In an episode of fascinating topics and exploration of evolving models, The AI Show hosts Paul Roetzer and Mike Kaput explore the rise of AI Emergent companies like Moderna and Asana, Microsoft’s introduction of the Phi-3 family of small language models, and the formation of the US government’s AI Safety and Security Board featuring tech leaders like Sam Altman, Satya Nadella, and Sundar Pichai. Our rapid fire section covers news from Eric Schmidt-backed Augment, Expansion of Meta Ray-Ban Smart Glasses Collection, Hubspot AI Updates and more.
Listen or watch below—and see below for show notes and the transcript.
Today’s episode is brought to you by rasa.io.
Rasa.io is the ultimate platform for AI-powered newsletters. If you’re looking to transform your email newsletter into a powerful, engaging tool that truly resonates with your audience, rasa.io is the game-changer you need. Join the 500+ organizations already making their newsletters smart.
Visit rasa.io/maii today.
Listen Now
Watch the Video
Timestamps
00:04:24 — AI Emergent Case Studies
00:20:20 — Microsoft Phi 3
00:29:19 — OpenAI’s Sam Altman and Other Tech Leaders to Serve on AI Safety Board
00:30:34 — OpenAI CEO Sam Altman talks AI development and society
00:35:04 — TSMC’s Debacle in the American Desert
00:39:53 — Expansion of Meta Ray-Ban Smart Glasses Collection
00:43:13 — Eric Schmidt-backed Augment launches out of stealth with $252M
00:45:59 — Musk’s xAI Is Close to Raising $6 Billion from Sequoia, Others
00:48:16 — Apple Intensifies Talks With OpenAI for iPhone Generative AI Features
00:50:33 — HubSpot AI Updates
Summary
AI Emergent Case Studies
We have new case studies detailing companies becoming AI Emergent by reinventing their businesses with AI. Moderna partnered with OpenAI to deploy ChatGPT Enterprise to thousands of employees, empowering every function with AI and creating novel use cases and GPTs that accelerate and expand each team’s impact.
Within two months, Moderna had 750 GPTs across the company, with 40% of weekly active users creating GPTs and each user having 120 ChatGPT conversations per week on average.
This spans hundreds of use cases, including GPTs for reviewing clinical data, summarizing contracts, and providing quick answers about internal policies.
Asana also integrated AI into every aspect of its operations, as detailed by co-founder and CEO Dustin Moskovitz. The company built bots for feedback, reviews, sales, customer experience, and content creation.
Like Moderna, Asana empowered employees to find their own use cases and build tools tailored to their specific roles and needs.
Microsoft Phi 3
Microsoft has announced the Phi-3 family of open models, which are noteworthy because they’re small language models, or SLMs.
According to Microsoft, SLMs “offer many of the same capabilities found in LLMs but are smaller in size and are trained on smaller amounts of data.”
Because they’re much smaller than LLMs, SLMs consume far less compute and, because of that size, can run locally on devices like an iPhone.
Phi-3-mini, one of the main models in the family, has 3.8 billion parameters, was trained on 3.3 trillion tokens, and performs better than models twice its size, according to Microsoft. Other Phi-3 models with more parameters (7 billion and 14 billion) are also coming soon.
While LLMs remain unrivaled for the most complex tasks, SLMs can perform well on simpler tasks and are more accessible and easier to use for organizations with limited resources. They can also be more easily fine-tuned to specific needs.
“What we’re going to start to see is not a shift from large to small, but a shift from a singular category of models to a portfolio of models where customers get the ability to make a decision on what is the best model for their scenario,” said Sonali Yadav, principal product manager for Generative AI at Microsoft.
OpenAI’s Sam Altman and Other Tech Leaders to Serve on AI Safety Board
Sam Altman, Microsoft CEO Satya Nadella, and Alphabet CEO Sundar Pichai are joining the AI Safety and Security Board run by the US government.
The board will be advising the Department of Homeland Security on how it can safety deploy AI within critical national infrastructure.
The board will also make recommendations on how operators of core infrastructure like power grids can protect their system against AI threats.
Almost two dozen political, tech and business leaders in total are on the board. Other notable names include Nvidia’s Jensen Huang, Anthropic CEO Dario Amodai, the CEO of Northrop Grumman, the mayor of Seattle and chair of the US Conference of Mayors’ Tech and Innovation Committee, the governor of Maryland, and noted AI research Fei-Fei Li.
Links Referenced in the Show
- AI Emergent Case Studies
- Microsoft Phi 3
- OpenAI’s Sam Altman and Other Tech Leaders to Serve on AI Safety Board
- OpenAI CEO Sam Altman talks AI development and society
- Taiwan Semiconductor Manufacturing Company Obstacles in the United States
- Expansion of Meta Ray-Ban Smart Glasses Collection
- Eric Schmidt-backed Augment, a GitHub Copilot rival, launches out of stealth with $252M
- Musk’s xAI Is Close to Raising $6 Billion to develop Grok
- Apple Intensifies Talks With OpenAI for iPhone Generative AI Features
- HubSpot AI Updates
Read the Transcription
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: You can be building your AI councils, you can be trying to lead within your department. But the companies that probably win here, it’s going to come from the C you’re going to have support from on high and you’re going to have a leader who has a
[00:00:12] Paul Roetzer: vision to transform the organization over the next five years, because I think that’s probably the window of opportunity for most industries. You got three to five years to figure this out, basically.
[00:00:23] Paul Roetzer: Welcome to the Artificial Intelligence Show, the podcast that helps your business grow smarter by making AI approachable and actionable. My name is Paul Roetzer. I’m the founder and CEO of Marketing AI Institute, and I’m your host. Each week, I’m joined by my co host, and Marketing AI Institute Chief Content Officer, Mike Kaput, as we break down all the AI news that matters and give you insights and perspectives that you can use to advance your company and your career.
[00:00:53] Paul Roetzer: Join us as we accelerate AI literacy for [00:01:00] all. Welcome to episode 95 of the Artificial Intelligence Show. I’m your host, Paul Roetzer, along with my co host, Mike Kaput.
[00:01:08] Paul Roetzer: are coming to you early this week. It is 7:20 a. m. Eastern time on Monday, April 29th in our world.
[00:01:15] Paul Roetzer: So, yeah, always the disclaimer if anything crazy happens on Monday that doesn’t make the show, that why. yeah. So it, you know, it’s interesting week, like it didn’t seem like a crazy news week, not a lot of breaking stuff, no, you know, major new tech emerging that we saw, no major funding announcements.
[00:01:34] Paul Roetzer: And yet there’s some really fascinating topics to cover that I think. continue to sort of set the stage about where we are and where we’re going, both terms of like regulations, some evolving models.
[00:01:46] Paul Roetzer: And, you know, the first big thing we’ll talk about today, Mike, is the, you know, a couple of examples of these AI emergent companies.
[00:01:52] Paul Roetzer: And I think this is the kind of stuff like I’ve been anxious to be able to to on the podcast, like the real applications of [00:02:00] companies that are actually Doing this the right way, like full blown adoption across the enterprise.
[00:02:05] Paul Roetzer: So, again, like no major breaking news yet, unless something happens we’re talking here, but, I think some really valuable stuff for people today, that can help you along your, your path.
[00:02:18] Paul Roetzer: So today’s episode is brought to us again by Rasa. io. Rasa. io is the ultimate game changer for AI powered newsletters. Rasa. io’s smart newsletter platform tailors your newsletter content for each and every
[00:02:31] Paul Roetzer: subscriber and automates tedious
[00:02:33] Paul Roetzer: newsletter production
[00:02:34] Paul Roetzer: tasks. We’ve known the team at Rasa. io for years now and think their
[00:02:39] Paul Roetzer: is well worth checking out.
[00:02:40] Paul Roetzer: Join the 500 plus organizations leveraging Rasa. io and get a personalized demo today at rasa. io slash m a i i. And again, I’ve mentioned this on one of the previous throughs on Rasa. io, but like Mike and I use it
[00:02:54] Paul Roetzer: as an internal newsletter. So. we don’t run our institute newsletter through it, but we actually
[00:02:59] Paul Roetzer: find it really [00:03:00] helpful just for internal research purposes, because it’ll send us, you
[00:03:03] Paul Roetzer: know, links look at and things like that. So it’s one of the sources that we use to actually curate the weekly news. So, is another potential use case you can think
[00:03:11] Paul Roetzer: about for smart newsletters like that. and then the second is, we’ve been mentioning this, our 2024 State of Marketing AI Survey is in the field
[00:03:20] Paul Roetzer: now. You can be a part of that research, this is the, what did we say, Mike? Fourth
[00:03:24] Mike Kaput: This is actually year four. I think I misspoke last Yeah.
[00:03:28] Paul Roetzer: Okay. So year four of this research. So we’ve got fascinating research going back, four years now, where we take a deep dive into what is actually going on in, artificial intelligence within the marketing industry. Look at use cases,
[00:03:42] Paul Roetzer: uh,
[00:03:42] Paul Roetzer: you know, how people are thinking about it, obstacles for adoption within enterprise. We’re asking questions this year around, do you have generative AI policies? Do you
[00:03:50] Paul Roetzer: responsible AI principles? Really trying to get deeply into that.The understanding of
[00:03:54] Paul Roetzer: where the market is right now with AI adoption. So we’d be appreciative if you have a [00:04:00] few minutes and can be a of that
[00:04:01] Paul Roetzer: survey. It is stateofmarketingai. com. you can go there, you can download the report and click the link at the top that says 2024 survey. And, And be a part of that research.
[00:04:14] Paul Roetzer: will be releasing that in summer of 2024, so coming up in a few months. So again,
[00:04:19] Paul Roetzer: stateofmarketingai. of marketing ai.com to be a part of that research. Alright Mike, it’s all you.
[00:04:24] AI Emergent Case Studies
[00:04:24] Mike Kaput: Alright so you alluded to our first big topic today, which is we’re seeing a couple new case studies come that detail companies becoming what we would call AI emergent, or these are existing firms that are reinventing their businesses with ai.
[00:04:42] Mike Kaput: The first case study details a partnership between Moderna and OpenAI to deploy ChatGPT Enterprise to thousands of employees across that company.
[00:04:53] Mike Kaput: According to OpenAI, quote, Now every function is empowered with aI, creating novel use [00:05:00] cases and GPTs that accelerate and expand the impact of every team
[00:05:05] Mike Kaput: They also say that within two months of ChatGPT adoption, Moderna had 750 custom GPTs they had created across the company. 40 percent of weekly active users were creating GPTs
[00:05:22] Mike Kaput: and each user had on average 120 ChatGPT conversations per day. per week. Now, within Moderna, this is happening across literally hundreds of possible use cases across the business. They’ve got a GPT to review clinical data using some of the data analysis capabilities that ChatGPT has.
[00:05:44] Mike Kaput: They have that their legal team uses to summarize contracts, and they have GPTs that help employees get quick answers about internal policies. at the company. Now, as we heard this, we also got an [00:06:00] article that contained a detailed breakdown from Asana co
[00:06:03] Mike Kaput: founder cEO, Dustin, Dustin Moskowitz, how the, about how the company evolved to integrate AI into every aspect of their operations. That included building bots for feedback, for reviews. for sales and customer
[00:06:19] Mike Kaput: experience, and for content creation. So it sounds like Asana also took
[00:06:24] Mike Kaput: kind of similar approach to Moderna despite them being very different types of companies in the sense that they empowered employees to go find their own use cases and build their own tools for their specific roles and needs. Moskovitz even details how everyone from sDRs to product marketers to HR professionals were able to actually identify high value AI use cases
[00:06:48] Mike Kaput: the company and build AI tools themselves support their work. And he noted multiple times, they often didn’t really have significant technical backgrounds. [00:07:00] So Paul, this caught our attention for a few reasons, but first up, can you kind of maybe talk a
[00:07:04] Mike Kaput: a little bit more about what we mean when we say AI emergent companies why it’s so important to be looking at these types of case studies?
[00:07:12] Paul Roetzer: Yeah, I’m so excited to see this case study. I mean, I get, like, there, it’s a case study from OpenAI, it’s going to be obviously very favorable, it’s certainly, you know, adoption and enterprises
[00:07:22] Paul Roetzer: isn’t this seamless, but, you know, the whole point here is just to see the depth of commitment that was made to the case study.
[00:07:29] Paul Roetzer: To integrate ai, and think it’s a great representation of what organizations should be thinking about and pursuing. So going back to the AI emergent concept. So I wrote a blog post, we’ll link to it in the notes, and I, I had to go back and look when I wrote this. It was May 16th, 2022. So set the stage two, may, May, 2022.
[00:07:50] Paul Roetzer: Midjourney was couple months old. We had, DALL-E had been previewed. I don’t think it was readily available yet. Mike and I were finishing the manuscript for our book. [00:08:00] So
[00:08:00] Paul Roetzer: our book came out in summer of 22, the artificial and marketing artificial intelligence book. and we were, what about seven months prior to ChatGPT being introduced? So that’s kind of the stage were
[00:08:14] Paul Roetzer: at, but we were already seeing at that point, the inflection point was arriving in our opinion as to like, we really were going to have future. Whereas I wrote.
[00:08:24] Paul Roetzer: You are going to be AI or obsolete. and so I’ll just like pull a few excerpts
[00:08:29] Paul Roetzer: from that post because I think it’s very relevant and what we’re now seeing is kind of these companies we envisioned are now coming to life.
[00:08:35] Paul Roetzer: So in that post, I wrote with each day that passes and each advancement in artificial intelligence, Language and vision technology is becoming more apparent
[00:08:44] Paul Roetzer: that there will be three types of businesses in every industry. AI native, AI emergent, and obsolete. I I keep running through examples in my mind. I said retailers, e commerce shops, marketing agencies, media companies, law firms, medical practices, keep
[00:08:58] Paul Roetzer: going. And I can’t come up with [00:09:00] an industry or business model where this won’t be true. I went on to write, take any of these or your own business and simply
[00:09:06] Paul Roetzer: look for inefficiencies and repetitive processes, opportunities to drive revenue through greater predictive models, such as, customer acquisition, retention and growth,
[00:09:14] Paul Roetzer: and ways to unlock ideation and innovation through previously unattainable creative possibilities. You could later add. degenerative capabilities to that as those became readily available to people. And then in that post, we went
[00:09:27] Paul Roetzer: on to describe AI emergent companies as established organizations that move quickly to adopt and scale AI across all areas of the organization.
[00:09:37] Paul Roetzer: They have visionary leaders who see the rapid advancements in AI capabilities and invest the resources needed to build a smarter business.
[00:09:45] Paul Roetzer: These emergent companies have expanding AI and machine learning talent pools. and at the time I referenced adobe, for example, had 300 AI ML employees at that time, according to LinkedIn. They innovate faster than the competition, potentially
[00:09:59] Paul Roetzer: through venture [00:10:00] studios or R& D labs on building AI tech, and they excel at personalization marketing, sales, and service.
[00:10:06] Paul Roetzer: They have the data, customer bases, and infrastructures to withstand the AI native companies, so the startups that are building AI first, if they move fast enough transform. Plus, they have the money to acquire the AI native companies before they grow to dominate. So that just
[00:10:22] Paul Roetzer: sort of sets the stage of like going back to 2022, how we were looking at sort of the future of business. And then you fast forward to today with this Moderna example.
[00:10:32] Paul Roetzer: And the thing I love about, I mean, I don’t know who, I don’t know if GPT 4 wroteThe case study, but
[00:10:36] Paul Roetzer: it’s a really well done case study. Mike and I used to write case studies and things like this for clients back in our, you know, agency days.
[00:10:43] Paul Roetzer: This is really well done because, What I loved they led with the why. So it said, Moderna is using its platform developing mRNA medicines to bring up to 15 new products to market in the next
[00:10:55] Paul Roetzer: five years. In order to achieve its ambitions, Moderna has adopted a[00:11:00]
[00:11:00] Paul Roetzer: centric, technology forward approach, constantly testing new technology and innovation
[00:11:05] Paul Roetzer: that can increase human capacity in trials. So they basically kind of lay out fact that they’re having this
[00:11:13] Paul Roetzer: um, unparalleled growth opportunity. And at one point they actually literally say that this would take thousands of people to do if they weren’t using AI. SoThe CEO is close as we believe very profoundly at Moderna that ChatGPT and what OpenAI is doing is going to change the world.
[00:11:31] Paul Roetzer: We’re looking at every business process, and this part I love, from legal, to research, to manufacturing, to commercial, and thinking about how to redesign them all. So this is it. Like again, go back to
[00:11:41] Paul Roetzer: the AI emergent definition, it’s a, it’s a visionary leader, like you need someone who’s looking out his head and saying this isn’t just like
[00:11:47] Paul Roetzer: a marketing this is an every function of the organization thing, and we’re going to like aggressively push this. Then it went on to say they’re, their objective was achieve 100 percent adoption and [00:12:00] proficiency of generative AI by
[00:12:01] Paul Roetzer: all its people with access to digital solutions within six months.
[00:12:06] Paul Roetzer: Quote, we believe in collective intelligence when it comes to paradigm changes. It’s everyone together, everyone with a voice, and no one left behind.
[00:12:14] Paul Roetzer: Then they said they had individual change management initiatives, including in depth research listening programs, as well as trainings hosted in person, online, and with dedicated AI learning companions.
[00:12:26] Paul Roetzer: Um, they had prompt writing competitions to find the top 100 AI power users. They had local, office hours with every business line and geography.
[00:12:37] Paul Roetzer: With over 2, 000 active participants, um, structured change management initiatives, including engaging Moderna’s CEO and executive committees. And then there was a quote in here from
[00:12:47] Paul Roetzer: their chief information officer, Brad Miller. said 90 percent of companies want to do Gen AI, but only 10 of them are successful. And the reason they fail is because they haven’t built the mechanisms of actually [00:13:00] transforming the
[00:13:00] Paul Roetzer: workforce to adopt new technology and new capabilities. This is exactly what we’ve been saying, Mike, like, like people complain that it doesn’t have the impact it should have, and the models aren’t smart enough, or they make mistakes. And the thing we
[00:13:14] Paul Roetzer: always say is, You’re not thinking about the right way. You’re not properly training everyone. Like, you can’t just buy the tool, give it to everyone, and think just going to happen.
[00:13:25] Paul Roetzer: So they went on to say, you’d cited a couple of those metrics, but, um, they said within two months, they had. 750 gPTs, 40 percent of weekly
[00:13:34] Paul Roetzer: users had created a GPT and each user had 120 chat GPT enterprise conversations per week on average. and then there was a section, the final section on the cloud is this idea that said
[00:13:48] Paul Roetzer: a team of a few thousand can perform like a team of a hundred thousand. And this is where the quote came from. said if we, With an ambitious plan to launch multiple products in the
[00:13:58] Paul Roetzer: few years, Moderna sees aI as [00:14:00] a key component their success. Quote, If we had to do it the old biopharmaceutical ways, we might need 100, 000 people today.
[00:14:09] Paul Roetzer: We really believe we can maximize our impact on patients with a few thousand people using technology and AI to scale the company.
[00:14:16] Paul Roetzer: So, I don’t know if anything else caught your attention, Mike, but this is like, again, the thing I keep saying and keep thinking when we talk to these big enterprises is like, You, this goes everywhere.
[00:14:27] Paul Roetzer: It’s every function of the business and you to be educating people first and foremost and then you have to give them like the resources to actually explore possible with this technology and then encourage all this sharing. So I just, I love everything about what they feature in this case study.
[00:14:41] Mike Kaput: Yeah, like you, like you mentioned, obviously, it’s never as seamless as it appears, just operating on
[00:14:49] Mike Kaput: single case study, but I think people would be wise to look at just, look at the numbers again. Look at how many conversations your average employee is having. If you are coming [00:15:00] away dissatisfied from these tools
[00:15:01] Mike Kaput: because you had three or conversations and it didn’t cut it, they’re having 120 per week. It’s going to take hundreds of conversations
[00:15:09] Mike Kaput: To even begin to get towards what you want. Is that kind of what you took away from this in terms of not only the education piece, but people do need to actually invest a large amount of time and trial and to get these tools to work well?
[00:15:23] Paul Roetzer: Yeah,
[00:15:23] Paul Roetzer: it’s viewing it as transformation, not tool use. You know, I think that’s the companies that go in saying, we believe this is going to transform. Our,
[00:15:32] Paul Roetzer: workforce, our strategies, our technology stack, and we’re going to approach it as such. We’re going to put the right resources, we’re going to build the right infrastructure, put the right governance in place, like not just go get five tools to be testing and hope
[00:15:47] Paul Roetzer: people figure out how to get some out of the 30 a month license that they’re using. So yeah,
[00:15:52] Paul Roetzer: I think it’s just the organizations that take that approach. And then you mentioned asana. It is a different approach because they’re coming at it most likely from a product first [00:16:00] standpoint because Asana is a project management system platform if not
[00:16:03] Paul Roetzer: familiar. we’re big fans. We’ve been using it since, I don’t, I don’t know, we started using it right before, the agency, so probably a decade or so, been using asana. we use it run,
[00:16:13] Paul Roetzer: the from a project management perspective.
[00:16:16] Paul Roetzer: Um, so Dustin Moskvich, who you mentioned, if people aren’t familiar, he was actually one of the co founders
[00:16:21] Paul Roetzer: of Facebook, and then he left founded Asana in 2018. He, you may see his news, his name in the news otherwise, because he’s currently has beef with Elon Musk.
[00:16:31] Paul Roetzer: Um, I will not disclose, like go check Elon’s
[00:16:36] Paul Roetzer: Twitter thread from Friday. I was messaging Mike when I was sitting at dinner. I was like, what, what happened? What did I miss? I
[00:16:42] Paul Roetzer: was in a talk at ohio university Friday afternoon and Elon Musk with this really inappropriate tweet up about Dustin. I was like, what, what happened?
[00:16:50] Paul Roetzer: And mike’s man, I have no idea. So I found out what happened. there’s been a lot of beef emerging between the two, but basically he, he, He said that Tesla is the [00:17:00] next enron, that it’s the
[00:17:01] Paul Roetzer: whole thing is a big scam. And Elon is basically just like selling this vision of full self driving that’s never going to happen. And so he basically called Elon out and said it was Enron. So Elon
[00:17:12] Paul Roetzer: was not very happy with that. So Dustin and news for other reasons you may hear about, but back to back to Asana. So the way they’re approaching it, similar thing. You a CEO with a vision, right? for how this is going to transform the company.
[00:17:27] Paul Roetzer: So they’re looking at it from a product perspective. But the I liked about this example is it was a letter from Dustin. So it was very clear that this is coming from the top. Um,
[00:17:37] Paul Roetzer: and they’re a company that has been investing in AI for a long time. Like we’ve talked about
[00:17:41] Paul Roetzer: this before, aI isn’t new. Like ChatGPT wasn’t the beginning of aI. There were companies like Asana, like Salesforce, like Adobe that had hundreds of aI and ML engineers way before ChatGPT emerged, but they were working on machine learning.
[00:17:56] Paul Roetzer: were working on predictive models. They were making predictions about [00:18:00] outcomes and behaviors that could be applied to like forecasting and recommendation engines and, optimization of pricing and like things like that and products.
[00:18:08] Paul Roetzer: Well, what they’re doing now is he’s saying like, okay, we were already in AI, but now we’re really in, and like, we’re really seeing the potential generative AI, not just within our product, but within our own company. And so there was one point he internally reactions to AI range from exhilaration to skepticism.
[00:18:24] Paul Roetzer: We knew we needed to build literacy and hands on experience to unite everyone around this transformation. Again, transformation, not just tools.
[00:18:33] Paul Roetzer: So we an internal AI community. community and immersive workshops, encouraging all employees to tinker with the technology.
[00:18:39] Paul Roetzer: stood stood up Slack channels. We we started evangelizing the of AI new internal use cases and stories of our personal breakthroughs that company all hands and in team meetings. And then it just goes on to say kind of how they got everyone involved and
[00:18:52] Paul Roetzer: began to glimpse a future aI moves beyond isolated chats to embedded contextual collaboration, kind of their own use within this.
[00:18:59] Paul Roetzer: So, [00:19:00] definitely another one to just keep eye on. They do say they have a work innovation summit on June 5th that we’ll track, where they said, quote, We’ll be showcasing the future of human AI teamwork. And then
[00:19:11] Paul Roetzer: ends with, Together, let’s build a future where every team has the power of AI at their fingertips, where human creativity and machine intelligence combine to solve the world’s greatest challenges, and where
[00:19:22] Paul Roetzer: work is not just more efficient, but more fulfilling, more impactful, and more profoundly human. That is like something I would read, like that is exactly what we’ve been saying all these years. More intelligent,
[00:19:32] Paul Roetzer: more human is the tagline we’ve been using. so I love the vision. I love what they’re trying to do. we’re not using any of the
[00:19:40] Paul Roetzer: AI features within Asana to my knowledge today. So, you know, I think we need to probably dive back in and see if there’s anything going on. Um Yeah, but certainly, you know, two good examples today of this whole vision of you gotta have, it’s gotta come from
[00:19:54] Paul Roetzer: top. Like, you can be building your AI councils, you can be trying to lead within your department. But the [00:20:00] companies that probably win here, it’s going to come from the C you’re going to have support from on high and you’re going to have a leader who has a
[00:20:08] Paul Roetzer: vision to transform the organization over the next five years, because I think that’s probably the window of opportunity for most industries. You got three to five years to figure this out, basically.
[00:20:20] Microsoft Phi 3
[00:20:20] Mike Kaput: So, in our next big topic today, Microsoft has announced the Phi-3, Phi-3Family Open Models, and these are noteworthy because they’re actually small language
[00:20:35] Mike Kaput: or S L M’s. According to Microsoft, S L M’s quote, offer many of the same capabilities found in large language models, but are smaller in size and are trained on smaller amounts of data.
[00:20:47] Mike Kaput: Now, these models are much smaller than LLMs, they consume far less compute and they can run locally, or have the potential to, on devices like an iPhone. So [00:21:00] one of the main models in the family, Phi-3 MINI, has 3. 8 billion parameters.
[00:21:07] Mike Kaput: It was trained on 3. trillion tokens, and Microsoft says it performs better than models twice its size.
[00:21:15] Mike Kaput: There are some Phi-3 models coming soon, with more parameters. 7 billion and 14 billion have been mentioned by Microsoft. And while large language models are basically unmatched right now for complex tasks and the most advanced use cases. The reason this is of a big deal that these small language models perform really well on a lot
[00:21:38] Mike Kaput: of simpler things that can make AI more accessible and easier to use. for organizations with limited resources. Small language models can also be more easily fine tuned. Sonali Yadav,
[00:21:52] Mike Kaput: Yadav, who’s Microsoft’s principal product manager for generative AI, said the following about this trend.
[00:21:58] Mike Kaput: Quote, What we’re going to start to [00:22:00] see is not a shift from large to small, but a shift from a singular category of models To a portfolio of models where customers get the ability to make a decision on what is the model for their scenario.
[00:22:13] Mike Kaput: So Paul, I found this trend really interesting in of what it could mean for AI adoption. Like, how do you see small language models potentially increasing or streamlining how organizations adopt AI?
[00:22:28]
[00:22:28] Paul Roetzer: For just some macro level context here and, reason why these models are important is because it costs a lot of money to run the big models. So you know, we’re building GPT 5 or Gemini
[00:22:40] Paul Roetzer: 1. 5, or, you know, Claude opus, like the biggest frontier models. to use those models. So training them costs hundreds of millions or billions of dollars. That’s one thing, but then the inference costs. So when you and I go to use the tool, um, for whatever it is, to write our emails,
[00:22:58] Paul Roetzer: to generate our images, [00:23:00] eventually to generate our, you know, 10 second videos. That’s the inference cost. So if you’re using these massive models, every time you need to do a discrete task,
[00:23:10] Paul Roetzer: it can get insanely expensive. So for Microsoft, who’s licensing OpenAI’s technology to do, say, 365 Copilot, it’s built on OpenAI’s GPT 4 or 4. 5, whatever it is, technology, it costs
[00:23:24] Paul Roetzer: them a bunch of money. to pay OpenAI to to use it every time you or I go in and use Copilot. So there’s a lot of motivation to build these smaller models that don’t cost as much money and
[00:23:37] Paul Roetzer: what the research seems be showing recently is, as you alluded to, you can train these things a little easier. There’s been a number of research reports just in the last few weeks that shows that the quality of data matters a lot and
[00:23:51] Paul Roetzer: that if can take a smaller model that costs less to train and costs less to run. And you can give it
[00:23:57] Paul Roetzer: more kind of hand picked data [00:24:00] sources that these things can perform like a much larger model in terms of their quality and reliability if the data is really good that goes
[00:24:09] Paul Roetzer: into them. apple actually has released a number of research papers showing this is the direction they’re going, where you build these smaller models that don’t cost as much to run, don’t cost as much to train, And they can run on device.
[00:24:23] Paul Roetzer: So you don’t have to go to the cloud every time you want to do something. So again, every time you or I go into ChatGPT wherever, we’re going and pulling
[00:24:32] Paul Roetzer: compute in the cloud somewhere, which costs money. What they’re saying is in the future, you’re, you could be on your device do not disturb or, or airplane mode and you could be running a model doing something on your phone.
[00:24:45] Paul Roetzer: That’s what this means. And so that’s why you’re going to hear a lot more about this we lead into the June developer Conference for Apple. Um,
[00:24:52] Paul Roetzer: this on device models that can do things for you on your phone without an internet connection is going to be a really big unlock, [00:25:00] and I think it will lead to a lot more adoption, and it’s probably going to be much more like the adoption we had in the 2011 to 2020 range
[00:25:08] Paul Roetzer: where we were all using aI in Netflix and Spotify and spotify and YouTube and Facebook. Like AI a AI was part of our lives we didn’t realize it. That’s what this will enable is like that these models can just be underlying functions on your phone
[00:25:22] Paul Roetzer: and you don’t even know you’re using AI. It’s not like you’re going a AI app to use AI. It’s just going to be embedded within everything.
[00:25:29] OpenAI’s Sam Altman and Other Tech Leaders to Serve on AI Safety Board
[00:25:29] Mike Kaput: So in our third big topic today, we got news that Sam Altman, Microsoft CEO Satya Nadella, and Alphabet CEO Sundar Pichai are joining
[00:25:40] Mike Kaput: called an AI Safety and Security Board that’s being run by the U. S. government. Now this board is going to advise the Department homeland Security on how it can safely deploy AI within
[00:25:53] Mike Kaput: critical infrastructure. So it’s going to do things like make recommendations on how operators of [00:26:00] core infrastructure, like power grids for instance, can protect their systems against AI threats,
[00:26:05] Mike Kaput: Those kinds of topics and advisory recommendations. Now, almost two dozen political, tech, and business are on this board. So some of the other notable names
[00:26:18] Mike Kaput: include NVIDIA’s Jensen Huang, Anthropic’s CEO Dario Amadei, the CEO of Northrop Grumman, the mayor of Seattle, who’s also the chair of the U. S. Conference
[00:26:29] Mike Kaput: of Mayors Tech and Innovation Committee, the governor of Maryland, and noted AI researcher Fei Fei Li.
[00:26:37] Mike Kaput: Now, Paul, when we talked about this before the episode, you mentioned that some, omissions from this group that stood out to you.
[00:26:47] Mike Kaput: Who, who was that?
[00:26:49] Paul Roetzer: Yeah. So, I mean, the obvious thing is Meta’s not on there, so no Zuckerberg, no Yann LeCun.
[00:26:54] Paul Roetzer: Uh, could say elon Musk, who’s certainly a lot to aI. So there was some [00:27:00] obvious ones there.
[00:27:01] Paul Roetzer: And then the biggest one that initially jumped out me is what were the open source people? Like, so a lot of the people on here from the AI perspective are, are, you know, not the big ones pushing for open source acceleration.
[00:27:14] Paul Roetzer: Um, so that seemed to be, um, one thing. There was, I did see a. Tweet from Gavin Baker, who, know, I follow pretty closely, he’s a managing partner and
[00:27:23] Paul Roetzer: CIO at an investment firm, but involved in AI. And he tweeted,
[00:27:29] Paul Roetzer: can’t whether it is funny, ridiculous, or sad the CEOs of occidental Petroleum and Delta Airlines are on the new AI safety board.
[00:27:38] Paul Roetzer: Less than half of 22 members have any real AI knowledge. Um, also odd that it’s not vaguely bipartisan, has zero open source AI CEOs, and excludes Elon. so that was my initial,
[00:27:52] Paul Roetzer: like, that was actually one of the first times I saw this list. And then, when I went back in read about it, it actually made a lot more sense, like what his critique, I [00:28:00] think, didn’t
[00:28:00] Paul Roetzer: necessarily hold up, at least from the business perspective, because this is all about critical infrastructure, as the power grid and transportation.
[00:28:07] Paul Roetzer: So, it makes perfect sense that someone from petroleum company, someone from an airline company, like, You would want diversity. They don’t have to be AI experts to able to explain
[00:28:17] Paul Roetzer: how transportation in the United States works or how the energy grid works, things like that. So it’s actually smart to have a diverse group of people that can bring that kind of knowledge to the table.
[00:28:29] Paul Roetzer: That doesn’t mean that there shouldn’t be open source people on there. Someone like a, you know, a Zuckerberg shouldn’t be on there. But I think they’re basically probably looking at which companies are we working with on the infrastructure of the US economy and the government. Thank you. And then what businesses are represented
[00:28:45] Paul Roetzer: that need to be there. So, yeah, I mean, my overall take was, I think it’s good that these conversations are happening. The infrastructure is something we don’t talk about on this show much. We will more.
[00:28:56] Paul Roetzer: It is a major problem. there are lots of [00:29:00] things that can go wrong with the infrastructure. And in many cases, the infrastructure is 70 years old or more, and that’s a problem and it’s an attack vector for, People use.
[00:29:11] Paul Roetzer: Um, so, yeah, and then the other thing, we’ll come back to this later on, but there was a, there’s a new bill, SB 1047 california, Safe and Secure Innovation frontier intelligence Models Act. That is,
[00:29:27] Paul Roetzer: uh, apparently moving pretty quickly in California. And this seems to be a of an attack on the open world as well. At least it’s being positioned by some open source advocates as such. Jeremy Howard, who is someone to follow on Twitter, is, posted an article about like
[00:29:49] Paul Roetzer: his thoughts on this. So he’s an AI researcher and entrepreneur, cEO of Answer. ai. Bye. and so he goes on to kind of do a takedown of why open source is so [00:30:00] important.
[00:30:00] Paul Roetzer: So I think, again, we’ll kind of come back to this maybe next week.
[00:30:03] Paul Roetzer: I think it’s worth talking a little bit more about this battle between open source and closed source as we start moving into regulations. But it’s going to be a really important
[00:30:12] Paul Roetzer: uh,
[00:30:12] Paul Roetzer: topic as we move forward, especially as we start getting into the elections and as the government starts looking at ways to, you know, protect infrastructure and things like
[00:30:21] Paul Roetzer: So yeah, really critical topics. keep an eye on sB 1047 in California. It seems like it’s moving real fast.
[00:30:30] Mike Kaput: Alright, so let’s dive into some rapid fire topics this week.
[00:30:34] Sam Altman Comment
[00:30:34] Mike Kaput: The first up is a pretty interesting public comment from Sam he recently made a public appearance at Stanford’s Entrepreneurial Thought Leaders. Seminar called
[00:30:46] Mike Kaput: ETL. and he some harsh words to say about ChatGPT At one point in the discussion, he said, quote,
[00:30:54] Mike Kaput: is mildly embarrassing at best gPT 4 is the dumbest [00:31:00] model of you will ever have to use again by a lot But it’s important to ship early and often, we believe in iterative thinking Deployment.
[00:31:08] Mike Kaput: Now Paul, this isn’t some groundbreaking knowledge interview that he’s dropping here, but it was I wanted to highlight because I think many people who are relatively new to AI don’t always grasp just how much leaders like altman believe existing AI technology is going to improve, and soon. And you also mention this point in talks all the time, saying this is the least capable AI you’ll ever use. Could you kind of unpack idea a little more for us?
[00:31:40] Paul Roetzer: Yeah, you know, I think it’s, it is just a good reinforcement that, you know, people look at GPT 4 today and maybe they are maybe there are some organizations or some leaders out
[00:31:51] Paul Roetzer: there who don’t see the real potential for transformation. They’re not doing what Moderna or Asana are doing, and they’re not really driving it because they just don’t see the value. [00:32:00] So one,
[00:32:00] Paul Roetzer: I think it’s really good reminder that it’s only going to get better. it’s better from here, and it’s probably going to get significantly better from here. And the other thing that I thought he gave some perspective on is, he said this in many,
[00:32:11] Paul Roetzer: interviews, that it’s better that they don’t just build So, again, their mission is artificial general intelligence, AI that is at or above human level at all cognitive tasks, generally speaking.
[00:32:22] Paul Roetzer: Um, and so his feeling is like, we don’t, we don’t want to just go away and build aGI and then just drop it on the world in, you know, two years, five years, whatever. And then all of a sudden it’s like, oh my,
[00:32:33] Paul Roetzer: Oh my God, like, where did this come from? Like if we had never seen ChatGPT, if we’d never experienced these tools as they were evolving.
[00:32:40] Paul Roetzer: So OpenAI does iterative deployment. So they’re a big believer in let’s ship this stuff early, ship it often, give people a chance to react to it. Now, if we rewind back to,
[00:32:52] Paul Roetzer: think it was GPT 2, when it first came out, they put out a paper and said, Hey, we’re not releasing this the world because it’s too expensive. It might be too [00:33:00] so they were worried about 2 being misused. it came now people are worried about GPT 5 and my gosh, what’s going to happen to society. And so Sam’s point is like, listen, there’s this initial
[00:33:12] Paul Roetzer: when, when new things come out, like ChatGPT comes out, everyone freaks out. And so, And so then he said, GPT 4, which came out a year ago, was met with two weeks of freaking out and people believed it was this crazy thing and the world had changed forever.
[00:33:24] Paul Roetzer: Now people are like, oh, it’s horrible. Where is GPT 5? So his whole point is like, As a species, we adapt, change is weird. I’ve referenced the one that Andrej karpathy said at one point where he was talking about like Waymo. Like, most people don’t realize, like, self driving is actually a thing. Like, there are taxis, basically, in
[00:33:44] Paul Roetzer: California that don’t have drivers and you can order them on phone and you can get in them. Well, the first time you see that, you’re just like, What is that? And then you continue walking down the street and like you go about your life.
[00:33:56] Paul Roetzer: And so I think like as much as AI is going to [00:34:00] transform everything, like it’s going to be this process where things just start to be weird and you’re going to look back two years
[00:34:07] Paul Roetzer: be like, oh my god, I can’t gPT 4 was like what we were using. That was such a terrible tool. But right now it feels like It’s very impactful. And so I think that’s the path we’re on with this technology.
[00:34:19] Paul Roetzer: Sam’s point is listen we’ll we’ll keep figuring it out. Now I don’t necessarily buy into this when it comes to like the jobs and the workforce and the economy conversation, which is a whole nother topic.
[00:34:30] Paul Roetzer: Um, but this is the approach of most of these technology leaders is like, Hey, we adapt, we figure things
[00:34:36] Paul Roetzer: out, it’ll all be okay. And like, we’ll have time to solve this. So yeah, that was, that was the, there was a few. quote worthy things from the interview at Stanford for sure.
[00:34:47] Mike Kaput: It certainly sounds like he does not think progress aI will be slowing down anytime.
[00:34:51] Paul Roetzer: he not. Yeah,
[00:34:53] Paul Roetzer: and
[00:34:53] Paul Roetzer: he’s been very clear that the leap to five is going to be
[00:34:56] Paul Roetzer: massive. but yeah, we’ll, we [00:35:00] will see hopefully, well, I don’t know, hopefully, but we’ll probably
[00:35:02] Paul Roetzer: see sooner than later.
[00:35:04] TSMC’s Debacle in the American Desert
[00:35:04] Mike Kaput: All right, another topic on the docket this week. We got a new in depth report from publication called Rest of World, and it details some really significant challenges that are being seen building out
[00:35:18] Mike Kaput: AI infrastructure in the US. The article does a deep dive into TSMC, taiwan Semiconductor Manufacturing Company. This
[00:35:27] Mike Kaput: is of the top AI chip makers on the planet that makes AI hardware basically possible. And it’s currently engaged in efforts to build and staff a chip fabrication plant in Arizona. So this plant has right now about 2, 200 employees and
[00:35:44] Mike Kaput: it’s kind of seen as this leading indicator of efforts to diversify AI chip manufacturing away from Taiwan given the geopolitics of that region and kind of great power competition between the U.
[00:35:57] Mike Kaput: S. and China. But this [00:36:00] appears to be a lot easier said than done, because this report, which is well worth a read, details tons of obstacles that TSMC is running into trying to get this plant up and running, and primary among these aren’t just, you know, technical hurdles. There’s serious cultural clashes between American and Taiwanese ways of working.
[00:36:23] Mike Kaput: The report details how Americans had to go train at TSMC taiwan for like year. But all of the training were in Taiwanese and Mandarin Chinese, so they were basically hacking it together with
[00:36:35] Mike Kaput: Google Translate to learn what they’re supposed to learn. There were also some significant clashes tSMC’s work culture, which is, quote, notoriously rigorous, even by Taiwanese standards.
[00:36:47] Mike Kaput: And Taiwanese engineers were also coming to the U. S. and then very critical, of their American counterparts, work ethic and technical skills.
[00:36:56] Mike Kaput: All of this is to say, you know, Paul, we hear a a lot of [00:37:00] talk aI leaders about the need to invest billions or trillions into AI infrastructure like this in coming years, especially in the US.
[00:37:09] Mike Kaput: But it just sounds like there’s some serious human obstacles here. I mean, could this type of thing hold back AI innovation?
[00:37:17] Paul Roetzer: Yeah, and it’s totally predictable. Like, so, if this is an interesting topic that wanted, it probably should be. just, if nothing else, you understand the supply chain that powers
[00:37:29] Paul Roetzer: smartphones, your cars, all the AI tools that rely on today, all the ones we rely on in the future, they are all dependent upon TSMC. and the supply chain of building these chips. also your retirement portfolio potentially you invest in NVIDIA.
[00:37:46] Paul Roetzer: Um, now I did hear an interview with Jensen Huang where he said that they are reliant on tSMC taiwan, but of the 35, 000 parts that go into each chip only eight of them are made by TSMC.
[00:37:56] Paul Roetzer: So it’s not like the whole chip is fabricated like in [00:38:00] in Taiwan, but it’s It is a critical part for sure. So this affects everyone. The article is insane. Like it’s a long read, but it is like, it’s comical at times, but, also sad
[00:38:15] Paul Roetzer: because it’s just, as you said, like, I don’t know how you’ve solved this. Like, and there are. The the CEO of TSMC has basically said for years like you can’t do in america like this isn’t going to work But i’ll take 10 billion dollars and we’ll try and try and build this in Arizona
[00:38:33] Paul Roetzer: you and we’ll see how it goes and it’s mainly a culture and and labor
[00:38:36] Paul Roetzer: issue Like they’re just very different in Taiwan than in in America and that becomes extremely apparent when you read this article that it is probably more misaligned than you could imagine to try and do what they do in Taiwan and in the U.
[00:38:52] Paul Roetzer: S. Um, so two quick recommendations. Chip War by Chris Miller, great book on the topic. And then there’s an [00:39:00] article from Forbes by Rob Tao’s called The Geopolitics of AI Chips
[00:39:04] Paul Roetzer: Will Define the Future of AI.
[00:39:06] Mike Kaput: AI.
[00:39:07] Paul Roetzer: that article is from May of and I believe the book came out 2023 well. So again, if you’re, if you’re fascinated by this thread AI, start with
[00:39:18] Paul Roetzer: the article on Forbes from Rob Tao’s great insight into it. He also had a TED talk on the topic, I believe. And then if you want to keep going, read Chip by miller. it a fascinating topic. It’s one of those, like,
[00:39:30] Paul Roetzer: I try and not too much time in because it’s like, I have no control of this whatsoever. And there’s times where you read this stuff. You’re like, Oh, this is going to go haywire. Like, this is not going to work out. And I feel kind of helpless. So I like, it’s kind
[00:39:43] Paul Roetzer: of like cybersecurity. Like I’ll dip in every once in while, read about it. And like, I got to get out of here. Like I got too many other things to worry about than the supply chain for AI chips.
[00:39:51] Paul Roetzer: But it’s a fascinating topic.
[00:39:53] Expansion of Meta Ray-Ban Smart Glasses Collection
[00:39:53] Mike Kaput: All right. up, Meta has announced that it’s expanding. Its RayBan Meta Smart [00:40:00] Glasses collection to include
[00:40:01] Mike Kaput: bunch of new styles. But more importantly, new, more powerful AI. So we had talked last week about MetaAI, the company’s intelligent assistant, and now in the U. S. and Canada, you’ll be able to use that right within your smart glasses.
[00:40:17] Mike Kaput: So you just say, hey, Meta, and you can then prompt the assistant with voice commands. Now, Meta. ai also gives the glasses the ability to access real
[00:40:27] Mike Kaput: information, and the company began testing a multi modal AI update, so you can actually ask your glasses about what you’re seeing. Now that update is now rolling out to users in the
[00:40:40] Mike Kaput: S. and Canada. So, Paul,
[00:40:43] Mike Kaput: it definitely seems like Meta’s Ray smart glasses could be, like, an actually useful way to engage with AI in the real world.
[00:40:51] Mike Kaput: We’ve talked couple times in, the last couple weeks about AI wearables and that whole trend. Like, how do you see these [00:41:00] stacking up compared with some of the other, unfortunately, rocky releases of aI wearables? definitely a better form factor, lot more positive buzz regarding these than, say, our humane AI pin or the rabbit, which we won’t get into, but the rabbit, if you recall, we talked about, I don’t know,
[00:41:21] Paul Roetzer: 10 episodes ago, 15 ago, when it was first previewed, said, It’s this device that supposedly does AI on it. At the time, I was kind of
[00:41:28] Paul Roetzer: skeptical that it was one needed and two would work. It is real bad so far. So they just started getting into the hands of people and it is, it’s not,
[00:41:38] Paul Roetzer: I don’t know if it’s as bad as the AI pin reviews, but it’s bad. just cause it only costs 1. 99 instead of 7. 99. So people are more like accepting of the fact that they just wasted 200. But, the Rabbit device is not going well. The Ray Bans, however, seem like Right form factor. Obviously, Meta has
[00:41:57] Paul Roetzer: endless money to throw at this, and they seem [00:42:00] to be being really smart. I tried to find sales data on these things, and the only thing I came up with was August 2023.
[00:42:06] Paul Roetzer: Uh, at that time said they’d sold 300, 000 of the devices, but only had 27, 000 monthly active users. So, adoption wasn’t real high. but again, it’s not
[00:42:17] Paul Roetzer: wildly expensive product, so it can be somewhat disposable if you’re a higher income person. It’s like, you know, 300 bucks at something, don’t like it, it’s okay.
[00:42:24] Paul Roetzer: Um, I could see these being more heavily adopted sure in, in the months and years ahead, and I think more people will get into this space.
[00:42:33] Paul Roetzer: I would be shocked if Apple doesn’t at some point apply their Vision Pro technology, to this. And by the way, we should come back to the
[00:42:40] Paul Roetzer: Vision Pro at some point. that’s not going well either. I mean, awesome tech, like at they have good tech, but it’s not being supported. Like there’s, I have like no new apps, no new immersive experiences. Like, I don’t know what
[00:42:53] Paul Roetzer: they’re doing. And apparently they just cut production like down to 400, 000 So yeah, I [00:43:00] think there’s just going to be, it’s going to be a rocky go with these, kind of immersive
[00:43:04] Paul Roetzer: experiences and, and AI devices, even for the big companies, but Meta seems be on track here more than most.
[00:43:13] Eric Schmidt-backed Augment, a GitHub Copilot rival, launches out of stealth with $252M
[00:43:13] Mike Kaput: So we also got some news that a new AI startup being backed by former Google, CEO eric Schmidt has just emerged from stealth with a whopping $252 million in funding. This company is called Augment. And it aims to challenge GitHub Copilot by offering a better version of an AI coding assistant to help programmers be more productive
[00:43:39] Mike Kaput: effective. So, essentially, AI assistants that will generate code for You debug code, and help you, create programs much faster. Now the company was actually co founded by an ex Microsoft developer and a former AI research scientist at Google.
[00:43:57] Mike Kaput: They are entering however a pretty [00:44:00] crowded and competitive market. GitHub Copilot has 1. 3 million paying individual customers and 50, 000 enterprise customers. Amazon and Google
[00:44:11] Mike Kaput: their own coding assistants and there’s a ton of other startups competing in this space as well. Now, Paul, while we don’t have a ton of details just yet on this company, it has raised a significant amount of money, has some notable investors, its founders have interesting and relevant backgrounds.
[00:44:28] Mike Kaput: These kind of tick all these boxes we look for when it comes to. Paying attention to certain companies, like, what are your first impressions of this?
[00:44:35] Paul Roetzer: Yeah, so we, on a recent episode, we talked about, you what startups get our attention, the the investors, the amount of investing, the founding team, sort of like check, check, check, this one, Eric Schmidt, if you’re not familiar, he was the CEO and chairman of Google from 2001, so just
[00:44:54] Paul Roetzer: couple of years after they were founded.
[00:44:56] Paul Roetzer: He sort of the adult brought into the room, basically, to kind guide [00:45:00] Google, and so he was the CEO till 2011, and then he stayed on as the chairman of the board, I think, until 2015 or 18. and so you know, a billionaire, and, he invests heavily, he advises the
[00:45:16] Paul Roetzer: and I think the Department of Defense, like an advisor for AI there, so he’s heavily involved, still. And so to see him leading this kind of round for a company to come out of stealth with a quarter billion dollars, that’s going to get your attention. And certainly the co pilot is, the GitHub co pilot is one of very
[00:45:34] Paul Roetzer: early, seemingly highly reliable uses of AI. Like we’re seeing it really impacting the coding world. And so you’re going to see some competition flow into this space. So, yeah, just, again, things that, you know, Kind of perk our ears up is when
[00:45:49] Paul Roetzer: hear a quarter billion in funding and see Eric Schmidt’s name tied to something and the founders from Google and microsoft. definitely a startup worth paying attention to. I
[00:45:59] Musk’s xAI Is Close to Raising $6 Billion from Sequoia, Others
[00:45:59] Mike Kaput: [00:46:00] Alright, you know, it wouldn’t be AI news without Elon Musk being in the news again. Elon Musk’s AI company, XAI, which makes Grok, the AI assistant that he has, is reportedly close to raising 6 billion from from investors that include Sequoia
[00:46:18] Mike Kaput: Capital. This raise would value the company at 18 billion dollars, and according to the the information, it’s expected to close in the next two weeks.
[00:46:28] Mike Kaput: Now, according to their reporting, the company is currently training the second generation Grok on 20, 000 NVIDIA H100 chips, and Musk has has also that the company needs 100, 000 GPUs to Grok 3.
[00:46:45] Mike Kaput: So likely where some of this money is going. But really, kind of big question I have here, Paul, is like, you’ve been, I mean, we both have been pretty underwhelmed, but with Groq, like in our initial tests, is this kind of valuation justified for [00:47:00] this company or
[00:47:00] Paul Roetzer: or this tool? I mean, Elon sells visions. like, no, Grok is useless right now still. Again, I, if someone has a use Grok, like please reach out to me
[00:47:10] Paul Roetzer: tell me how you’re using thing. I just don’t understand, like, what, what it’s supposed to be doing.
[00:47:15] Paul Roetzer: Um, But, he sells visions, like, we’re going to go to Mars, going to electrify the world with cars, and, you know, that’s his thing, and so I’m sure he’s just pitching some grand vision for aGI and, and its embodiment into robots, and how
[00:47:33] Paul Roetzer: going to accelerate, you know, getting to Mars, and, you know, saving planet.
[00:47:38] Paul Roetzer: So, and he can get money from people, whether it’s Sequoia or, you know, other government, funds, is certainly a place he tends to tap. He, which creates some of friction between him and
[00:47:52] Paul Roetzer: the U. S. government is Elon’s very friendly with, Other governments that the US doesn’t necessarily want
[00:47:57] Paul Roetzer: to be friendly with and because a [00:48:00] source of money for him. So, yeah. I don’t know. I guess keep keep watching grok. Maybe it’ll, do something
[00:48:07] Paul Roetzer: some point. I don’t know. But yeah, I I wouldn’t be shocked if he raised this. I wouldn’t be shocked if it was more money than this. Like, he’s, he can raise whatever he wants. Probably
[00:48:16] Apple Intensifies Talks With OpenAI for iPhone Generative AI Features
[00:48:16] Mike Kaput: All right. Next up we. have gotten news that Apple has started back up discussions with OpenAI about using its AI to power new iPhone features coming later this year. This comes from
[00:48:29] Mike Kaput: reporting by Bloomberg. These discussions revolve around the terms of a possible agreement between the two companies that would integrate OpenAI features into iOS 18, which is the next iPhone operating system.
[00:48:43] Mike Kaput: Now earlier this year, Apple and OpenAI apparently had talking, but it looks like those talks had stalled until now. Apple is also in discussions with Google to potentially license their Gemini models for use
[00:48:57] Mike Kaput: their products. So Paul, these are [00:49:00] still discussions and rumors, not any actual deals yet, but you’re a long time. Apple Watcher. Like,
[00:49:07] Mike Kaput: is Apple engaged in trying to use AI from other companies rather than building their own?
[00:49:14] Paul Roetzer: I think they’re doing all of the above. I mean, who hasn’t Apple talks with at this point, at at least reported to have talks with? So, I don’t know. I mean, but Apple’s done deals with everybody for the iPhone at different
[00:49:27] Paul Roetzer: So, nothing would shock me if they do a deal with OpenAI or Microsoft or Google, like, I mean, they’re frenemies. They’re, sometimes they do deals together and other times they’re competing against each other. So I’m not surprised by any of it. I have no idea what they’re going to
[00:49:44] Paul Roetzer: It seems like it’s going to be a mix of own models.
[00:49:47] Paul Roetzer: So again, if you read the recent research reports coming out of Apple, which they’re not historically one to put out a bunch of research.
[00:49:54] Paul Roetzer: So they’ve been talking a lot about their, their advancements internally aI. So I think there’s [00:50:00] just going to be a mix and it may be that they’re not fully baked yet with their internal models and
[00:50:05] Paul Roetzer: they’re going to do deals with other companies until feel that their models are you know ready for prime time. I don’t know but they’re going to do something like they’re going to make
[00:50:14] Paul Roetzer: some major announcements in June and I would expect that by this fall the way your phone works is probably going to start to evolve like we’re going to experience AI on device um as soon as this fall. So one way or the other. definitely something to keep an eye on.
[00:50:33] Hubspot AI Updates
[00:50:33] Mike Kaput: Alright, our last story this is about HubSpot because HubSpot just unveiled its most recent spotlight. So this is a product showcase. company says it’s rolling out twice a year to kind of highlight new parts of the, its various hubs and other products.
[00:50:50] Mike Kaput: And AI played a starring role in this one hubSpot says it has started to embed AI across each one of its hub products across marketing, [00:51:00] sales, and service. And this spotlight showcases some of the latest
[00:51:04] Mike Kaput: AI and these include things like a feature called Clip Creator, which generates videos from text prompts, an AI powered automatic reply in their service hub,
[00:51:15] Mike Kaput: the ability to turn written content audio. with AI powered blog post narration, the ability to reports automatically using generative generative AI and text prompts, there’s AI powered brand voice which generates content sounds like your company,
[00:51:32] Mike Kaput: they have now predictive AI forecast sales performance, and much much more.
[00:51:37] Mike Kaput: Now, Paul, we’ve obviously got a very long history with hubSpot. You know, this isn’t the first time they’ve released AI features, but do you see these updates kind of fitting hubSpot’s overall AI trajectory? Yeah.
[00:51:52] Paul Roetzer: out a lot of updates. um. The really logical use cases is it’s good to see them making [00:52:00] innovation. I saw some chatter in like the partner channel. I’m still like observe, not involved it anymore, but
[00:52:06] Paul Roetzer: just that like people are struggling to up with the update. I think it was like over a hundred things in the spotlight and the way they released it is like this, this page that’s just like infinite scroll page of updates. I don’t know if there’s like a way to download.
[00:52:20] Paul Roetzer: This like I was having trouble honestly, like kind of even figuring out what was going on and what, what all the updates were and what we have and don’t have as a company. because again, we,
[00:52:29] Paul Roetzer: as you said, we use it and I’d love to know like, okay, which are things that mattered to us? I’m not going a hundred things and and figuring out what to actually use here.
[00:52:38] Paul Roetzer: Maybe they could build an AI bot that advises you on what to use. connected to your hub that says, Hey, you’re, cause I know, I mean, God, going back like 2008, they used to track like, individual usage individual apps. It’s how they knew how, you know, their happiness factor for a customer is based on your
[00:52:55] Paul Roetzer: app usage within the platform. And that was reported to partners. Like, they know [00:53:00] what we’re using, what we’re not. Like, maybe I could talk to their little AI chatbot and it could tell me, like, here’s ways you could
[00:53:04] Paul Roetzer: AI. Like, that would cool. because this is a lot to process, but good to see them continuing to push out AI updates for sure.
[00:53:13] Mike Kaput: right, Paul, that’s a wrap for this week. We really appreciate, as always, you breaking down what’s going on in AI this week. just some final notes here. I would just highly encourage anyone
[00:53:24] Mike Kaput: who’s getting value out of the podcast please leave us review on your podcasting platform of choice. It helps us get the podcast into the ears of more listeners.
[00:53:36] Mike Kaput: And I I would also just reiterate, if you do have a few minutes and haven’t, taken our state of marketing AI survey yet, please go to stateofmarketingai. com. We are keeping the survey open for the next about six weeks here and we’d love to hear from you about how you’re using artificial intelligence.
[00:53:54] Mike Kaput: It should only take a couple of for you to fill out and it helps us move the [00:54:00] industry forward by publishing really robust research on AI adoption. And then last but not least,
[00:54:07] Mike Kaput: Please go check out our newsletter this week marketingaiinstitute.com/newsletter.
[00:54:15] Mike Kaput: It covers more in depth all of the news we discussed today and also all the topics we don’t have to get to.
[00:54:22] Mike Kaput: it’s a single, comprehensive brief to get you caught up on AI in just minutes that comes out every week.
[00:54:29] Mike Kaput: So if you haven’t signed up for that, I highly, highly encourage it. Paul, until next week, thank you again for breaking it all down for us.
[00:54:38] Paul Roetzer: I have a feeling this week’s going to like make up for last week. I think there’s going to be a lot happening this week. So we will be back with you next week reporting all the news that matters to you.
[00:54:46] Paul Roetzer: Um, yeah, have a great week everyone. We’ll talk you again soon.
[00:54:49] Thanks for listening to The AI Show. Visit MarketingAIInstitute. com to continue your AI learning journey. And join more than 60, 000 [00:55:00] professionals and business leaders who have subscribed to the weekly newsletter, downloaded the AI blueprints, attended virtual and in person events, taken our online AI courses, and engaged in the Slack community.
[00:55:12] Until next time, stay curious and explore AI.