AI, VC, and the next big opportunity in tech, with 'Startup Project' host Nataraj Sindam

This episode dives into the latest trends and opportunities within AI and venture capital, featuring insights from Nataraj Sindam, host of the Startup Project podcast.

1. Generative AI is reshaping the investment focus, with a significant pivot towards AI-enabled SaaS solutions in early-stage ventures.
2. Seattle is emerging as a crucial hub for pre-seed AI investment, particularly in AI-enabled software services.
3. The competition in foundational AI model development is intense, with only top-tier VC firms positioned to make substantial early bets.
4. There's a growing trend towards AI applications that enhance business operations, moving beyond foundational technologies to practical, market-ready products.
5. Startups have potential opportunities even amidst dominance by tech giants, especially if they innovate in AI application and efficiency.

1. Introduction

Todd Bishop introduces Nataraj Sindam, discussing his roles and the focus of the episode. They touch upon the dynamic shifts in the venture capital landscape influenced by AI technologies. Nataraj Sindam: "AI is a big opportunity, but it's about finding where the puck is going, not where it is."

2. AI's Evolution in the Venture Scene

Sindam elaborates on the evolution of AI from data science to generative AI, highlighting investment trends in AI-enabled SaaS and the strategic focus of venture capital in this area. Nataraj Sindam: "We've moved from calling it machine learning to generative AI because it can generate new things, creating intelligence as a service."

3. Startups vs. Tech Giants

Discussion on whether startups can compete with giants like Google and Microsoft in AI development, with insights into the strategic approaches of emerging companies. Nataraj Sindam: "Startups may not immediately compete with giants in foundational models but have a significant role in AI's application layer."

4. The Future of AI in Business Applications

Exploration of how AI is poised to transform business applications, with specific examples of how AI can streamline operations across various industries. Nataraj Sindam: "AI is not just about the technology; it's about the applications and innovations that drive business forward."

5. Conclusion and Closing Thoughts

Wrap-up discussing the broader implications of AI in technology and business, with Sindam emphasizing the importance of strategic innovation in capturing market opportunities. Nataraj Sindam: "The real competition and opportunity lie in creating AI solutions that enhance productivity and operational efficiency in businesses."

1. Explore AI-Enabled Solutions: Companies should consider integrating AI to streamline operations and enhance productivity.
2. Focus on Niche Markets: Target specific industry needs with tailored AI applications to carve out competitive advantages.
3. Stay Updated on AI Trends: Keep abreast of the latest developments in AI and venture capital to identify emerging opportunities.
4. Invest in AI Talent: Cultivate a team skilled in AI and machine learning to leverage the full potential of technological advancements.
5. Prioritize Practical AI Applications: Develop AI tools that solve real-world problems and offer tangible benefits to users.

As the creator and host of the "Startup Project" podcast, Seattle-area tech senior product manager Nataraj Sindam talks with a wide range of investors and entrepreneurs, contributing to his broad perspective on the future of artificial intelligence, investing, startups, and business technology. He's also author of the Above Average email newsletter, and venture partner with Incisive Ventures.

People

Nataraj Sindam, Todd Bishop

Companies

Microsoft, Incisive VC, OpenAI, Anthropic, Xai

Guest Name(s):

Nataraj Sindam

Content Warnings:

None

Todd Bishop
That's an especially loud plane. So my joke is we need to sell a sponsorship to Kenmore Air. Yeah. Hey, maybe you should, ah, the sound of Kenmore Air. Book your trip to the San Juans, Victoria.

Welcome to Geekwire. From geekwire.com in Seattle, I'm Todd Bishop. We are coming to you from Seattle where we get to report each day on what happens around us in business, technology and innovation. What happens here matters everywhere. And every week on this show we talk about some of the most interesting stories and trends in the news.

I'm pleased to be joined this week by another podcast host, Nataraj Sindham. It's great to have you here. Hey, Todd, thanks for having me. So Nataraj is a senior product manager at Microsoft, but he's also the host of the startup project podcast about the business of technology, the author of the above average email newsletter. I love that name.

And he's also a venture partner with Incisive VC, among other projects that you've got going on. And I just want to mention too, we've talked in the past, and of course, you're here in your individual capacity, not in your role as a Microsoft employee. First off, I wonder if you could just give us a sense for what you're hearing these days from the venture capitalists that you talk with on the startup project podcast. Thanks, Todd. Thanks for the introduction and this amazing opportunity to be here and in person.

Nataraj Sindham
This is a first for me. I've never done a in person podcast recording. Isn't that amazing? Yeah. It is also very different, right.

Because all you have to care is that one little box on your screen and now you have to care about your body language and everything else. Right. So it feels a little different. But yeah, for your question, I think, I mean, the race is on. Like, everyone thinks that AI is a big opportunity, but I think one famous investor, I forgot who said it, that you have to go where the puck is going and not where the puck already is.

Todd Bishop
Yeah, I think it was Wayne Gretzky. Yeah. I don't know who said it, but I remember pretty well. It's sort of adopted by the tech industry for sure. Yeah.

Nataraj Sindham
So I think the point being a lot of people are now focused on AI. I think a lot of SaaS investors are now focused on AI. I think there has also been a rebranding of AI in general. Like, we have gone through different phases of AI. We started with data science, machine learning, and we just call it AI.

Now we are calling it as generative AI because it can generate things. I think in terms of what I'm seeing, I think a lot of AI enabled SaaS is where lots of pre seed and seed stitch investors are investing. Because I think there are only few funds or very top percentile funds who can invest in really long term crazy bets, which requires a lot of early investment, like companies which are talking about going towards AGI and doing really fundamentally unknown things. And those are top tier. And I think in Seattle, we have very few firms which can actually write that sort of checks.

So I think in Seattle, pre seed, what we're really seeing is, okay, now we have this interesting LLM as a technology, and it can generate things, it can reason, it could do things that we could otherwise not do it. So it's like, almost like you have given a new ability of intelligence as a service, I call it. So we have cloud as compute as a service, storage as a service. Now we got intelligence as a service. So this allows you to sort of create some novel applications that were not possible before.

What we're really seeing at pre seed primarily is, I would say, and I think geekware already covered some of these companies, is a enabled software as a service companies is where a lot of precedency dollars are going, apart from those big ideas of foundational companies and data orchestration, you know, lang chain and companies like those which fall into the enablement and tooling and infrastructure side of things. But talking about Seattle, I think we are seeing a lot of AI enabled SaaS ideas which are being funded a lot. Is there any hope in this day and age for a startup to compete on the foundation model development level and the infrastructure level in the world of AI? Or is it more in the realm of the application of the intelligence, as you're saying here? I think there is.

I think there is. I think we sort of often get caught up by the moment, because everyone says we have these seven companies now which are doing foundation models. If you remove Microsoft, Google, Amazon, then we have anthropic cohere, OpenAI, Xai by Elon, and a bunch of other companies which are optimizing LLMs. But I think we have to remember, like, Google is the 16th search engine, you know this better than I do, and we don't know, like who can, which method will optimize compute. Right now, everything is compute intensive.

Like someone can figure out how to run an LLM with less compute more efficiently. So I think when like, even I've heard this thesis of like, you know, every LLM will be commoditized, I think probably in 2025 years, but not immediately. There is this argument of everyone has sort of same capabilities. I think we were just talking about cloud reaching a new step change or improving from its previous version. But I think there is still a long way to go.

We often throw around AGI as an example, but we don't know how to achieve AGI. Like, we can't squint our eye and see how AGI will be formed. I don't know if we have the technology to actually purely form AGI, but we are seeing agentic capabilities which are from generative AI. So people are extrapolating that and saying that if you have a model which reasons and which can find tasks to do and it knows how to do those, then we got good enough technology. I think that's where we can see the future.

Whether that translates into Aji with existing technology or something new has to come up is a whole different question. But I think there is opportunity. I mean, just yesterday I think Ilya Satsky just announced a new company. This is the former OpenAI board member and engineer who really is now going to be targeting this whole idea of artificial general intelligence and this superintelligence. And the way that he announced this startup was interesting.

Todd Bishop
This is not going to be a minimum viable product. I mean, this is like the opposite of MVP. It's like the maximum viable product. Like we're not going to hear from them until they achieve AGI. Yeah, I mean, it's.

Nataraj Sindham
It's almost like what Google had inside its incubator, right, where they didn't really pursue things as, you know, a product, but they just went on pure research mode, I think. Yeah, I think SSI, which it's called, I think is also trying to be that. It's also, for me, the interesting part was, will the best talent go towards an idea like this versus a company where the stock vests into millions of dollars? And it's in short term. So I think it's going to be an interesting human experiment that we are running.

And I'm glad that we have these multiple companies competing with each other. And we see like there are absolutists who wants to do just pure research. The people who want to create a product, iterate on it, the traditional version of product development, and see what is good, what is bad. We understand, learn and come out of it. And big tech companies sort of did both, right?

They have their main business, which are making money. And then we had these incubated projects, like Google had their incubator, Microsoft has some incubators inside it, and Amazon, famous for creating multiple big projects all the time. So I think this is an interesting example between OpenAI and anthropic and SSI, that where is the human incentive going towards? Well, to your point, Ilya Sutzkever was explaining that they are not going to have a problem raising capital. They will face many challenges in the course of their startup journey.

Todd Bishop
But the big ones are technological, perhaps organizational, maybe scientific. But the fact that he is coming into this with such a reputation really helps. And if you think about it, too, anthropic had a very similar journey where the founders there left OpenAI, started their own thing. They've raised $4 billion from Amazon alone. So in some ways, this answers the question I was asking, where it's like, can, can startups compete with the tech giants in the development of these foundation models?

And in some ways, the answer seems to be yes. Although those startups have to have this pedigree that makes investors go, wow, you are the next OpenAI. You are the next. Well, anthropic in the future, right? Yeah, I mean, pretty much.

Nataraj Sindham
You can trace back a lot of AI companies now back to OpenAI. It's almost like OpenAI alumni which is creating all these companies. Right? And there's also, like this famous anecdote where Elon convinced Ilya to come and join from Google to start OpenAI. So it was him who convinced Ilya to actually move from there in the first place.

But I'm a little bit skeptical of this idea of being a complete nonprofit. I think, you know, if you look at an average PhD graduate, is he motivated by doing lifelong research, or is he motivated by stock options that will vest in a couple of years? Probably the latter. 90% are probably the latter, right? I'm sure, like, there are folks like Ilya who want to join from the mission, but I think the practicality is 90% will, you know, go the other way.

And there's also this problem of, if you're a PhD, you want to optimize your career, right? Then what is the best place to join? Like, even if you join Xai, which is valued at 16 billion, it's a bad choice for you to make, even though a lot of good talent is going there, right? Even though it's backed by Elon, technically it's a bad choice for you because you're going at a very, very high valuation. So if you are a PhD, really good PhD, however, base salary is higher.

And how much ever stock you're getting, you're getting, you're still undervalued because it's already priced at 16 billion. Unless this is the only company which should dominate the future, which, I mean, Elon has created multiple companies, so it could be. But as an employee, you have to make your own clever decision. So you're better off joining coherent or anthropic at a lower valuation. You'll probably have a better shot at an exit, right?

If you just think about it logically, where is the bigger chance of exit for you? So I think those will eventually play out. If not now, sometimes. All right, I want to take a quick break here, and when we come back, let's talk about the implications of this boom in AI for the cloud and some of the fundamental underpinnings of the cloud infrastructure that a lot of these startups are using. You're listening to Geekwire and we will be right back.

Speaker C
I wanted a career in it, but I didn't know where to start. WGU makes it simple. Their accredited online degree programs cover all kinds of IT specialties, and they have valuable industry certifications built in at no extra cost. The payoff? Having those certs back up my degree makes me look even better to future employers.

A nonprofit university that includes top industry certs in their programs. I choose WGU. Learn more at wgu.edu. itsertsincluded.

Todd Bishop
Welcome back. My guest this week is Nataraj Sindham. He is host of the Startup Project podcast about the business of technology, author of the above average email newsletter, and a venture partner with Incisive VC. I know one of your theses maybe not in a formal investment sense, but one of the ideas that you have about AI is that there's going to be some fundamental shifts, or we're already seeing some fundamental shifts in cloud infrastructure. What kinds of things are you seeing there?

Nataraj Sindham
I think with generative AI or AI in general, there are a lot of second order effects that we are seeing which haven't fully played out yet, but the one that is playing out aggressively is how the cloud is changing. I think it's becoming to a point that maybe in next five years we will ten x our data center size. And I'm not talking about Microsoft, but just if you look at how much Elon wants to invest, he wants to create a gigafactory of chips. I think you can easily correlate how much compute we need if you really want to achieve things that are quite achievable from the current capability. Even if we didn't unlock any new technology, transformative technology beyond LLMs, there's a huge need of compute.

So I think it's easily going to be that we'll ten x our data centers because of just existing use. And on top of that we want to create this intelligence layer. If you look at the capital spending among a lot of the big tech companies, it's certainly on that trajectory right now. Yeah, so I think that's one of the sort of second artifacts. But I think fundamentally, I think AI is giving cloud a new momentum.

It's sort of changing. What are the requirements the cloud has to do from storage to compute, which are two important parts of consuming cloud. For example, in storage, unstructured, where I work a lot, everything in the world is just unstructured data. And unstructured data is basically your files, your documents of all types of form, anything that is not basically a database. Storage is unstructured and pretty much everything, most of the data is actually in the form of unstructured data.

That means your appetite to store and consume from unstructured data is basically going way up because of this fundamental shift. And you see that with companies like unstructured, which is aptly named, they are enabling how to use this unstructured data. Because if you look at where most of the data that we use for day to day work is, it's an unstructured format, it's with the companies that are actually owning that data, it's not the data on the Internet. So that's why working with unstructured data becomes hugely important. And I think the battle also is slowly starting to shift from the chat GPT thing to a b two B enablement.

Because it's sort of like if AI really has to change how we live, it has to impact what we do. And we spend eight to 10 hours a day on what we do. And all the data is actually b two B. It's not really on the Internet, it's all private data, it's in cloud mainly, and even cloud. If you look at latest Ghana reports, they say most of the data is actually in private data centers.

85% of all's data is still not in public cloud. So you couple that with the appetite what AI needs, then you can go to that 1st 2nd order effect of we will ten x the need for data centers and I think you'll also ten x for compute and storage and unstructured specifically. And that means in general we will switch to the competition actually between anthropic and cohere. And OpenAI to a lot of extent is not actually about chat GPD. I think chat GPD is just one form factor and consumer form factor and maybe that's where like search gets disrupted.

But there's a world beyond search, right? There's 8 hours to 10 hours of our day where we spend our time, which is work. Work gets changed by whoever creates the AI platform for our eight to 10 hours of a day. So I think the comp, the real competition really is getting kicked in there and it will be a very slow cycle because Chajubbid is consumer, everyone is testing out consumer and everyone wants to comment on how this answer is better than that. If you look at real dollars eventually I think b two B is where the competition is headed and that gets.

Todd Bishop
Into the skating to where the puck is going, not to where it is. Are there startup opportunities there, are there vc opportunities in what you're saying? Definitely. I think we have seen some investments like for example Jasper AI came out a while back, but I want to give that example because Jasper AI was specifically targeting marketing as a job. There are companies which are now targeting marketing.

Nataraj Sindham
But what if you can create AI enabled marketing campaigns? Having a good marketing campaign is not straightforward. Like if you go to Facebook's marketing portal and try to set up a campaign it takes a while and some iterations to get the right fit for the thing that you're actually trying to market. But if you have some AI trying to optimize it, iterate it and you just give it instructions. So if my dataj job is a marketer who runs Facebook marketing, or if I'm an ad agency which does Facebook marketing for my clients, now I have this new tool which makes, you know, my life whole lot of easier.

So from Jasper to you can go there and there are a couple of companies, I think one started in Seattle but they moved to SF, it's called validator AI and they're sort of attempting at this. And then now you extrapolate this to other roles like if you're a financial analyst, how much of your work is now crunching excel, going and finding and sourcing data? And it's all little, little small tasks. It's almost like a death of thousand, death by thousand cuts. Your day is basically cut across 100 Google tabs, Google Chrome tabs in your day.

But all those tasks are now can be achieved through agent plus extracting insight or extracting tasks from your instruction. So that is possible, it's not as good but it is possible. So that means you can really create a tool, workflow tool for finance professional. Now you go to legal, I don't need a lawyer always. But I have so many legal questions, whether I'm navigating immigration or if I'm a HR in a company, I don't have to wait three days to get a response from my legal team.

There might be some questions that are about the policy of my company that I want to know now that could be addressed by sort of like a Microsoft copilot, or it could be addressed by a legal specific tool. And that can also help lawyers by outsourcing to their paralegals doing more work. So you'll see, almost like, I think Lexion got funded. But there are a lot of companies doing this in this space. There are companies just targeting EB one IA applications, for example, or companies just trying to provide immigration guidance using AI.

So as you extrapolate each role, you can create new sort of workflows, essentially like if I'm a designer or even if I like podcast. And I think we talked this about, you know, in our podcast episode when we did. No one has fully fledged out the podcast workflow, which pretty much everything can now be automated. But right now I do it through using individual tools. But there's no reason why it all has to be an individual tool.

You can almost create a workflow based tool, which pretty much does everything for you by, on an instruction basis. So I think we're going to see a lot more applications. We haven't seen any application innovation right now using AI. We have seen just chat interface, general. Purpose, general purpose, generative AI.

Todd Bishop
And to your point, like I'll jump in sometimes and I'll try making a GPT for something and I'll upload some knowledge through a PDF. And more often than not, the incremental value that I get from even that, versus just the general chat GPT or Microsoft copilot or anthropic, there's not that much more incremental value in that. So there's an opportunity for somebody to say, okay, let's go beyond the GPT, the sort of on the fly app, and make something truly created by a software development team for this industry. Vertical for this application. Exactly right.

Nataraj Sindham
It's almost like instead of creating 1000 GPTs, you'll have ten vertical apps for each role. The thousand GPT scenario, it doesn't seem like where we are going. Maybe it's almost like in any app industry, any app category, the top five dominate the category, even in Apple Store or Google Store. And then there's a huge long tail, but each one has significantly lower use case or audience, same like media so I think we are just seeing that about to happen even with AI also. And remember, like even just as simple as like task management, we have like 20 software companies, I think, like from Monday to workday to Asana Trello.

How many companies are there? Just today I was telling Taylor Soper, Geekwire's editor, I was saying, hey, wait a second, Slack has a new Trello style app. Like we're spending, you know, $800 a year on Trello. Like should we be doing this in slack anyway? That's a rabbit hole.

Yeah. And you can do the same thing on notion, which also allows you to do this. So like just for that aspect, you have so many applications. That means if you now put, okay, if you can do AI enabled abstraction layer and come up with a new app for either task management or just marketing or just legal or finance, now you can imagine a whole number of applications that haven't been built yet. I think we only saw two innovations.

One is chat and summarize generate and that, and then I think Microsoft came up with recall, which is like something we haven't seen before, right, and haven't. Actually seen it yet. But that's a whole other topic because they pulled it back from the initial launch of the Copilot Plus PCs, which was another news item this past week. Yeah, but there is no real new product consumer or b two b yet other than copilot or just chat GPT. So I think that's where real use cases come.

Todd Bishop
Well, this gets into something that I want to bring back, which is a segment that I've been calling my AI as you talk about use cases. So I want to do that when we come back. You're listening to Geekwire and we'll be right back.

Welcome back. It's Todd Bishop. My guest this week is Nataraj Sindham. He is the host of the startup project podcast. I highly recommend it.

You've got this really super low key approach and it's really interesting to hear when you interview folks, you bring out a lot of interesting insights from these VC's. You had a great one with one of the principals of breakwater VC talking about the missing middle layer in the Seattle venture capital market. And so I just highly recommend that people check out the startup project podcast. I'll link to it from the show notes thestartupproject IO. So check that out.

Okay. My colleague John Cook and I always joke that we need a theme song for this. And maybe one of these times I can use an AI music generator to create a theme song. And John always does a little ditty on my AI, but I'll spare everybody out there. But I love this idea of these practical applications, even just the small nitty gritty things that make a difference in your workflow from just the general purpose tools that are out there.

And I found one recently. It happens to be from your day job. Not that you're involved in it, but it's your employer in your day job, and it's one that you can use with Microsoft Copilot in Edge for free. I'm not on the paid version of Copilot anymore, at least, and it's something I love to use on YouTube. So I'm on my desktop, I go to a YouTube video in Edge, I open the copilot sidebar and where typically it says generate page summary.

That's kind of the default if you're on YouTube now it says generate video highlights. This thing is so slick and I was actually using it to help prepare for this interview with you, this podcast with you, because I was going back to some of your past startup project podcasts on YouTube that I hadn't seen you hit generate video highlights and it just almost instantaneously says, okay, here's what happens in this segment. You know, at minute three, this is the discussion, three discussions points, you know, and it just goes right down and gives you basically an outline and you can click on that segment and go straight to that part of the video. And I recognize you can do this in part with the automated transcript in YouTube, but it's not quite the same. This is clearly, yeah, a case where an application of generative AI is superior to something that's just more raw data.

Nataraj Sindham
Yeah. One other hack is, I mean, obviously we use teams, but teams, the way it summarizes when you don't attend a meeting is actually very, very good. It actually tells you who talked about what and what's the summary, what's the takeaway? It's a very well done feature. I feel like it's also a shortcut if you have not attended a meeting.

Just looking at teams summary is pretty awesome. It's funny, so I haven't used that, but I will give it a shot. We end up using a lot of Zoom. We're a Google apps shop at Geekwire, so we end up using Google Meet. I incorrectly called it hangouts for somebody earlier today.

That's Google's faults, not yours. Exactly. It's funny, though, this is a totally different topic, but I keep trying to get teams to stop adding itself to my startup items every time I do a Windows update. This is totally not your realm. That's just my rant.

Todd Bishop
It's like, I don't know why teams keeps getting checked as a box for my Windows startup items. I'm a Windows user. I'm not a Big Mac person. But anyway, in the realm of I'm. A Mac never person.

Never, never. Yeah, I mean, my wife uses it. The reason I why don't encourage people to use Mac, I mean, not because I'm working in Microsoft. This was my opinion even before I was working at Microsoft, is because it's objectively a costly and pricey lifestyle. Once you get into that lifestyle, it's a very, very pricey one.

Nataraj Sindham
If you want to go cheap, you go into the Windows Android space. You have all kinds of range. You can buy the high end, you can buy the low end. Yes, but the problem with Mac is a college graduate who doesn't have any money and has to do student loans will have, will end up in this circle of paying humongous amounts for dongles and sort of some ridiculous upgrades apple will give. Apple always tends to price.

They'll give you 100 gb at 999, but 100 gb is not enough. So you want like 200 gb, but there will not be 200 gb. There will be 500 gb at $2,000. So it's some ridiculous pricing like that. So I'm with you to an extent.

Todd Bishop
My problem with PCs, just the hardware themselves, is you can really subject yourself to an awful experience if you go cheap. At least with Apple, your chances of that are very slim. I am a real aficionado of the ThinkPad lineup. I mean, that's what I've got here right in front of me. And the Carbon X one from Costco, I love getting, boy, they're not a sponsor.

Nataraj Sindham
Not the Kaskos episode. But seriously, like, just the ability to know that if something goes wrong, it's gonna be easy to return. And they've typically got a pretty nice, in the realm of $1,500, 1300 bucks carbon x, one with 500 gigabyte or more, sometimes terabyte hard drives that they've got in their lineup. And I tend to go for those. And there's also, like, another sort of subtle point here, which is like the Windows Android world enabled a lot more people in the world than what Mac enabled, even though in the modern world, Mac sort of dominates and everyone uses Mac.

But if you look at how many people it sort of enabled because of being cheap. Like, if you look at Android, for example, if you go outside us and outside Europe, everything is Android. So in terms of world impact, the fact that it's cheaper, open, I think that sort of enabled more people in the world to use technology and access technology and get benefit out of it. So that actually is a bigger reason. So these are two big reasons.

One is it's a costly lifestyle once you start engaging in it, and then it's like it enabled more people. I can tell our podcast producer and editor, Kurt Milton, who's sitting here and is an avid Mac user, I'm just glad he doesn't have a microphone. You're so lucky.

Todd Bishop
Just stay over there, Kurt.

So along this theme of my AI and useful everyday AI tools, is there anything that you've been using lately that's been especially helpful for you in your work? Not for work, but in general for my personal use cases? Gemini is very interesting, especially the, I think mainly because of the 1 million context, because it sort of automatically fine tunes on your data and you don't have to do additional fine tuning steps. It opens up use cases that were otherwise not open until you write code. I mean, you could do fine tuning, but you have to write code to do it, and you have to get API access and token access and do that.

Nataraj Sindham
But the 1 million context is pretty interesting, and I think we can expect other models to open up to 1 million, 2 million contexts pretty soon. So my understanding with this, and when you're talking about the 1 million context window, that's 1 million tokens. And the idea is that it essentially increases the amount of short term memory that the system has when you're making a query. There'll be times when I'm using some generative AI tools and it's like, wait a second, I just told you that five minutes ago. Why don't you remember that?

Todd Bishop
I think that extended short term memory can be a real benefit if you're trying to do a complex query. Yeah, and it's also like, for example, like I'm doing a podcast, right? I want to sort of, let's say research a guest and let's say they did ten podcasts. I want to take all those ten transcripts, upload and say, find interesting questions for me to ask this guest that have not been asked or something like that. Like if you want to play around with a huge set of data, then the 1 million context matters because you can feed all that context right in the window and just ask a question on it.

Nataraj Sindham
And it's pretty good and it's very well fine tuned. I don't know what technique Google is using, but it's better than when we fine tune. I've done fine tuning with different stacks and with different vector databases. It is still better because obviously I'm not at the cutting edge of using fine tuning in the way they are using it. So I think that's very interesting because it opened up a lot more use cases.

I think that is really fascinating. It's more about how many use cases it opened up and less about quality, I would say. Okay, I want to circle back to apple real fast. You said something in one of your tweets recently. I know you joke that, like we could have fodder for days from your tweets, but this one caught my attention because I think I actually disagree with it a little bit.

Todd Bishop
Tech media in the US over indexes on Apple and underindexes on everything else. Maybe Elon. Elon's an exception for that. Elon is an exception, probably like everyone is over indexed on Elon and what he does. That's true.

And it doesn't hurt that he's got a social media platform that seems to be a little bit skewed toward him in terms of what it surfaces for him. Yeah, that's also true. But I think in general, coverage about Elon is never ending. It's almost like every day. Obviously he's important person and has to be covered, but I don't know.

Nataraj Sindham
I think tech media wants to cover it because of clicks. Yes, I think there is an element of that. Here's my take on some of the Apple coverage, and I don't do a lot of it, in part because they're not here. They have an engineering center here that actually is focused heavily on AI and machine learning. So actually, maybe we should be covering it more.

Todd Bishop
But so much of that is just table stake Silicon Valley coverage. And to me, the reason why it makes sense for the tech media writ large to overindex on Apple is with their latest AI announcements. When the new version of the iPhone comes out, that's probably going to be the best chance for my mom to be using AI in an everyday way. And the general pervasiveness of Apple technology means that even if they're not first, that oftentimes what they do popularizes technology. So in that way, I can see where there's some justification to the fixation that a lot of the tech media have with Apple.

Nataraj Sindham
So even both in Elon and Apple, they're very important topics to cover. So you can expect them to be covered more than because if us population is majority using Apple devices, then obviously it makes sense to cover them more. And you sort of expect it to be covered more because the tech media is in their own filter bubble. They tend to sort of overcover Apple. They tend to not criticize Apple's features as they should be.

That if, let's say, when Facebook snaps, takes a feature from Snapchat and copies it, they sort of go all the way and they'll cover it endlessly. The fact that Facebook is a copycat. Yeah, like when Instagram launches stories or something like that, they can't stop talking about it. But when Apple does it, when Apple pretty much has been doing that for past seven or eight years from Android, they tend to not do it because they tend not to see it because they're not invested in Android, they're not using two mobile phones. If I was a tech media journalist, I would be using two mobile phones, one for Android, one for Apple, and switch between them if I was on the beat of mobile coverage.

Of course. Yes. Our next episode will be our podcast producer and editor, Kurt Milton, just ranting for an hour on everything that we're saying. I can see him turning red over here. No, there's another classic example here is the criticism Google got that Google is late to AI is partly because tech media didn't cover Google I O events, especially the Android part of features as well as they should be, or they didn't understand them because the best AI, I say, is not visible.

And Google's mobile phones had better AI than what even Apple launched last week, and sort of tech media sort of ignored it, and which created an interesting opportunity in the market itself, like undervaluing Google in a lot of ways. I mean, it doesn't make Google scot free in here. Like, Google obviously makes a lot of mistakes in their strategy, like, when it comes to AI or in general products. You said it has hangouts. We know how many chat products Google launched and ended.

There's like ten products probably in chat alone, that we can talk about. Google is ending its Google pay product now, and it has Google Wallet. So the nightmare continues with Google products. But having said that, I think tech media has a bias towards Apple in the US specifically, or even in the west coast, probably the big ones. This is interesting because I've been covering Microsoft for 20 years, since 2002, more than 20 years.

Todd Bishop
And at the beginning, I kind of all felt bad for Steve Ballmer and some of the executives of that era, because when they went to media events, when they were presenting to the assembled press, corps. They would look out and see a sea of Apple logos. And not that I care, but it was like, it was like, it was a little bit of a cringe worthy moment. I will say the advent of Microsoft Surface changed that. It's much more of a mix now.

I see a lot of Windows PCs in the mix, whereas before it was purely Mac. So everything's a little different and it depends on the context. But it seems like there's at least a little more diversity of tech use among the mainstream media that I see, and perhaps the trade press too. Some of the YouTube sort of reviewers like MKBHD and those guys, they're often switching between Android and Mac to stay sort of objective. So I think that's a good practice for them.

Nataraj Sindham
If you're like in a professional reviewing job role, I think it sort of comes with the territory. Yeah, we'll have to have you back just for a smartphone gadget consumer tech conversation because this is good. Well, thank you. This is great. I highly recommend that people check out the startup project podcast.

Todd Bishop
I'll link to it from the show notes. My guest this week has been Nataraj Sindham. He is a senior product manager at Microsoft. In his day job, host of the startup project podcast, author of the above average email newsletter, and venture partner with incisive vc. Thank you very much for joining me.

Nataraj Sindham
Thanks, Todd. Thanks for having me. And I just want to reiterate one thing, that all these opinions are just mine and it's not Microsoft. I'm not talking as a Microsoft representative. Noted.

Todd Bishop
And good. I'm glad you clarified that. Yeah, thank you for listening. Kurt Milton edited this episode. I'm Geekwire co founder Todd Bishop.

We'll be back next week with a new episode of the Geekwire podcast.