Hey Everyone,
Listen on the Go 🎧 15:54
NotebookLM, the Research Assistant from Google Labs has gotten a lot of traffic in recent weeks, and we were all wondering when Audio Overviews might get more choices and customization. Well, we didn’t have to wait long.
(☝—💎For a limited time get a Yearly sub for under $5 a month✨— ☝).
Note: Most of my articles are now for premium readers, with some unlocking (temporarily) sometimes on Friday or Saturday for all readers.
Google on Thursday updated the audio summarization feature of its AI note-taking and research assistant with a custom instructions option. You can go to the Notebook guide icon, where you can now ask your AI ‘hosts’ to talk about a specific topic or tailor their discussion to a certain audience.
Google chose to improve Audio Overviews and launched a Business paid subscription related called NotebooksLM Business. The company is expected to reveal general availability and pricing details later this year.
NotebookLM went Viral in September
SimilarWeb suggests NotebookLM saw an over 371% increase in its traffic in September to 3.07 million monthly visits, up from 652,181 a month ago. Those aren’t ChatGPT like numbers, but for a Google AI product that’s still a prototype, it’s a lot. But what will Google Labs do with it?
How to use Custom Instructions
To customize the chat, open up a notebook in NotebookLM, select “Notebook guide”.
Then navigate to the “Deep dive conversation” option.
From there, hit “Customize,” and enter instructions that you want your AI hosts to follow. NotebookLM will then generate a conversation based on the directions you’ve provided.
We have written about NotebookLM here and our guide here.
NotebookLM now works in other languages.
Until now, Audio Overviews would automatically generate AI conversations from users’ sources that by default were super general and somewhat generic sounding without any ability to personalize the output besides the sources given.
Background listening: A new feature allows users to listen to the generated audio while continuing to work within the app. This means you can query your sources or explore relevant quotes without interrupting the audio playback, making it easier to multitask.
Customizing audio summaries may also help reduce hallucination to some extent according to the product people involved in the project.
NotebookLM is branded as a Personalized AI Research Assistant that can be useful for students, researchers and learning about topics.
With custom instructions for Audio overviews (the synthetic podcast): you can experiment with tone, target audience, specific topics, sources, etc.
OpenAI makes most of its revenue B2C, while Anthropic makes most of its revenue so far B2B.
AI Manifestos are having a Moment
In late September Sam Altman the CEO of OpenAI wrote this Op-ed blog. Copying him with the prefunding ritual was Anthropic’s CEO with a longer Op-ed with the stylized name “Machines of Loving Grace”. OpenAI, Anthropic, Perplexity and others pitch investors with fairly unrealistic valuations looking for sizable funding rounds.
Founded in 2021 by former OpenAI executives, Anthropic has already raised $7.6 Billion mostly from Amazon and Google. The Anthropic Manifesto had shades of Google DeepMind, especially the last bastion of hype in techno-optimism as a movement as it relates to AI in healthcare and biotech.
Axios noted that “writing a manifesto predicting a glorious AI-driven future has become a mandatory ritual for AI leaders.” It’s not a great sign. The 15,000 word essay by Amodei while better than the immature writing of Sam Altman, is a contradiction in terms.
It’s awkward because pushes back on the idea that he’s a “pessimist” or “doomer” by outlining some grandiose claims for the future of AI. As Anthropic hires big-time product people, OpenAI recently copied them with Canvas eerily similar to Anthropic’s popular and functional Projects.
Amodei’s blog post isn’t for the average AI-curious reader, nor is it a convincing roadmap for the future of AGI, it’s not clear who he is trying to convince? Probably investors.
People are trying to convince us with social posts that NotebookLM is useful in their work? I’m not buying it. Even as Google search and Google SGEs feels very dated.
Perplexity on pace to Raise another $500 million
Perplexity has become one of the big winners of the Generative AI movement. With competition like Neeva opting to get acquired early, it’s left them alone as a pioneer. I personally also use Genspark (many times a day).
They are looking to raise roughly $500 million, with a surprising $9 Billion valuation. SoftBank Group Corp.’s Vision Fund 2 invested in Perplexity earlier this year at a $3 billion valuation. It’s not clear when OpenAI’s SearchGPT will be a direct competitor.
Perplexity Spaces seems to have evolved surprisingly well. Recently Internal Knowledge Search (IKW) has been added. See the official product video for more details.
Perplexity’s Internal Knowledge Search is a new feature introduced for Pro and Enterprise Pro users that allows them to search both their internal files and the web simultaneously. Perplexity’s Enterprise Pro is for companies of over 250 employees. Here is the Perplexity Page FAQ on Internal knowledge search.
Mira Murati OpenAI’s former CTO will start a new AI Startup
It’s not surprising any longer when former OpenAI employees start their own company. Mira Murati was at OpenAI for 6.5 years. Murati was reportedly among those, along with ex-OpenAI chief scientist Ilya Sutskever, who approached the rest of the board prior to Altman’s ouster to express concerns about his behavior. As female leaders in Silicon Valley find out, going against tech bros isn’t a great idea.
So it’s not surprising Mira Murati has been pushed out or left or a combination of the two. With most of the original cofounders no longer at OpenAI, it’s clear Sam Altman is not an easy person to work with or his goals might not be very open.
Of course not much is known about Mira’s actual new startup yet, though it likely includes other former OpenAI team members. This startup will reportedly focus on building AI products based on proprietary models and could raise more than $100 million in the seed or pre-seed round.
BigTech Barnyard Variety Factions – America’s AI Exceptionalism Farm
Whereas Microsoft Corp. and Apple Inc. have invested in and partnered with OpenAI to provide generative AI services, Microsoft’s own relationship with OpenAI becoming even more complicated, Google and especially Amazon seem closer with Anthropic. BigTech are not only funding the AI chip demand driving datacenter expansion and Nvidia’s stock price, but fueling the survival and future of these expensive Research labs building the LLMs and smaller models.
The only other AI startup with similar levels of funding is xAI which is like being funded by the BigTech known as Tesla, that isn’t even a real technology company but an automaker. Monopolies (and tycoons) are driving and controlling the next era of AI monopiles.
U.S. Nationalism appears to be hurting other regions of the world in terms of innovation and economic leverage in an unbridled grab for power and datacenter colonialism by Cloud Giants. Just ask Europe and China.
It’s not clear how much Amazon could invest in Anthropic’s next round but it could be a lot. Microsoft has invested $13.75 billion in OpenAI since 2019 but that’s not counting the latest absurd valuation round. Microsoft Corp., OpenAI’s largest backer, put in about another $750 million. OpenAI gives 20% of its revenue to Microsoft and that will become 75% of its profits if it ever becomes profitable (expected in 2029). It’s not clear how OpenAI could legally execute its AGI clause (given that it’s a marketing gimmick).
The problem of the AI arms race in an era of U.S. national and exceptionalism is very dystopian. Learn more about AI Nationalism. So we can assume Microsoft has invested north of $14.5 Billion in OpenAI now, probably over $15 Billion.
On the topic of AI Nationalism:
🌟📋🌱 (October, 2024)
Popular Articles on AI Supremacy
“Who to follow in AI” in 2024? 🎓
Prompt Engineering from 2020 to 2025 ✨
AI for Non-Techies 🛠️: Top Tools for Meeting Notes
Goldman Sachs and Economists are Backtracking on Generative AI’s Value
OpenAI and Microsoft Legal Battle
The race for AGI is getting complicated in the Altman regions. The two companies each hired investment banks to help answer how Microsoft’s roughly $13.75 billion worth of investments in OpenAI since 2019 will translate after the startup converts from a nonprofit to a for-profit corporation, according to the Wall Street Journal.
How much of OpenAI does Microsoft own exactly? That probably has yet to be determined for 2025. What will the antics of Ed Zitron say about it? Both the critics and fan-boys of OpenAI are missing important details.
This has the vibe of something heading to a hotly contested legal battle, with a warning sign of the talent exodus. OpenAI hired Goldman Sachs to advise it during the process, and Microsoft brought on Morgan Stanley. I don’t like the look of this Uncle Sam!
The negotiations over Microsoft’s ownership stake come as OpenAI’s valuation has soared. OpenAI is now worth on some kind of paper, $157 Billion. While reports suggest Microsoft is entitled to up to 49% of the for-profit arm’s profits, this does not equate to a 49% ownership stake. For the record, their last funding round was another $6.6 Billion. Which won’t last long at the current levels of burn. Losses in 2025 and especially 2026 are expected to soar.
There may be unusual strings attached with both the latest funding round and Microsoft’s equity stake that have not been made public. As OpenAI transitions from a non-profit to a for-profit structure, both companies are engaged in complex negotiations to determine how Microsoft’s previous investments will translate into equity. OpenAI’s products compete against Microsoft’s own products, to make matters worse. ChatGPT likely left Microsoft’s various Copilots dead in the water. Having early access to OpenAI’s models doesn’t appear to have been very useful for Microsoft. It also makes them look incompetent, which we already knew they were in AI.
Sam Altman could get as high as a 7% equity stake in the new restructuring as investors want him to have more skin in the game. After Microsoft and Snake oil Sam get their share, how much is left? Not a hell of a whole lot to be honest.
Mistral and Cohere CEOs respectively.
Exit or Die
With several Generative AI companies already being acquired, there are fewer and fewer first wave Generative AI startups left playing on the monopoly table. Inflection, Adept, Character.AI, Aleph Alpha and Stability AI are no longer seriously in the game any longer. How much longer can model builders with limited revenue like Cohere, Mistral and others survive?
Sovereign AI startups are sort of on their last legs. This stuff is getting expensive. The cost of talent, cost of compute, competition, the market and the lack of revenue generating products is a problem. Cloud companies can push demand for AI chips higher, but for Generative AI startups AI isn’t just eating software, it’s eating itself.
By 2030 how many of these Generative AI startups will even be left? Which are next to get acquired? BigTech is damaging the innovation spectrum of America as U.S. Nationalism has become the major priority. Of course this has been going on for some time as the American internet is dominated by the same names for at least a decade. AI startups are basically just glorified research labs for these Cloud players. The only play is literally to sell out to them. Ask Mustafa Suleyman if he’s building real products at Microsoft? He’s mostly selling his latest book. It’s embarrassing to watch the turmoil and lack of real innovation in the founding years of the great datacenter boom among AI startups. If Tesla’s robotaxi program was itself an AI startup, it would have failed a long time ago.
U.S. vs. China in AI’s End-game?
Curiously as U.S. based Generative AI is eaten by the Cloud Titans, this leaves more market share for Chinese AI startups. American BigTech is the U.S. flavor of a great firewall. But it’s fundamentally defensive which over time erodes, and it’s only a matter of time. Just look at the models of Alibaba Qwen or 01.AI to understand what’s happening on the ground.
Recode China AI
By (related to Baidu).
I hope to get Tony to share more guests post with us.
AI Nepotism is the new Flavor of Dystopia
Here’s the dilemma right, how much of all of this is even real or sustainable? If OpenAI’s fate is to be owned in part by Microsoft and Anthropic’s fate is to be owned in part by Amazon? You need a sponsor? Now you are no longer truly independent. The tech oligarchy is like the new nepotism that rules the world. All of AI innovation is governed by it.
AI slop everywhere, a lack of consent and debatable value even for consumers are all rather serious symptoms of the age. This is not exciting for young people or giving them much of a sense of hope in civilization’s future. More AI monopolies owned (and governed) by Cloud Titans and Ad duopolies is probably not going to build a great new world.
Generative AI is not a MAGA Moment
With forced RTO at Amazon and more layoffs at Meta, even techies are leaving BigTech. This is not what we worked so hard to build, AI Nationalism and BigTech authoritarianism will have a steep price for society and the future. When you prioritize profits and National moats at all costs, what consequences are you putting into motion?
Hint, it’s not free market capitalism and democratic values any longer. How Generative AI is unfolding is making that abundantly clear.
Who are the AI startups that look Sustainable?
Based upon my own observations and my opinion only:
Suno AI (see Mikey Shulman)
ChatGPT as the flagship product of OpenAI.
Glean (I wrote about them here).
ElevenLabs (a category winner)
Perplexity (an acquisition target)
Midjourney (bootstrapped so more sustainable)
Runway (I predict a category winner)
Poe by Quora (among the most useful AI apps)
Udio (a competitor of Suno AI)
Scale AI (has grown up with the Gen AI ecosystem)
Anduril (a National Defense winner, e.g. smarter drones)
Synthesia (AI Avatar category, Nvidia backed)
Etched (speculative very early stage AI chips)
Cerebras Systems (plans to go IPO soon)
Microblink (Croatian creator of the Gen AI app Photomath)
HeyGen (Video generation and personalization)
Meitu (Chinese, app has Gen AI features. It’s model is called “MiracleVision”)
Gauth (homework app, made by ByteDance)
Last words on AI Startups
Big funding rounds are not a great sign for the sustainability of these startups for the most part as many of them struggle to generate real-world revenue in a realistic timeframe. Consumer behaviors are changing rapidly and B2B and Enterprise AI revenue is a safer bet. The AI talent needed is very expensive and each use case is clogged with competitors.
The idea that this all translates into mind-bending real innovation isn’t actually true. The failure rate of the first wave of Generative AI startups is going to be a stark reminder of hype trends. Will solutions and startups from China fill the void as Cloud Titans poach and grab? Certainly if ByteDance is going to have its way, who understand how to scale apps globally.
Western social media is full of half-truth and exaggerations about the Generative AI products and ecosystem. Most of the AI News are just public relations campaigns. As the dust begins to settle there won’t be many companies or startups left standing as VCs gamble at their own risk. BigTech is eating AI and it’s a kind of a downward spiral for the world at large.
Even the most diehard techno-optimists (pretend or otherwise) will have to give way to the reversion to the mean of techno-realism likely by 2027. So few winners are emerging and even most of these won’t exist in five to ten years. Generative AI has made it faster to grow but also likely shortened the lifetime of a company in many cases.
Read More in AI Supremacy
