Content About Stocktwits
No Image Available

'Striking Out' at Google. Mostly in a Good Way. ARD #135

Author
Michael Parekh AI: Reset to Zero
18 hours ago
15 MIN READ

Today’s theme cuts both ways. ‘Striking out’, in the good sense and some think bad.

A bunch of people at Google are striking out on their own, in new directions. And Mr. Market is asking the other version of the question: is Google striking out on its AI opportunity with all these people leaving?

Three events today, all at Google, and my take on each. All in this AI Tech Wave.

(1) Sir Demis Hassabis Steps Up, His Number Two Takes the Models

MP TAKE: As I wrote back in May, Hassabis was already stepping up to ‘the foothills of the Singularity’, aka AGI. This one reads as succession by elevation, not an exit.

Google DeepMind’s Demis Hassabis, the founder/CEO of DeepMind and head of AI for Alphabet/Google until now, is stepping down as CEO. And his number two, the chief scientist, is becoming his replacement.

Demis has been the head of AI across Google, reporting to CEO Sundar Pichai. He is a Nobel Prize winner. He also leads Isomorphic Labs, the AI drug discovery company spun out of DeepMind under Alphabet.

The model development consolidates under Koray Kavukcuoglu, DeepMind’s longtime technology chief. He takes responsibility for all of the AI model development, reporting up to Sundar Pichai directly.

Sundar had positive words on all of it: “We have to continue to move fast with clear purpose,” and “We are super focused on these areas where we need to improve.” He also reinforced Google’s commitment to competing at the frontier, a question people have raised with Anthropic and OpenAI setting the frontier pace.

Remember as mentioned above, Hassabis has talked about Google ‘approaching the foothills of the singularity’. The Nobel laureate founder has been moving toward the science from his operational roles for a while, and today’s events basically formalize that. The founder becomes chairman and chief scientist, the number two takes the models.

We are likely to see other management shifts from Google’s deep bench across its various entities, particularly in light of event number two below. And all of this keeps Demis’s focus on drug discovery, which has been his primary interest, in things like AlphaFold and all the wonderful innovations Google has done under DeepMind.

So Hassabis is striking out upward. That is my good version of striking out.

Sources:

For longtime readers:

(2) Four Google Builders Strike Out to Found Discovery Loop

MP TAKE: The builders are striking out. And note carefully who is investing.

A number of Google veterans are striking out to found a new company called Discovery Loop, in which Google itself will be an investor, alongside a who’s who of venture capitalists.

Four of Google’s most important builders, some who have been at the company from almost the very beginning, are leaving to start their own company. The mission is recursive self-improvement, or RSI.

This is a hot new area. I have written about it in earlier posts. It is supposedly the next big thing after AI reasoning and AI agents, on the way to AGI. The idea is that the current models can teach future models in a recursive loop. There is a lot of science that needs to happen here. And there have been a number of startups, some with other people from Google and other tech companies, founded at multi-billion dollar valuations. Discovery Loop is now one of them.

The key people leaving. Jeff Dean, Google’s chief scientist, literally employee number 30, there since 1999. He helped build much of Google’s best technology, things like MapReduce, BigTable, Spanner and the TPU architecture. He co-founded Google Brain in 2011. He will be Discovery Loop’s CEO. And Sanjay Ghemawat, his number two for over two decades at Google, joins him, along with a couple of other senior scientists.

Google people striking out have been in the news all year. In recent months Google saw Noam Shazeer, the lead author of the iconic 2017 ‘Attention is All You Need’ paper that kicked off this entire AI tech wave, including redirecting OpenAI onto its current path. The T in GPT is literally the Transformer that Noam and the team of eight wrote up in that paper.

John Jumper, who shared the AlphaFold Nobel Prize with Demis Hassabis, went to Anthropic. Peter Norvig, Google’s head of research for 25 years, joined Recursive Superintelligence, another RSI startup. There are two or three other companies founded of late with ex-Google people.

So this is now being seen as a trend. And the other aspect: Google is an investor in most of these startups. As is Nvidia, by the way. I have talked about Nvidia being the ‘Kingmaker’, using its immense free cash flow to invest in the most promising AI companies further up the tech stack. That is one of the things going on here too.

Sources:

For longtime readers:

(3) Is Google Striking Out on AI? Mr. Market Asks Again

MP TAKE: I have covered Google’s AI ups and downs for years. I have been, and remain, very constructive on Google’s opportunities in AI.

The question has a lot of echoes, because it has been a theme ever since AI got started with the ChatGPT moment almost four years ago. There was a code red at Google about three years ago, right after ChatGPT was introduced. Google then went to work, built Gemini and all the technologies around it, took the leadership back in a lot of these areas, and induced a code red of its own at OpenAI last fall.

And now, with the frontier models running away with AI agents and AI coding, the question has returned on Google. The stock reacted negatively to some of these departures and changes, down a little over 5%.

Here is my take on that. Google, along with Apple, is going to be one of the core leaders of consumer AI at least over the next two or three years, because of its immense vertical stack, from chips all the way up to applications, serving billions of people daily across search, mail, photos, maps and half a dozen plus applications.

And Google has a very deep bench. Sure, a lot of these people are very high profile, and they are striking out in new directions. But Google is a unique platform. And one of the things about a platform at a time of tremendous change and tremendous opportunity is that it also attracts people, which gets less attention. People coming in to Google, because of the unique set of platforms and the scale at which they can participate.

Sundar Pichai has done a great job, and he is only in his early 50s. There will be questions about whether these changes mean a recast succession plan. There are articles talking about that, in the sources below. Potentially Thomas Kurian, the head of Google Cloud, maybe a candidate down the road. But Pichai has still got a long way to go at Google, and is currently running high, both internally and in the views of the market externally.

So Google, I do not think, is striking out on AI. It is early days. My read is that they have a lot of opportunities ahead, and these are ebbs and flows in the marketplace.

Sources:

For longtime readers:

My overall take on all of this at Google, happening almost concurrently: some of the founders are stepping up. The builders are striking out on their own. And Google itself is leaning into the AI Tech Wave in every possible way with Gemini.

The bench the company is keeping is producing. Yes, some of the people are leaving, but a lot of the people staying will have a lot more opportunity with these changes. So stay tuned for additional things happening.

Tactically, one of the things we will be watching for is Google Gemini 3.5 Pro, their AI coding related effort, which has been delayed a few months. I have written about that, and we will see how it progresses shortly.

So I still believe Google is striking out in new directions totally, mostly in a good way.

GADGET AI

Google Assistant Hands Off to Gemini

MP TAKE: This is a positive direction, and a big move years in the making. It also involves Google AI and Google the company.

Google is retiring Google Assistant in September. It has been their equivalent of Apple’s Siri and Amazon’s Alexa on the Echo, the thing we use on our Google apps and devices to ask questions of the world. On the wake words, hey Google, it answers all kinds of questions. The weather, the time, set timers, play music, and a lot of other things besides.

They are replacing it with Google Gemini. So a large language model replaces Google Assistant, which has been based on deterministic computer science, built on core Google search technologies over the last couple of decades. And it is going to be effective across Google’s platforms, whether search, Google TV, or Google in cars.

My take: Google has built enough hybrid technology around Gemini and its large language models to be in a position to do this transition. Gemini will be able to answer a far wider array of questions, and will be one of the most competitive things serving billions of people going forward.

So a good transition, and something to keep an eye on. Stay tuned if your Google Assistant starts to behave a little bit differently, hopefully for the better.

Sources:

For longtime readers:

Q1. What has been my favorite use of Google Assistant?

I have been a longtime user of Google Assistant. And like most people, I have found it very capable, just like Siri and Alexa, on the weather and music and timers and the like.

But Google Search was generally under the hood. So I have had good use of the product on broader questions that the other products from Amazon and Apple sometimes could not answer. Those are my fond memories of Google Assistant. It has generally been more capable than some of its peers.

Q2. What is the core improvement I have wanted in Google Assistant?

It is an improvement I have wanted from all of them. They all have depended historically on wake words. You say hey G, A, or hey S. You know who I am talking about, each of the companies.

And typically you are referring to the device that is closest to you, but it is usually the device furthest out, in other rooms or across the room, that answers. That has been a frustration. There are deep technological reasons for it, particularly because each company uses siloed vertical technologies to build its capabilities.

As Google moves to Gemini, Apple improves Siri with AI, and Amazon does the same with Alexa, I fervently wish they would fix this issue of wake words and triangulation, of which device you are actually talking to. That would be my big wish for these devices going forward.

WRAP

Today’s AI-RTZ #1170 is up, on encouraging recent data showing students and schools starting to come around on AI. You will remember I have written a lot about this technology wave being received more negatively than past waves, particularly on fears among students for jobs, and among teachers around cheating. That is changing, with surveys from Gallup and others now moving in the right direction. It is something I posited would happen a couple of years ago, and now we are seeing the data. If you are interested, do take a look at that post today.

AI Ramblings Daily on AI-RTZ is here to think through AI and reset. Together.

Tomorrow, ARD 136 and AI-RTZ #1171.

Thanks for joining us today, AI Curious Folk. Stay tuned.

— MP

Full Source Reading

For the broader context, see the canonical sources for ARD 135, in today’s narrative order:

Event 1. Hassabis Steps Up

Event 2. Discovery Loop

Event 3. Is Google Striking Out?

Gadget AI. Assistant to Gemini

Clips from today

Clip 1. AI’s Next Big Thing: Recursive Self-Improvement

Four of Google’s most important builders are leaving to found Discovery Loop, and the mission is recursive self-improvement.

MP Take: RSI is the hot new area after AI reasoning and AI agents on the way to AGI, models teaching future models in a recursive loop. A number of RSI startups have been founded at multi-billion dollar valuations, and Discovery Loop is now one of them. Note who is investing: Google itself, top venture capitalists, and Nvidia playing its ‘Kingmaker’ role further up the stack.

Clip 2. Google’s AI Leadership: A Closer Look

I have covered Google’s AI ups and downs for years, and I remain very constructive on Google’s opportunities.

MP Take: Along with Apple, Google is going to be one of the core leaders of consumer AI over at least the next two or three years, on its immense vertical stack from chips to applications serving billions daily. The bench is very deep. And a platform at a time of tremendous change also attracts people coming in, which gets far less attention.

Clip 3. Google Assistant to Gemini: A Big Move

Google is retiring Google Assistant in September, and Google Gemini takes over.

MP Take: A large language model replaces two decades of deterministic search based technology, across Google’s platforms from search to Google TV to cars. Google has built enough hybrid technology around Gemini to make this transition, and it will answer a far wider array of questions for billions. Watch if your Assistant starts behaving a little differently, hopefully for the better.

Clip 4. Google’s AI Future: Pichai’s Leadership & Gemini AI

Sundar Pichai has done a great job at Google, and he is only in his early 50s.

MP Take: There will be succession plan questions, with some articles pointing at Google Cloud head Thomas Kurian down the road. But Pichai has a long way to go and is running high, internally and externally. Google, I do not think, is striking out. It is early days, and these are ebbs and flows in the marketplace.

About AI Ramblings Daily (ARD), and AI-RTZ

Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.

AI-RTZ is the morning text, a deeper written take on one idea, published by at least 5 AM EST. Today: post #1170.

AI Ramblings Daily is the afternoon video + podcast, my ad hoc takes and perspective on the day’s AI issues and news flow, around 20 minutes, with short 1-2 minute clips for quick topic views. Today: episode #135.

Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here.)





Want the latest?

Sign up for Michael Parekh's Newsletter below:


Subscribe Here

More like this

BROWSE ALL ARTICLES