma somebody who can give you career advice so this isn't a substitute for that but this will give you a simulation of that i mean we let people simulate uh war or being a criminal in grand theft auto like literally you could be a gangbanger in grand theft auto and you can have that experience you could do red dead redemption and be a gunslinger well why wouldn't you allow people with proper disclaimers hey this is what a therapy session might be like for you this is what a career coach might be like for you um but of course it's not so click this button to go talk to somebody and here are your options this would be amazing so great great catch there i do think replica um is trying to upsell too much i think what they should do is make it time-based i agree they should make it time-based give everybody everything for 30 days get them on the hook and or 15 days whatever it is or or maybe it's not maybe they are they they they're just experimenting right now but my best advice would be let people have like 48 hours with it and then you know make it time-based uh you know um not not let them understand how great it could be you know i think this is an interesting conversation and we can probably finish off on this but um the the drive to monetize ai this early in the process is i think a little bit misguided in the short term.

1:12:33Explain why. Explain why to founders, why it's misguided. We've gone through this like Zirp policy for a couple of years. We've gone through, you know, very cheap capital for years. And even in the early days of Google and whatever else, I mean, they were making money, fine. But there was still a need to sort of have the industry mature a bit more to understand what the right business models were. Like when Search came out, no one knew what the right business model for Search was. CPC wasn't even on the horizon. Google was losing. they had like 20 months worth of runway left but they had billions of users or billions of searches a month and so it took a while for them to sort of find the right model they you know they bought applied semantics they they did a bunch you know they copied overture of goto.com right and and they built out search and they monetized the product i think by prematurely monetizing products you you lose out on the big uplift that you're going to get from mass model users trying it and the data and the data exactly yeah i mean you gotta get that data yeah so so that so that that i think is um is a critical part of like the cycle we're in right now yes unfortunately capital is not available to everyone easily some people can raise capital some can't some some have to make a profit and grind away um i mean just from my experience with wait room right now we're a small team we're burning a small amount of money we're just grinding out features the product's free you can have free video conferencing on wait room we don't have an hour limit a 40 minute limit like zoom you can do it because we just want the data we want to see what features they're using what ai features are popular feedback from customers that's more important than trying to sort of prematurely monetize the product and what i'm seeing with all these products is and i get the reason right money's tight we're in a tough situation but be thoughtful about if you need to make money off the product there are underlying costs that you have to cover be thoughtful of not price getting your product too early on so that people who would be paying later on never get to the point that they're willing to pay.

1:14:28Exactly. If you think about the Zerp environment, people went crazy and didn't even think about monetization. They didn't even think about their spending. And now they're so obsessed with getting money in the door because they're afraid. The pendulum has swung too far yet again. Yeah. I think it's well said. It's a really important thing for people who are founders. Great job. You brought some great demos. We talked about some great news. for folks who are running businesses. I want you to drop everything and go to weightroom.com and sign up for Vinny's service. And then I want you to come up with the three things you like best and the three things that suck.

1:15:01And then I want you to at mention Vinny. Vinny, remind everybody of your Twitter handle. At Vinny Lingham on Twitter. At Vinny Lingham. And give Vinny Lingham some, I'm sure he's Vinny at weightroom.com. Give him some feedback on his product. What could be better? What's great? And let's get him some early beta users. here finally um we are doing founder university uh again on starting it's i guess this is uh our 12-week course is starting again july 31st and then october 23rd what's our founder university course it's 12 weeks it's basically free uh you pay 500 bucks vinnie when you start the program if you come to all 12 monday sessions you can miss one probably just ask for a forgiveness but if you come to all 12 talks we give you we charge you back the 500 do you know what our completion rate is now our graduation rate 100 90 94 when we gave it for free it was like five percent so it turns out if we you give us 500 we pay you 50 bucks every time you come to the course essentially uh but what's happened is uh we've had about half the more than half the teams that come to this and we try to have builders come to this uh 250 builders come to each cohort we do it four times a year um half of them were more aren't even incorporated we watch the progress vinnie and we give 10 to 20 people in this program 25 000 checks as investments from us and we become the first investor in your company which is something i always have a lot of pride about doing being the first fund to put 25k into these companies if you want to join us apply by going to founder.university.

1:16:40You can have an idea. We want you to be a builder. So UX design, developer, growth hacker, can be a copywriter, marketer, but you gotta be a builder. We don't just want idea people or you can be an ideal person, operations person with a builder. We give preference to people who are two person teams. So go check out founder.university. And if you wanna read the deal memo for my fourth fund, we got a lot of plugs here at the end. I tried to share Vinny. uh the deal memo that i send to like top lps because i like people to read my deal memo how i'm thinking about how i'm going to deploy the fourth fund and you can read it at launch.co slash memo launch.co slash memo you can read the launch fund for deal memo this is my strategy now why would i give the strategy away vinnie you're probably saying we're not in competition in the early stage it's us y combinator tech stars nobody's really in competition pair vc uh um Sequoia has got their new, they're experimenting with a new accelerator type program.

1:17:41All these companies need 20, 30, 40 investors in their first two or three rounds of funding. So we're really not in competition with anybody. We're in collaboration. So go read the deal memo. And if you want to be an LP, Vinny, are you an LP in my funds? I don't know. Maybe you can come. I invested in some of your deals. Okay, some of my deals. But let's see. Let's see. Maybe I get a$250K check from you right now. You spend$1 ,000 on a stake with gold leaf on it. Maybe I get you, I'll lend you as an LP on my fund. We can chat about it. Okay. Let's read the deal now. I look at the deal. I actually like the Deep Sentinel security one that you did.

1:18:17Ah, yeah. Deep Sentinel is a great company. What a great idea. And speaking of AI, I got to have him on. But I think, you know, those companies are going to do really well. If you haven't seen Deep Sentinel camera, which I have at my offices, cameras that have two-way microphones somebody comes up they know if it's vinnie and jason if we work in that office every day but somebody comes up and it's susan or john and they're like hey can i help you not like in an aggressive way but if it's three in the morning and you're wearing a hoodie and you're trying to crack the door open like hey please leave the vicinity we're calling the cops they're on the way and it can have an alarm go off so they have like virtual monitoring through a two-way essentially nest cam uh but it's their own hardware uh at this point and uh Incredible company.

1:18:59Well worth checking out. And I think obviously AI is going to be really interesting as well for those type of companies. All right, everybody. We'll see you next time. Bye-bye. On behalf of the producers and the partnership team, thanks for listening to episode 1768. We'd like to take one more time to thank our partners, Notion. Try Notion projects for free today at notion.com slash twist. LinkedIn Jobs. Post your first job for free at linkedin.com slash twist. And Intercom. Try Finn Intercom's new AI customer support chatbot at intercom.com slash fin. If you're looking to become a partner of This Week in Startups, you can email hannah at hannahatlaunch.co.

1:19:45That's hannahatlaunch.co. Thanks for listening.

1:19:52Thank you.

From the publisher

This Week in Startups is presented by:

Notion just launched Notion Projects, which includes new, powerful ways to manage projects and leverage the power of their built-in AI features too. Try it for free today at notion.com/twist.

LinkedIn Jobs. A business is only as strong as its people, and every hire matters. Go to LinkedIn.com/TWIST to post your first job for free. Terms and conditions apply

Fin can’t burn its mouth on hot pizza. Or wave at someone who wasn’t waving at them. Fin can resolve half of your customer support tickets instantly before they reach your team. Meet Fin. A breakthrough AI bot by Intercom – ready to join your support team today. Visit https://intercom.com/fin

*

Today’s show:

Jason is joined by Vinny Lingham to break down IRL shutting down after faking 95% of users (1:12), ZIRP fraud (12:59), Databricks acquiring MosaicML for $1.3B, and some AI demos! (1:02:31)

*

Check out Waitroom: https://waitroom.com/

Follow Vinny: https://twitter.com/vinnylingham

*

Time stamps:

(0:00) Vinny joins Jason

(1:12) IRL's 19M fake users

(5:45) Twitter bots and the fake account problem

(11:50) Notion - Try it for free today at https://notion.com/twist

(12:59) Diligence in early-stage startups

(20:59) Databricks acquires MosaicML for $1.3B

(27:02) LinkedIn Jobs - Post your first job for free at ⁠https://linkedin.com/twist

(33:09) Testing Google’s new generative search

(37:47) Fin - Try Fin, Intercom's new AI customer support chatbot, at ⁠https://intercom.com/fin⁠

(38:31) Testing generative search on the most expensive steaks in Los Angeles

(41:56) Vinny's thoughts on Toxic Wealth and the Titan tragedy

(53:34) DeepMind CEO says Gemini is more capable than ChatGPT

(1:02:31) Vinny demos Colorize and Replika

*

Read LAUNCH Fund 4 Deal Memo: https://www.launch.co/four

Apply for Funding: https://www.launch.co/apply

Buy ANGEL: https://www.angelthebook.com

Great recent interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland, PrayingForExits, Jenny Lefcourt

Check out Jason’s suite of newsletters: https://substack.com/@calacanis

*

Follow Jason:

Twitter: https://twitter.com/jason

Instagram: https://www.instagram.com/jason

LinkedIn: https://www.linkedin.com/in/jasoncalacanis

*

Follow TWiST:

Substack: https://twistartups.substack.com

Twitter: https://twitter.com/TWiStartups

YouTube: https://www.youtube.com/thisweekin

*

Subscribe to the Founder University Podcast: https://www.founder.university/podcast

More from This Week in Startups

All 653 episodes
IRL to shut down after faking 19M users, ZIRP fraud, Databricks acquires MosaicML for $1.3BThis Week in Startups · 1 h 20 min
Listen in VO