David Friedberg: the question is can chat gpt open ai and microsoft that duopoly can they beat apple and meta

Chamath Palihapitiya: open source your answer no wow wow stunning yep snap snap you snapped it you called you shoved all the chips in you believe meta and apple will beat the mighty microsoft open ai duopoly

SPEAKER_07: and look we've seen this movie before this week in startups is brought to you by vanta compliance and security shouldn't be a deal breaker for startups to win new business vanta makes it easy for

SPEAKER_10: companies to get a sock to report fast twist listeners can get one thousand dollars off for a limited time at vanta.com twist scalable path want to speed up your product development without breaking the bank since 2010 scalable path has helped over 300 companies hire deeply vetted engineers in their time zone visit scalablepath.com twist to get 20 off your first month and linkedin ads to redeem a 100 linkedin ad credit and launch your first campaign go to linkedin.com this week

SPEAKER_13: in startups all right everybody welcome to this week in startups it's time for our ai roundtable where we do demos me and sandeep have been doing this and the crowd loves it the audience has been going crazy for these ai demos every week we try to find five demos that y'all will be inspired by or intrigued by and just to keep tabs on the absolutely blistering pace of ai innovation in our industry you can see all the demos at this weekend startups.com ai you can go there see all the demos and just really catch up with what's been an incredible i don't know what is it like we're 14 15 months into

SPEAKER_16: this since chat gpt 3.5 was released and took the world by storm yeah how are you doing sonny yeah you

SPEAKER_18: get your super bowl you got your super bowl behind you you were very very excited about the super bowl

SPEAKER_21: you know there was a lot of opportunity and it all went out the door with that last play

SPEAKER_18: oh man uh i don't yeah i mean was it was this a good super bowl oh i mean i know you're a 49ers fan but yeah putting that aside was this a good super bowl i mean the amount of excitement around it with

SPEAKER_28: this taylor swift stuff was extraordinary yeah that was probably a bit too much but you know anything that goes into overtime well come from behind over time both teams have a shot both teams score but one

SPEAKER_29: scored a touchdown that's about all you can ask for so yeah yeah up there okay i think it was the most

SPEAKER_33: watched event ever wow you see yeah 135 million or something i don't know what the number was crazy

SPEAKER_28: is like an average mr beast video gets more views so that's how much our society has changed gets more

SPEAKER_35: views but not live so i guess the question is like of people watching live yeah because how do you think

SPEAKER_13: about that um it's just you could look at the first 24 hours of a video and you know how different is that than watching it live probably not much different everybody's at the water cooler seeing it the next day cumulative videos also matter you know if you were to go on to youtube the top most viewed videos are like baby shark gangnam style yeah yeah things that kids listen to over and over and over again so i think mr beast falls into this category it's not taking anything away from them it's

SPEAKER_46: repeatability so mr beast doesn't fall into repeat it you can't watch that video again oh i think

SPEAKER_18: kids do watch them over and over again yeah i don't not like baby shark where like they listen to it you

SPEAKER_35: know every day or 10 times in a row yeah but i do think that there's probably some repeatability there

SPEAKER_13: i think young people like to go back to the archive and watch the previous episodes kind of like

SPEAKER_50: sopranos right like i've watched the sopranos from start to finish three times so okay yeah okay you

SPEAKER_13: know i think i think it is rewatchable but maybe not like a an audio song but either way it's pretty amazing when you think about mr beast releases a super bowl every week in terms of viewership from from that perspective yeah and it's a the duration is also interesting you know the duration of the

SPEAKER_58: super bowl is you know five hours four hour experience whatever yeah he's now got 237 million subscribers pretty amazing and there's a bunch of these channels from india that are also in the top 10

SPEAKER_60: t-series yeah t-series is like uh like a recording label in india got it so it'd be like uh uh what was it back i vivo uh or yeah sony records columbia records and then set india 168 million subscribers

SPEAKER_13: so wow i don't know exactly i think that's because the denominator right you have just so many people in india you got a billion people it's you know 25 of the population 17 of the population 237 million subscribers to mr beast that's 330 million people in the united states so he's got roughly if half of those are us he's got half the u.s population maybe subscribed i wonder how many of the 207 237 million subscribers are active u.s subscribers that's really the statistic i would look at so he

SPEAKER_28: just released a video on saturday let's take a quick look at that and we can see how many views

SPEAKER_71: that one has for just a rough sense of you know how he stacks i mean i think he gets like a hundred

SPEAKER_19: million um yeah in the first couple of days so yeah he got 77 million from three days ago right so

SPEAKER_71: and two weeks goes at 128 million wow that's wild yeah so that's you know it's basically within a week

SPEAKER_74: yeah he's got a super bowl there's also you know one of the other things to you know when you're looking at these statistics and how they're recorded you know there are drive-by views so

SPEAKER_13: you know if you were to discount people who watched 30 seconds in in both cases the super bowl yeah and mr beast you know you might get a lot of drive-bys on social networks like youtube or certainly twitter you know i think maybe half the views we get on twitter for all in or for these

SPEAKER_14: podcasts you know people might be dropping off in the first couple of minutes and so do you really count those as views at what minute do you you know i mean they are views technically but you know are

SPEAKER_16: they viewers so to speak right so duration of views matters and percentage complete that's one thing he obsesses about is a percentage completion complete yeah yeah all right let's get into our ai uh demos

SPEAKER_28: here let's just do it yeah we have some really really cool stuff today it's a lot different than

SPEAKER_29: some stuff that we've seen before so i'm very excited so let's start off with one that is called

SPEAKER_28: agent hub dot dev agent hub dot dev got it and so they have templates and so they have them for

SPEAKER_29: different types of use cases i'm gonna do like a simple one just for demonstration sake daily stock report generation and so i can say i want an agent that generates a daily stock report so i'm just going to use their template i'm going to zoom in here and what this does it's like a you know like you can

SPEAKER_28: imagine like a little bit of a a workflow tool and i'm just going to say every day i want to see apple amazon and nvidia right we do that and over here we hit run what it will do is it has a workflow here which walks through you know what it does including web stripes scraping including some batch workflows

SPEAKER_60: so this looks like your standard flow chart you know chart making software here if this then do that

SPEAKER_29: yeah exactly and they have all those templates there so you can look at the templates but while this comes together as generating the files you'll see i will get a report so i've got three files here and so i've got apple so you know if i were to let me just download these three

SPEAKER_28: and if i were to download them and and look at nvidia and you'll see i got this little oops that one

SPEAKER_93: didn't come out correctly so let's look at apple oh well it bugged out on me it bugged out but here it did

SPEAKER_95: a script where it said the average price is this yeah so if it did do it successfully it would have

SPEAKER_28: filled that in yeah and it was working for me before i want to give them credit so i don't know if i if i messed it up or something but uh yeah it was pulling the price on these things and it was

SPEAKER_95: incredible so so it's going out into the open web and doing a search scraping that data normalizing

SPEAKER_28: and putting into a report exactly and you know like they have a whole bunch of different templates in

SPEAKER_29: here you know i just i just picked like a simple one but they have ones for automated sales emails and outbound email creator so these kind of multi-step flows with agents which i think is really really

SPEAKER_13: cool so this reminds me a little bit of zapier or if this then that yes combined with what we used to call workflow software in the 90s and 2000s lotus notes was a original piece of workflow software that i actually used to script i wouldn't call it coding but it was scripting like this hey go collect this information and then send it to this person so if you think about like paying an invoice you would create workflow software an invoice came in uh you know there's somebody it's this is a you know office supplies invoice it has to go to the person who requested the staplers uh then it has to go to their manager to approve then it has to go to the office person to approve then it has to go to the cfo accounts payable person you would just send that invoice around so you know this is kind of stuff like that where you could build a workflow very interesting um and this is where you can start to we've seen a couple approaches to this already right sunny we had somebody who was looking at uh recreating a web browser remember that one we had with the chrome extension just a couple weeks ago

SPEAKER_18: that will kind of blew our minds yep you know this is kind of the mind-blowing okay so automation

SPEAKER_111: this is oh here we go so this is analysis yeah yeah so the analysis so basically you can so this is

SPEAKER_82: more than just giving you the stock price the stock price of amazon has been fluctuating the past few

SPEAKER_16: weeks on the last date the stock closed at 172.34 showing a slight decrease from the previous day the volume of trading has also been relatively high indicating active market participation yeah the historical trend we see there was a spike in stock price on this date reaching 174.45 however there's a downward trend after that with the stock closing i mean i i don't know if you

SPEAKER_13: noticed this but there are automated stories in uh apple stocks where they tell you like this is an automated story have you seen those yeah and so like you know go check my uber stock or my robin hood and it's like yeah it's kind of like this where it's like a little synopsis of what happened in the market i'm like okay well this isn't providing me enough value just i would rather you know i can

SPEAKER_16: look at the stock and know that but i could see this being valuable over time and this is going to

SPEAKER_115: get better and better what you'd want to do is you'd want to customize it listen a strong sales

SPEAKER_116: team can make all the difference for a b2b startup but if you're going to hire sharks you need to let them hunt and you can't slow them down with compliance hurdles like sock 2. what is sock 2? well any company that stores customer data in the cloud needs to be sock 2 compliant if you don't have your sock too tight your sales team can't close major deals it's that simple but thankfully vanta makes it really easy to get and renew your sock 2 compliance on average vanta customers are compliant in just two to four weeks without vanta it takes three to five months vanta can save you hundreds of hours of work and up to 85 on compliance costs and vanta does more than just sock 2. they also automate up to 90 compliance for gdpr hipaa and more so here's your call to action stop slowing your sales team down and use vanta get a thousand dollars off at vanta.com twist that's vanta.com twist for

SPEAKER_28: one thousand dollars off your sock two you have the jcal yes chef and if i worked for you and you know there's something you want on a daily i would build something inside agent hub and it would

SPEAKER_23: basically get you something every single day that you wanted to see could be an output from um yeah i

SPEAKER_12: could see this working for like hey collect all of the reviews of our product let's say we were you

SPEAKER_13: and i were running a cpg company yeah every day go to amazon go to our shopify store go to target tell us all the new reviews give us how the reviews are trending and basically write a report which we might have you know a 30 40 50 an hour employee doing previously yes who is you know our our review analysis now the review analysis could be scripted and automated without a developer and then that person doesn't need to have a job or they can move up the stack to doing more important work so this is uh

SPEAKER_14: this is an a for me yeah it's an a i mean i just like the structure of it too um where it's like drag and drop great interface and yeah i could see this eliminating a lot of repetitive work slash jobs and i think that's what we're seeing in all these layoffs that are happening i wrote i did a tweet today

SPEAKER_58: about it i think these layoffs are really about three things i'm talking about media and tech layoffs let me run this by you and see what you think i think it's about three things one people had very high salaries during peak zerp right there was a massive competition for employees and a lot of

SPEAKER_13: people i don't want to i'm not you know trying to be cruel to anybody but when there's a competitive market and you have unions maybe at the media companies fighting for people as they should maybe salaries got a little inflated 10 20 30 now that doesn't seem like a lot but if you can lay off 10 of your workflow force and then replace them with people who are 40 cheaper well or 50 cheaper or they don't have the stock options which we saw evan spiegel i think do at snap you know he kind of got rid of top heavy people who had big stock packages i've heard that people who've gotten laid off from google have been asked to reapply for their jobs with a fraction you know like a third less

SPEAKER_29: comp have you heard about this boomerang trend yeah so it's interesting that you use the word boomerang because another thing that i've heard layered onto that was there was an echelon of folks that were so highly compensated that in it certainly weren't the younger folks but it was like the folks that were 20 years and mid 20 years and like mid 40s and up that were at like really insane compensation

SPEAKER_28: levels of stock and salary and what i'm finding is in some cases are being asked to come back reapply so that they can be reset in other cases they've created a lifestyle for themselves and they can't do that because of you know overhead mortgages and overhead and you know they got used to making like i mean you know jay kelly you know even in our careers it was insane to think about making a million to a million and a half dollars a year and this was pretty common in that kind of echelon

SPEAKER_77: of i've been in the tech business for 25 years i'm a gen xer you know i did three job hops over 25 years

SPEAKER_13: or 20 years i went from google to facebook to uber and each time i got a quarter million dollars more

SPEAKER_14: and i'm this elite person but yeah and so then the second piece of this is i think people are looking at it saying can i automate what this person does and so here are these tools you know i look at

SPEAKER_13: these tools and i always just squint a little bit and i'm like okay uh 10 exit so here it did like a very basic analysis highs and lows or whatever but if you asked it hey tell me about the contempt forget about even like programming it but you could say to another lm what question should i ask this what would be common questions to ask about stocks and i would say oh how does it compare to

David Friedberg: its competitors yes how does this compare and it could even come up with the questions to ask yes

SPEAKER_139: and then put them all the way down from ai all the way down yes so just like you get suggested searches

SPEAKER_13: oh you did a search for i don't know burritos and then it said well here's burrito recipes here's burrito low carb recipes keto recipes like you didn't know to even ask oh is there such a thing as a keto burrito and it's like yeah there is here it is boom you know burrito alternatives whatever yep so that's what i see with this is okay this replaces a 30 an hour person right and if you ever

SPEAKER_16: want to do back of the envelope math as a leader in a startup my trick is always 2 000 hours right the average person puts in 2 000 hours generally accepted practice so 30 buck an hour employees

SPEAKER_14: 60 000 you know you add 20 to that so it's 36 an hour whatever 36 times two you can easily do the math 72 000 a year or whatever you start looking at that i think this replaces probably like that level of employee yep yeah you know now what's the next one it goes after so you have the resetting the automation and the third piece is globalization and people finding people offshore who maybe

SPEAKER_148: really want work and so those are my three things and the best way to think about what it goes after

SPEAKER_150: is look at the templates because okay let's see youtube video processor oh seo optimized article

SPEAKER_152: generator generator these are exactly 30 40 an hour jobs in my experience and when you hire if you hire

SPEAKER_30: consultants they then they tend to charge double imagine i think you wanted one of these a linkedin

SPEAKER_24: profile processor reads right you know warm oh my lord yes linkedin warmed reader oh this is incredible wow and i was just doing this one right yeah oh this is stuff so now we we talked about this category

SPEAKER_160: business process or outsourcing yeah bpo this is what happens in india this is what happens in the

SPEAKER_13: philippines there's been an entire industry of doing this now i think some of those people were doing their own scripts but now this you're going to be able to do this yourself so you know inbox

SPEAKER_163: categorizer uh inbox summarizer uh email order replier man this is next level um yeah it is it is like they're literally atomizing people's jobs right and so they're atomizing people's jobs and so in if

SPEAKER_14: you are the leader of a company and you want to do seo and blog posts and social media and whatever competitive analysis like i could see doing competitive analysis here like let's say

SPEAKER_13: there was another podcast about you know i'm a food podcast and you're a food podcast and we're competing against 10 other chef podcasts you could have this go out and summarize what the other chefs are doing then have the ai ask who are their guests make me a list of their guests you know which of their videos are breaking out above average and you can do competitive intelligence competitive intelligence boom wow i mean if you are going into the job market today this is a list of what you could

David Friedberg: avoid or this is a list of how you present yourself as the uh grim reaper yeah i'm coming to your startup

SPEAKER_13: as the grim reaper you know the guy with the doors the meme with like knocking on doors okay knock on

SPEAKER_48: the door for competitive analysis knock on the door for yeah whatever yeah what jobs are they coming for

SPEAKER_163: wow okay i mean i give it a name boom congratulations i really also i think interface matters here

SPEAKER_13: yeah i could see this interface like the top 50 of my team members are good at this kind of interface

SPEAKER_14: yeah and that's one of the things i've tried to work on on my team is just please learn how to use zapier please it'll make me happier if you learn how to use zapier zapier ah i love zapier and i'm just constantly trying to get my team members to learn zapier or basic you know coda scripting notion

SPEAKER_23: scripting you've got some cool scripts that you told me right that take your expenses from your credit

SPEAKER_13: card and put them into your slack and i mean it's really basic but like just you know people game changer it's a game changer like if we're doing an approval process i worked on getting credit cards for my team members that could be turned on and turned off like you know those categories of cards and then i like to have them sent into a slack room yeah with that person and with you know the operations person so then everybody sees your expenses as they're coming in if you forgot and you subscribe to something you know and you subscribe to the wall street journal last year but somebody else has a wall street journal and we don't need to and whatever you know at least we see it and you

SPEAKER_19: can have a conversation about it right um so transparency yeah really interesting stuff okay all right let's

SPEAKER_188: keep going hey well start up shrunk start shrunk start it's hard to balance hiring top-tier developers and keeping your burn rate under control but these days i see a ton of founders successfully doing this

SPEAKER_190: by hiring remote talent so let me tell you about scalable path it's a software staffing company that can help you build an awesome remote developer team and the right developer isn't just a list of technical skills we all know that it's about their personality it's about their work ethic their motivation and their fit within your team and scalable path knows this so here's what they do their team will get to know your vision they're going to get to know your needs and then they're going to develop technical challenges tailored to the roles you're hiring for and these challenges are conducted live and on video so there's no gaming of the system you're going to get great people they also evaluate each candidate's soft skills like communication attitude and work style scalable path has completed more than 300 projects for their clients and they have a network of 30 000 developers they've been doing this for over a decade they know what they're doing so you're going to be in great hands here's the best part twist listeners get 20 off their first month if you're ready to scale your dev team and your business check out scalablepath.com twist once again that domain name scalablepath.com twist

SPEAKER_29: 20 off so the next one this is super interesting let's put it like in a in a very similar category okay so this is exactly no i apologize to people i apologize to people with jobs that we're saying

SPEAKER_28: are going away so this one creates like these four types of um basic analysis you can do a business overview competitive overview financial review or buyer's list for stocks and so basically you put

SPEAKER_29: it in here it's a lucite.app and i'm gonna have to share just a different window because its output is

SPEAKER_88: like a powerpoint presentation and so now so i i did this okay i have a computer presentation here okay

SPEAKER_28: and i basically did nvidia and so i have this seven slide presentation that does nvidia competitive landscape so it's like gpu based products these are the folks gaming products uh data center and cloud wow and you know basically feels accurate yeah i mean like let's look at it these are gpu competitors

SPEAKER_29: amd right intel makes sense you know these these are gaming folks that are leveraging you know their ecosystem and these are folks that compete compete within data center feels right then a little bit of a

SPEAKER_28: breakdown in each company for each section valuations and then recent m a transactions this is done by

SPEAKER_13: an ai took three minutes now the question is how much of this is correct how much of is hallucinating and that really is kind of where i think the industry needs to get to is there needs to be ai

SPEAKER_14: checking the facts of ais is this true yes and that's i i don't i haven't seen that but this would be a really good product would be an ai copilot that fact checks other copilots i know this sounds

David Friedberg: stupid that's very meta it's very meta but imagine there was a fact checking a copilot so i'm using

SPEAKER_14: microsoft copilot and i say give me the competitive analysis for nvidia and it gives it to me now i say to

SPEAKER_13: my fact checker please go into this report and check each of the facts and find me sources for it on the open web etc and tell me if there's things wrong and so that would be interesting is to take that pdf upload it to claude or chat jpt4 and say can you tell me if any of this is factually incorrect

SPEAKER_14: do you think it would be able to do anything like that yet or we're not there yet so i believe

SPEAKER_29: that the you know the foundational models are capable of doing this the challenge that they're running into is in order to fact check they have to use third-party data systems they haven't licensed

SPEAKER_211: and so yeah because you know in the training data they don't have like the current stock price

SPEAKER_212: of they're pulling it from some article somebody wrote previously in the open crawl index it's not

SPEAKER_88: correct yes and so in order to fact check it they have to subscribe to like a google finance api

SPEAKER_28: or bloomberg or you know someone with an api and that's where we haven't really crossed the chasm yet we've been talking about this where yes you know the closest is like perplexity who has some data relationships but like no one's really fully done it with like a real data provider and i think the minute someone does that we're gonna really see a distinct jump in the capabilities and i think everyone's been on the edge of it but yeah it feels like it's coming very soon i think it's a great idea

SPEAKER_35: yeah i would like to have a fact checking api so this is something in journalism or just in general

SPEAKER_69: like is this a true fact so this is a request for startups here is this true this is true is this true

SPEAKER_13: dot ai it's a toolbar it's a copilot it's a system tray and it is reading everything in real time like grammarly does yeah and it will tell us it will it will just look at facts in the background and just try to independently determine if they're correct and you could i mean spelling is one thing that right is this word spelled correctly is something grammarly is doing grammarly is also doing is this you know is the comma in the right place so here when we're looking at this nvidia financial snapshot i would like it to know this is nvidia no this is finance data so it says what is this okay this is nvidia's five-year historical chart yeah and it says go find it and then compare it to this

SPEAKER_14: um and that would be super interesting to me as a as a device um and then just highlight things it's something i've always wanted as a as a product or a service as a fact checking as a service and having a human plus ai do this would be really fascinating fact checking as a service

SPEAKER_28: i wonder if that is like a good startup idea this one i just shared both uh the the sheet and the google stock price i got the the stock sorry the apple stock stock price i just did the share there so i got those things accurate which were 721.28 it's it's at 0.3 and it got the market cap

SPEAKER_231: rate and then if i go back to the financials that's a little bit harder to do uh through these tools but

SPEAKER_234: yeah what is the name of this one lucite dot app lucite l-u-c-i-t-e dot app yeah i'm gonna give it a b

SPEAKER_16: okay i think it's like really a great start yeah the fastest way to produce investment banking style

SPEAKER_14: materials it's a great idea remember i was talking before about these kind of projects and you know the dollar amount of the employee it's replacing now we've gotten to uh you know a hundred thousand dollar employee like an analyst at a bank who makes these slides and so i pulled this one up oh my

SPEAKER_240: god yeah this goes back to your previous point mckinsey puts 3 000 staffers on reviews citing concerns over performance as the recently frothy consulting business slows yeah i mean if people

SPEAKER_18: don't read consulting reports and the consulting reports are just a hodgepodge of web research

SPEAKER_28: well you might as well just run it through chat gpt or you've got an example right from all the stuff you've ever made so you've got a corpus and then you can just let train an ai on that and say go make things that look like this you know what's interesting about lucite is that's what they did right they trained and said these are like the five different basic you know investment banking um

SPEAKER_13: style presentations make them for us right they they had the templates and they made a verticalized tool to just focus on that yeah just like we have a company saga um that is doing just screenplays and storyboards and i told them i was like you know this whole ai wrapper thing is nonsense you're building a verticalized app you're going to build so much on top of it that you know it'll it'd be like going to google and saying you could do a spreadsheet at google like you could use google search to do a math equation right you can put a math equation in the search box or you could use google to search for flights or whatever but that doesn't mean you're not going to do expedia kayak or a google sheet or microsoft expelled thing like you can get some basic functionality done with chat gpt4 or claude but if you verticalize this man they can really really refine it and build features around it like fact checking like collaboration tools etc so i love these verticalize i mean if the ai wrapper

SPEAKER_260: vertical i think is going to be very big i think it's going to be awesome it's going to be huge

SPEAKER_71: yeah i think so as well yeah but that's where we're going okay so i give this b i liked it okay

SPEAKER_28: yeah i like it's very solid i'm like b plus you know like okay i was probably higher but i like the idea you said if there'd be so if there could be a way that you it could verify that the data is correct

SPEAKER_68: because now you have is yeah citations i think that would be really good i mean it's like really cool

SPEAKER_16: it creates a deck time is money it's a really nice yeah all right job blue side b2b marketing is not an

SPEAKER_116: easy job it's much different than b2c why well enterprise buying cycles are much longer it's not like you see some cool mug and you buy it as a consumer no enterprise buying cycles are long and it's hard to find and reach decision makers through most marketing channels we all know that so that's why you need to check out linkedin ads linkedin recently passed a billion users which includes 180 million senior executives that's right about 18 percent are senior executives but there's also 10 million c-suite executives on linkedin ceo cfo cto's chief strategy officers you know i'm talking about this means you can find and target these decision makers in a respectful environment linkedin and according to linkedin their ads platform generated 2 5x higher return on ad spend when compared to other social media platforms and it's easy to understand why linkedin equals business business equals linkedin you understand this when people are on linkedin they're there to do business they're not there to talk about politics or pop culture they're there to do business right so your business message is going to be in the right place make b2b marketing everything it can be and get a hundred dollar credit on your next campaign go to linkedin.com this week in startups to claim your credit that's linkedin.com this week in startups terms and conditions do apply jake now this is

SPEAKER_29: going to be a two-part demo two-part i like it well our friends at google released gemini ultra

SPEAKER_163: uh explain what that is to the audience because i saw this um and i was a little confused about

SPEAKER_14: what's the difference between gemini ultra and google bard and then i saw a bunch of people were

SPEAKER_28: losing their that they got access to ultra gemini ultra yeah on like their google app okay so there's a few different things bard's been renamed to gemini.google.com okay fyi okay then gemini advance which is and this is where this gets a little bit wonky gives you access to gemini ultra 1.0 ultra 1.0 is their most advanced model which is great the issue that i have is i have a gmail account that's tied to a google fi account yeah i'm going to pull this up here and this is like totally wild it's crazy because there's no way to try it so when i go here and i go gemini advance which i want to try and i go upgrade i get this crazy message saying google one's not available although i'm a google one customer huh because if you're a google one customer through google fi so i'm going to hand this one over to you jake how okay if you can go to gemini.google.com and see if if you can pull it up then we can try out gemini ultra together or a gemini advance

SPEAKER_168: which is gemini ultra okay so bard is now gemini i got that alert on my gmail account but i don't see

SPEAKER_24: ultra you should be able to click beside the hamburger there should be a little down arrow beside gemini oh it says upgrade gemini advance i click upgrade takes me to google one oh here we

SPEAKER_113: go 19. yes so you gotta pay it all for you 19 bucks after so i start my trial for gemini events okay so

SPEAKER_277: now they have a chat gpt4 competitor yes and uh okay so i subscribe and let's see what happens here and

SPEAKER_83: so i haven't been able to try this one because this is my little bit gripe to google so i gotta give it

SPEAKER_28: a really low grade because there's no way for me to subscribe and try it so you've got to you've

SPEAKER_13: got to do this demo okay here i am how can i help you today so okay i'm in there you go yeah all right so what should i do as a search here what would be how would we test this well we just did nvidia

SPEAKER_58: competitors right yeah uh make me a table with nvidia's top competitors and the most important information about them so i told it to make me a table and we got a little spinning uh gemini wheel so see if it even understands a table absolutely and patterned our amd yeah i mean it's yeah but like you know

SPEAKER_290: not great um how about what's what's cool is like lucite did a better job that's what's interesting

SPEAKER_58: i did a better job so this shows you how vertical apps are doing better than just man line what are some event spaces in napa i can use for a 200 person corporate retreat you know i met with a company that is doing like an ai for event planning and i've been using event planning so here it goes it's going to

SPEAKER_291: google maps for some reason and uh let's see it's giving me web pages so this is like a google search

SPEAKER_72: uh no that's not what i wanted terrible answer searching the web here are some results yeah give me

SPEAKER_50: 10 great locations to host a corporate off-site for 200 executives in napa slash wine country

SPEAKER_19: put the slash wine country there see if i can fuse it let's see if it gives me 10 i mean i'm giving a very explicit instructions here i mean so far this is a total fail yeah well i mean like literally

SPEAKER_295: wasting my time i could have gone to the web i mean if i wanted to google search i would have done it

SPEAKER_30: yeah one of the things that it's meant to be really good at is multimodal so maybe ask it a question

SPEAKER_28: where it's you know bringing back some images or some ideas it did do a map of some items it didn't

SPEAKER_277: give me 10 it didn't give me a number list these are not very good yeah this is not not a very good

SPEAKER_13: result it's doing what i intended it to do it just didn't give me the fidelity of results i wanted like if i were to ask an event producer for this information i would have gotten a much better result like a

SPEAKER_300: human could do this much better this is yeah maybe try your burrito example and ask it for

SPEAKER_277: yeah like you know something like that or sushi restaurants what are some great sushi restaurants with omokasa in the bay yeah these are like things you would search google for right and you wanted to

SPEAKER_262: give you the answer you don't want to deal with all the links and everything else right so yeah

SPEAKER_277: no so i mean yeah this is okay it's hitting google maps again uh yeah not great which restaurants have

SPEAKER_58: peking duck in the bay area let's see if i got like peking and get very specific i'm getting a dish

SPEAKER_15: right maybe they'll show me photos of the dish yeah same nonsense i would have been better just doing a

SPEAKER_58: google search what are the top peking duck dishes in the world at four star restaurants show me five

SPEAKER_35: to ten photos so again multi-modo trying to get this i mean yeah we gotta i think we're gonna have to sit

SPEAKER_19: sundar down again remember i told him he's gonna have to bring people in and do these searches in front of them he he fixed the name yeah i mean it's fixing the name is not as important as fixing the answers i mean this is also slow yeah oh man yeah this isn't good yep i mean maybe hakasan i did get hakasan i do think hakasan is one of the top peking duck places in the world mat32 okay wait this is more

SPEAKER_104: promising okay i take it back if you had asked me the best peaking duck in the world i would have said mot32 and i would have said hakasan both known for having like very elite peaking duck and it did

SPEAKER_277: give me the picture so this is slightly better i will say like generally speaking continue to be disappointed and they need to do more work on this i'll give gemini like a b nah i'll give it a c plus

SPEAKER_323: yeah i just don't feel like it understands what i'm asking here and i'm asking it very google centric

David Friedberg: versions of this and it feels like they're just not not getting it right enough yeah it's like it's

Chamath Palihapitiya: like close maybe it's a b minus i'm gonna go b minus and unfortunately i'm not i can't grade it

SPEAKER_190: because i can't try it and they got to fix that too i mean there i feel like google is needs to lay

SPEAKER_208: off another 10 000 people i mean i don't mean to be cruel but i just feel like it's still bloated and

SPEAKER_14: like not elite teams building elite product you know i feel like facebook and chat gpt microsoft

SPEAKER_18: tesla twitter like more elite teams and i just don't feel like this like google has eliteness going on right now it's not elite enough for my expectation of google this is where sergey samurai sergey's got to get in here and just high high expectations right now yeah for sure my expectations

SPEAKER_35: of google are real i think that's why i'm giving it lower grades is because i've got higher

SPEAKER_28: expectations of google than anybody else well i want to try it so i'll reserve to come back next week so if someone is listening or watching and you're in google please what's going on subscribe i want to pay you 20 to try it out but i mean this is the thing you know what they have so much

SPEAKER_13: bruffed at google i know how this happens like they're like oh okay we have this and i brought

SPEAKER_163: this up with sundar one time where i was like listen i did i said you know you treat google domains as second class citizens when we're paying you for google docs right like yeah if i've got my company

David Friedberg: at google docs yeah and then i try to use google fi it's like no you have to have a gmail account use your personal gmail it's like you're taking your best customers and you're making us jump through goddamn hoops so now if i want to use google fi on my org you're making me tell everybody to use their personal ones and then i don't have control over it as the you know but i'm paying you for this like

SPEAKER_163: why is there is a separation between the and i understand why there's a separation between the

David Friedberg: namespaces but like this is the same thing that happened with nest you know like you want to use google docs with nest you can't do or whatever google home yeah yeah it's just the interoperability

SPEAKER_14: and i mean maybe this is just what happens when companies get too big and there's too many product lines and cruft and stuff like that but they have to simplify this and make it easier yeah it should be

SPEAKER_29: so easy pretty straightforward i'm i'm surprised that that's happening right and you know um yeah

SPEAKER_211: the error message is fake please someone get in touch we'll try it out we'll do another demo

SPEAKER_347: another demo we'll give you another shot i mean but yeah gemini advanced i'm just not feeling it yet

SPEAKER_57: what was your grade again i'm going to give it officially a b minus okay officially a b minus but

SPEAKER_350: officially officially i can't grade it i you know i can't really give it a grade so you know but i

SPEAKER_28: want to and i'll you know like everything else i played with it before i just couldn't do it so

SPEAKER_351: yeah let's keep our round table going here you got another one for us yeah i do and this should give

SPEAKER_18: startups a lot of hope yeah big companies with their unlimited resources trip over themselves

SPEAKER_13: and we saw two companies lucite and agent hub are much better products than the mighty googles gemini you're yeah so just to give those two companies their flowers like google can't get out of their way to compete with you yet now they might figure it out eventually and just like take some giant leap forward but keep being scrappy startups you can you can beat google you can be chat gpt you

SPEAKER_58: can beat microsoft you can beat you know apple you can beat all the meta you can beat them all just by being scrappier and focusing on the customer experience everything is changing just do it

SPEAKER_29: two more quick ones okay two more quick ones here we go lightning right so the next one really cool i actually just saw this earlier today on twitter friend eric mitria i think he funded it over at

SPEAKER_28: benchmark and basically it does pcb layout using ai and so we're not going to do a layout here but you

SPEAKER_29: know there's try instant example and it gives you some very popular things i can say i want to do this or do we know you've seen these things right i'm sure you've had some this hardware just explain what it

SPEAKER_24: is yeah what an arduino is so arduino is like an open source design to run uh android right it's like

SPEAKER_13: a open source computer that you can like build off of the standard and it's got things like a usbc port

SPEAKER_14: and exactly you know uh whatever hdmi and memory so it's like a computer on a stick arduino is like a computer on a stick that hackers would use yes but here you're taking an arduino and well this is so

SPEAKER_28: when you're creating this is you know again we're going verticalized again when you're creating the pcb that's what the chip and all the surrounding components go on the usb this uses ai to do the layout of the pcb and this is a four-layer pcb that has you know these rules put it and it did a bunch of different candidates which is this candidate can be produced by the most fabs this candidate has the fewest layers this one has the shortest number of traces and all these things are factors in you know when you're when you're making hardware who like who's capable of making it more layers more expensive shorter traces means you know everything is closer together wow like to see this being done by ai

SPEAKER_367: is wicked yeah and so this is where the flywheel gets out of control because you have ai

SPEAKER_13: building the hardware that could fuel and enable future ai models now it's doing this with arduinos

SPEAKER_28: yes not nvidia chips or not and the pcb not the chips this is the pc or the board that they go on

SPEAKER_14: right right but you know it will eventually be ai all the way down and so at some point this ai will be pointed at chip manufacturing and pcbs and everything else and it's just going to it's going to have i think the first phase will be able to have insights that will inspire humans you know and save them time in making this stuff but then at some point it's going to have breakthroughs right you know i'm saying like insights are different than breakthroughs or efficiency is different than insights is different than breakthroughs so this to me is an efficiency play right like the person who's making these is going to be much more efficient at their job correct and could be breakthroughs because

SPEAKER_06: the ai can think of things in those different ways as well so you can have it efficiency plus

SPEAKER_14: breakthrough you can grow on both dimensions yeah but do you think it's actually doing more than efficiency and insights and is that the breakthrough level yet i mean you'd have to ask them or we'd have

SPEAKER_23: to pull someone in so i i think you should have these folks on because yeah uh you know look i'm

David Friedberg: not an expert in in pcb just tweet at us just tweet at the twi startups account when we put the clip up

SPEAKER_13: and just um let's start these conversations on x so x.com sandeep slash jason slash twi startups if this is your company or we mention your company when we put the video clip up or you can take the clip and put it up on your handle follow up with us and let's start the discussion on x to just keep these demos moving what do we get right what do we get wrong and where are you headed with us and my question is you know has this thing had a an insight or breakthroughs yeah not just efficiency right and

SPEAKER_35: and so i'd love to hear that but when you think about it like i was i was talking about saga this you know screenplay writing software it's giving people insights and breakthroughs right you put in

SPEAKER_13: your screenplay or you say hey this is the tv show i want to make it's going to be nine episodes i want this character arc and i i just gave it like well i want to do the sopranos plus blade runner you

SPEAKER_14: know i want to have like a mafia story and i want it to be a mafia story in a cyberpunk mafia story seems pretty interesting to me like a crime family in the blade runner setting right so just it would be pretty dope right yeah like who's tony soprano in blade runner you know and what are they doing and what's their crimes and you know well what are they trafficking in right are they doing

SPEAKER_29: modern seinfelds like when i just want to entertain myself i just have one of these just write a modern

SPEAKER_82: seinfeld episode and so for a writer who's in there like it's going to give you insights that you

SPEAKER_14: know i was talking to the founder of saga and he's like you know i was like how do people get these

David Friedberg: insights currently it's like you know they're in the shower they watch movies they read books they go on long walks you know and that's why you know a great screenplay might take a human two or three years but if you're sitting here using you know some verticalized ai maybe you get a really cool insight with the ai helping you oh yeah you know we should make an anti-hero oh no what if it was a female lead oh no what if it was you know this person double crossed them you know and it might

SPEAKER_14: have taken the creative sopranos to have the relationship with the mother and tony soprano and the psychiatrist like those were incredible breakthroughs that he had i don't know how long it took to david chase like to come up with those but you know maybe you you know he could have had seven other ideas for sopranos episodes right um that would be really interesting i would love to watch more sopranos episodes and if they could do it all in ai and give me oh man oh man to get one more

SPEAKER_18: season of sopranos or just to fill in one episode per season just give the ai you know but i don't know how many seasons that lasted but it was it was under 10 i think just give it each season and just say can you give me 10 more minutes per episode of awesomeness and i could watch it with some things i didn't remember but 10 more minutes of backstory oh give me like so good it would just be so like that would fill my bucket i would fill my bucket i would be really happy and that's what i think like you know we're on the cusp of here is like yeah so anyway i give this a i give this a b

SPEAKER_394: plus just for efficiency i'm not even going inside i think it's a hard problem i think it's a very hard

SPEAKER_28: problem yeah i'm a neophyte in this ai being used okay all right last one last one so apple this week released a model uh designed for modifying images all right we had talked about that this was coming yes and so the reason yeah exactly and so the reason i want to show this is we can imagine because we have a bet going on what we're going to get in the next iphone and so basically you know i took a picture and i won't run it here and i just did a simple one remove the background and what this model does is two interesting things it removed the background and it also kind of gives the expressive instruction around what was required there so we've seen things like this before but

SPEAKER_88: apple coming out and putting this out there means it's a little teaser it's a new way to do a teaser

SPEAKER_399: to the breadcrumbs yes of what's coming our way follow the breadcrumbs so apple is doing maggie maggie is their image processing open source uh ai yep you can do things like their model if they're

David Friedberg: building this model and they're doing it that means you're going to be able to go into apple photos on iphone 16 i would predict yeah and say hey for this photo can you remove the background can you

SPEAKER_13: brighten the faces and can you yeah whatever you know uh make it crispier or something or change

SPEAKER_14: give me some ideas of better lighting and that would be built into the iphone and they'll probably have an ai chip or your maggie will be built into your phone so you don't have to go up to the cloud

SPEAKER_18: to do this yeah oh that's what you believe you believe this will run on a phone yeah and that's

SPEAKER_16: part of our bet for whatever it is june 1st exactly right this week in startups.com bets you can look up

SPEAKER_14: our bets but yeah so this is going to be down to the wire we know that it's coming yes just explain to the audience what it means to have a language model slash language model plus chipset that can do

SPEAKER_58: this kind of stuff on a phone what would that mean you know for consumers you know what does that

SPEAKER_24: unlock why is it important explain it to people well i think we're going to get the following i think you can get like sort of unlimited memes and so i think you were trying you were asking about this

SPEAKER_29: in one of our chats the other day and so i just did this right now i said give the man a mustache

SPEAKER_12: yeah so how about give uh him a mustache uh give him a goatee that looks realistic that's not funny

SPEAKER_13: you know like an actual because he gave me like a character mustache but give it a ted lasso mustache

SPEAKER_414: that would be funny give the man a ted lasso mustache yeah this would be interesting if it

SPEAKER_14: knows what ted lasso is yeah here it goes and it's running what is this running on your macbook or in the cloud no this is someone just posted this you know on hugging face exactly yeah it's a hugging face

SPEAKER_58: running this and it's taking looks like about 30 seconds to do it and it gave me not a ted lasso

SPEAKER_28: that is not a ted lasso more work to do um but that's what's cool because they're putting it out there because they're going to get this feedback and that's a different thing from apple and it shows what generative ai is about you can't do everything inside a box you've got to be open you got to embrace open source you got to get feedback from the developers open source is going to win

SPEAKER_58: open source is going to win you heard it here first i think that what this means is like stable diffusion at 10 billion or whatever i don't know what valuation they had but these these models are

SPEAKER_13: going to be open source and if they're open source and remember open ai when closed ai and they're not giving the the code or the weights right the only person who has the code and the weights is microsoft

SPEAKER_425: because they put 10 billion in because of their deal correct because they're dear so microsoft

SPEAKER_18: has the source code i think and the weights yes from open ai that's the reports now meta and apple

SPEAKER_13: two of the magnificent seven are putting all of this out open source google is not open sourcing

SPEAKER_58: their gemini bard stuff right that's proprietary amazon doesn't have a large language model they

SPEAKER_16: might be participating with other folks yep uh so and tesla's model for self-driving is obviously proprietary i don't think that that's open source it can't be so the question is does microsoft

David Friedberg: versus chat gpt beat can chat gpt open ai and microsoft that duopoly can they beat apple and meta open source

Chamath Palihapitiya: your answer no wow wow stunning yep snap snap you snapped it you called you shut all the chips in you believe

SPEAKER_05: meta and apple will beat the mighty microsoft open ai duopoly and look we've seen this movie before

SPEAKER_290: operating systems microsoft you know sun all that seeds to linux right from what's powering almost

SPEAKER_29: everything out there's some some type of linux base right saw it in databases we saw it in you know programming languages we've seen this movie through our yes entire history it's very hard to have a

SPEAKER_14: proprietary platform yeah model beat an open source one now for desktops yeah microsoft windows and

SPEAKER_428: apples apples is built off of open source open source it's like a unix base yeah bsd base but android's

SPEAKER_24: open source right and yeah it seems to me android dominates the world in terms of phone yes you know apple gets all the profit from ios being closed yes i think that was sort of um you can't win it every

SPEAKER_18: time okay yeah you can't expect it to win every time but in this case i think i have to agree with

SPEAKER_13: you because knowing what i know about startups if you're a startup yeah you're going to use the best product available which is chat gpt4 right now but long term you're not going to want to be tied to their whims and they could change the rules on you and so you're going to probably which would be like you know like listen oracle's great oracle databases run like some very sophisticated stuff but i watched as startups used hadoop whatever uh mysql you know there's a whole series of open source database stuff and so you know to see a oh there you go put some glasses on them what kind of glasses

SPEAKER_260: are those i don't know oh we're just gonna put sunglasses on yeah yeah maybe so you know i think

SPEAKER_58: apple doing this is um if you're behind you go open source is what i've always said if you're trying to

SPEAKER_13: catch up you want to create chaos and you go open source so i think apple and meta are behind so they

SPEAKER_58: go open source and they also have lock-in with their brands so it makes total sense what did you

SPEAKER_57: think i didn't see the microsoft copilot ad there was a copilot ad how was it it basically was a

SPEAKER_262: direct competitor to oh my god that is not some ray-ban sunglasses oh you know you have sun

SPEAKER_447: yeah you don't have sunglasses so we're gonna have sun eyes because you should be able to create them

SPEAKER_277: on your phone laser eyes yeah so what we it's going directly after what what is microsoft copilot

SPEAKER_28: going directly after the chat gpt like it's a it's a full experience that you couldn't really

SPEAKER_13: differentiate at all so now i'm gonna have to make a decision am i paying for microsoft copilot

SPEAKER_14: am i paying for claude yeah poe or chat gpt or gemini because now i'm paying for all of them

David Friedberg: correct correct that's where we're going isn't it's so weird that copilot microsoft copilot is built off of chat gpt just playing it here yeah oh here we go yeah yeah i didn't see it let me tell you we

SPEAKER_60: could keep keep talking but like i just don't know here we go or build something yeah yeah whoa whoa whoa

SPEAKER_15: uh they say i'm too old to learn something new yeah too young to change the world yeah so they're

SPEAKER_288: really going after the consumer use case right that's what we're seeing here is there they want

SPEAKER_290: to go after the consumer this is all you know generate images uh yep here we go yeah write my

SPEAKER_193: code for my 3d game wow they're literally just basically because mean organic chemistry so when you see this

SPEAKER_14: did they pull the rug out from chat gpt by buying into it and saying we have the rights to it we're going to make a consumer product and then try to beat them with the consumer because it's the same product with a different wrapper and microsoft has massive distribution chat gpt doesn't correct

SPEAKER_29: i think that's the case and you know the only thing we don't know is you know the deal that they've struck is it for every model going forward or is it just for the existing models is it just for the next one i think what's going to be really interesting in that relationship especially after seeing that ad is how they coexist because before what was being sort of outlined was microsoft was going to do enterprise stuff and and open ai would go after startups plus the consumer

SPEAKER_19: now it's you know game on game on game on and is microsoft charging for copilot is it like 20 bucks

SPEAKER_58: a month right heads up against chat gpt or are they going to just make it free and just i think it's been

SPEAKER_29: free so far i don't know if they have a plan but so far it's been free so if you're willing to go get

SPEAKER_28: a microsoft login and do that it's been free so you don't have to use you know gpt4 yeah that's wild

SPEAKER_13: yeah i mean it's really interesting like as a as a chess board because remember this also when sam altman was ousted as ceo for 15 minutes they were all going to go work at microsoft so satya was going

SPEAKER_19: to hire them all which means satya maybe satya is hiring them all or maybe he's just hiring people

SPEAKER_35: and then he builds up his team and then when he gets the team to a certain level it doesn't matter so maybe like he did this interim deal i put 10 billion into it but i know this is a trillion dollar opportunity yeah so i just get the code base i get the weights and then i just replicate it and

David Friedberg: build something better yeah very fascinating very fascinating times yeah well and then also you just think about bundling right microsoft bundled the browser and we've seen this movie before so they

SPEAKER_14: bundle copilot and it's just like it's everywhere because i got the little copilot on my desktop my windows desktop and then apple if it's just built into the operating system i don't know if apple's

SPEAKER_163: going to just make siri their copilot and then where's amazon and all this with alexa why can't amazon get alexa they i mean shouldn't they just rebuild alexa with a language model and just start

SPEAKER_60: over there was an article a little bit recently that they let go of a huge chunk of the alexa team

SPEAKER_21: right um yeah and i think that's what they must be doing right they must be preparing yeah perfect

SPEAKER_288: example of being too early right siri and alexa were too early and didn't work and they worked in like

SPEAKER_477: constrained environments did you get a vision pro yet apple vision pro yet no you know um i i

SPEAKER_28: listen to you right now on the part pick mine up so i did the same thing and i went to the apple store to buy it which they would let me buy it but they wouldn't let me try it huh and i said well can i just try it and it's kind of like i was a little bit like perturbed yeah like i'm like you want me to spend 3 400 and they're like it was on the saturday and i came out on the friday so they have them

SPEAKER_30: available because they're expensive and i was like i'm willing to do it and you know i i built i think

SPEAKER_334: we should do it i'm getting mine just go buy it if you want i'll pay for yours um so oh yeah sure

SPEAKER_58: i said thank you for you uh i said thank you for you doing this and i don't pay you for this you're getting a lot out of it with your uh but i get more out of it i think so i will pay for you submit

SPEAKER_153: an invoice to us and we will pay for yours but you got to use it and i want to incorporate mondays

SPEAKER_288: will be ai and vision pro demos going forward vision three oh okay four ais one vision pro how

SPEAKER_13: about that is that a good ratio four to one three to two okay but i just want to because i think it's the real deal after hearing friedberg who's like a you know he's a bit of a weirdo you know um i mean

SPEAKER_18: that in a good way you know like freeberg's got a good sense for like weird i think yeah and

SPEAKER_487: they had a good use case like for work yeah see i'm going with freeberg on this one i think freeberg

SPEAKER_13: knows what he's talking about and um i think you and i doing the demos for both of these emerging platforms every monday makes us even more people tell me that mondays are essential listening for

SPEAKER_35: this week in startups and if we do this maybe i mean you're you're gonna wind up being the co-host here because it might be we have to do vision pro one day and ai one day and you're gonna get nothing

SPEAKER_493: done like me you're just doing podcasts constantly i mean yeah yeah we'll pick it up buy it send us

SPEAKER_495: the bill i'll pay for it and then we don't have to you don't have to have the cognitive i was ready

SPEAKER_496: i was ready to buy it i just wanted to demo it i was like let me try it out i know i just don't want

David Friedberg: that don't be you know i get it i get what you're saying but you know it's kind of like we got to make the jump here it's clear that something's going on with this okay there's something going on with it and i think it's a desktop replacement that's what my position has always been is this is a desktop educational replacement i think yeah you know and then let's see if we can do an episode inside of it let's see actually here's what we do this is what we should attempt for next week yeah we do the ai demos inside of the vision pro and we release it so like idiots we wear the we wear the goggles on zoom on our desktops but we do the demos i don't know if there's a facetime

SPEAKER_502: feature or zoom inside of it it renders your face and all that yeah but could we do the demos

David Friedberg: and record the demos inside of our vision pros and then stitch it together so like the demos we did

SPEAKER_83: today you and i looking yeah you know it would be on like a screen or something right yeah it'd probably look terrible for the audience but it'd be interesting we'll give it all right ready we'll

David Friedberg: have to do some we'll give it a try okay so we have to do some prep we'll see you next week bye bye