Hacker News new | past | comments | ask | show | jobs | submit
GodelNumbering | on: Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs
So, in last several months, all the prominent names Google lost: Demis Hassabis (technically still with google but these things are usually presented with a spin), Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, Quoc Le, Noam Shazeer, John Jumper, Jonas Adler, Alexander Pritzel, David Silver, Denny Zhou, Fernando Pereira, Alex Turner

And all the prominent names Google gained: NULL

Combined with no gemini frontier GA release in about 14 months. You have to have created an environment pretty hostile to innovation for this to happen

hintymad | parent | next
This funny post reads surreal, but it may carry some truth: https://x.com/signulll/status/2067446889956430273?lang=en
JumpCrisscross | root | parent | next
https://xcancel.com/signulll/status/2067446889956430273
yyaakkqq | root | parent | next
I wish their deep tech teams moved as fast as the privacy violating ones.
loading story #49198795
loading story #49199803
bjackman | root | parent | next
this prompted me to search and find: https://addons.mozilla.org/en-US/firefox/addon/xcancel/
spwa4 | root | parent
Thank you. Twitter has become such a terrible website in the last years. Well, since Musk, really. Real disaster.
snthpy | root | parent
Yes, but now we have AI to summarize tweets.
adityazero | root | parent | next
> sundar, softly: “we can create a permission working group.”

so true. i have attended meetings to decide on meeting topics for the next half.

imglorp | root | parent | next
At $work, r&d once had a retreat with a consultant to discuss why our productivity was too low per some spreadsheet. The consultant was told how everyone is in meetings all day so we all stole time from our families at night to get any work done at all. :(
brobdingnagians | root | parent
Someone I know who did a lot of consulting said consulting was very easy 1. Go in and talk to employees about what is wrong 2. Tell management what the employees said without mentioning it came from the employees (so that management wouldn't dismiss it out of hand) 3. Charge a lot of money so management felt like it was very valuable feedback
loading story #49199744
loading story #49200323
atmosx | root | parent | next
That’s how the Brussels bureaucracy works. There is even a TV series about it.
Melatonic | root | parent | next
Sounds like its time to schedule a meeting to talk about how you can reduce the number of meetings
Bendy | root | parent
Reminded me of a bit from the GOAT political comedy Yes, Prime Minister:

"But he's the Prime Minister!"

"Indeed he is Bernard. He has his own car, a nice house in London, a place in the country, endless publicity and a pension for life. What more does he want?"

"I think he wants to govern Britain."

"Well stop him, Bernard!"

titanomachy | root | parent | next
Oh man, that show has some amazing writing.

"Tobacco-related diseases cost the NHS £150 million per year!"

"Yes we've looked into that, it turns out that if those people had survived they would have cost us billions in pensions and healthcare costs! From a financial perspective, it's vastly preferable that they continue to die at the current rate."

rob74 | root | parent
This is indeed what the current US administration is failing at: keeping Trump from governing the US...
galangalalgol | root | parent | next
I would not have chosen the word governing for that sentence but I'm not sure what fits best. Plundering, pillaging, gutting, ransacking, Rick rolling, despoilng, or perhaps molesting.
sysguest | root | parent | next
...molesting? a country? ...how?
loading story #49203125
MyHonestOpinon | root | parent
"If you are famous they let you do anything!"
shevy-java | root | parent
Trump is indeed effective at pillaging. The interesting thing is that Yuri explained and predicted (!) this in the 1980s: https://www.youtube.com/watch?v=yErKTVdETpw
neutralx | root | parent | next
I can see this as a screenplay for one of the KRAZAM[0] videos.

[0]: https://www.youtube.com/@KRAZAM

utopcell | root | parent | next
I remember a time when a tweet was 140 characters.
ethbr1 | root | parent | next
140 characters was likely more efficient, in terms of forcing people to focus their thoughts.

But it was unpopular, in that people don't like mental effort.

conception | root | parent
No people just posted around it (1/36).
jorisw | root | parent
I remember GemTOS
dnchdnd | root | parent | next
So believable until the last line, alas Gemini being this useful is completely implausible.
nunez | root | parent | next
Nice post, but this reads as the Google of yesterday and the OpenAI of some near-distant future.
throw-the-towel | root | parent | next
This is so good, it gave me flashbacks.
_the_inflator | root | parent | next
You made my day. As tragic as it reads, it is on point.
loading story #49196966
loading story #49199900
loading story #49197791
trueno | root | parent | next
lmao that's actually so good
loading story #49197202
deepsun | root | parent
Regardless of how much I hate Musk, he did a great job of firing 80% of Twitter staff.
maeln | parent | next
To be fair, I think it's pretty hard as an AI company to secure your top worker right now unless you have a significant equity compensation. Once your name is known, and you can say you are/were "top AI researcher at OpenAI/Google/Anthropic", you can probably just make you own company and raise enough money that even if the company fails, you will probably make more off of it than what you would have at your previous place of employment.
loading story #49198350
loading story #49198088
squidbeak | parent | next
Gemini 3 Pro was legitimately frontier and released 8 1/2 months ago, not 14.
mips_avatar | root | parent | next
It was inspiring watching the Gemini 3 pro team launch the model and bike away into the sunset never to launch anything again
loading story #49196961
egillie | root | parent | next
I remember after that release everyone was saying "well obviously google is going to win this, we all knew it, they have the data and the infrastructure"
zarzavat | root | parent | next
All of that is true, Google should be winning. We underestimated Google's determination to fail.
galangalalgol | root | parent | next
Is Gemini really that bad? I mean, I would always prefer anthropic for agentic coding, and apparently openai has the math thing cornered, but if I'm using ai like a search engine gemini does that just fine. That may be all they were actually aiming for whatever else they say.
scottyah | root | parent | next
Gemini works amazingly well, and is better suited for most use cases. When most people speak of it negatively (especially here), it's just with the task of coding, which is a small subset of what google should be training its AI for.
wongarsu | root | parent | next
If you need a less censored model that's still pretty smart and capable of some lateral thinking, it's great

Same niche as Grok, which is also great at search and less censored. Gemini is maybe a bit smarter and better on long contexts

trollbridge | root | parent | next
It’s not bad. It’s just not better than other models. It’s very much in the same class as Grok, Cursor Grok Build, etc.
WarmWash | root | parent
It's vision capabilities are still top notch, and it seems pretty natural that most everyday human users are going to want strong vision over strong language/coding.
ethbr1 | root | parent | next
One of the most corrosive effects of being large and old (and Google is old by this point) is institutional inertia and individual internal optimization.

The former caused by the latter.

And the latter a problem once you begin to get executives and leaders who have mostly worked inside the company, because then they subconsciously prefer the Company Way(tm) to alternatives.

And Google has some previously-optimal, now-detrimental company ways.

0xpgm | root | parent | next
From my understanding (I may be wrong) what is happening now is that frontier labs are doubling down on doing RLHF on their models using the user interaction with their own tools e.g. Anthropic making their models better at producing code with claude code. Perhaps the top Google researchers do not see this kind of narrow focus as a path to more generally beneficial AI technology.
weatherlite | root | parent | next
Is it failing ? Take a look at the stock price
Filligree | root | parent | next
Their stock price won’t matter once OpenAI launches the paperclippers.
loading story #49193937
ghtbircshotbe | root | parent
I feel like the majority of ai use by normal people is through Google search, so that's kind of a win.

I'm confused by meta's capex though. Are they planning on becoming a cloud provider, or just throwing money away like with the metaverse?

WarmWash | root | parent
In an interview, Zuck said that the cost of not having super intelligent AI is far more than the cost of trying to develop it.
marcuschong | root | parent | next
What's more annoying is that so many people say it with a passion. Like it's about their favorite soccer team or something.
DrewADesign | root | parent
More like sects of a major religion. Or, a minor and incredibly expensive religion, in this case.
puzzlingcaptcha | root | parent
Tribalism is human nature, for better or (usually) worse.
loading story #49196267
TacticalCoder | root | parent | next
Google owns 15% of Anthropic. They'll be winning either way.
ozgrakkurt | root | parent
It is pretty surprising if you see this as “pretty much everyone has figured out the way to success on any field”. Our way of seeing life is really bizzare
GodelNumbering | root | parent | next
Correct, but it was preview release. I was referring to GA in my comment.
notatoad | root | parent | next
Does that mean anything at all? They released it. Customers paid for it. Sticking a preview label on it doesn’t change that it was released
npn | root | parent | next
btw, GA usually means it is suitable for production, and available globally.

the last time a gemini model was available globally in multiple datacenters was gemini 2.0 era.

aenis | root | parent
Use google pre-ga stuff at your own risk. Fine for POCs and cool demos, absolute misery in scaled prod.
loading story #49197749
clhodapp | root | parent
That's another problem Google have: they won't stand behind their products until they are outdated.

That can work in the B2C space but it's horrible in B2B.

aleph_minus_one | root | parent
> That's another problem Google have: they won't stand behind their products until they are outdated.

> That can work in the B2C space but it's horrible in B2B.

My experience differs: (conservative) companies like stability, so calling some very new model "Preview" is a good idea to make it clear to the customer that this frontier model should be treated as more experimental than the default offering.

loading story #49194944
clhodapp | root | parent
That's fine, but it should still have the same technical availability level
loading story #49191926
chaostheory | root | parent | next
It’s a good substitute for search, but its effort is so low that it’s near useless for work compared to the other frontier AIs. Gemini focuses so much on response speed that everything else suffers.
mountainriver | root | parent | next
It was always benchmaxxed
Analemma_ | root | parent | next
I mean it was legitimately frontier for all of a week or two, and then OAI and Anthropic made better releases, and then did that several more times over the year. Google’s pace is not cutting it.
frollogaston | parent | next
Part of me thinks Google's entire problem is crappy internal tooling, not really an anti-innovation environment. Just making a dev take 2x as long to get something done has a bigger effect than you'd think. With LLMs it's more like 10x now because even Gemini doesn't understand Google-internal tooling.
Cidan | root | parent | next
Google's internal tooling is still, hands down, better than anything that exists for the scale they operate at, and it's not even close.

Google's processes, however, is hands down the worst thing to exist for the scale they operate at, and it's not even close.

loading story #49200017
bushbaba | root | parent | next
the processes are a result of repeated lawsuit
| root | parent
{"deleted":true,"id":49194902,"parent":49191567,"time":1786012523,"type":"comment"}
coderatlarge | root | parent
except for most other companies in the world.
fooker | root | parent | next
Google has close to the best internal tooling in the industry for a decade or so.

Then the Google engineers who joined Facebook missed it so much that they built a better replacement.

frollogaston | root | parent | next
Replacement for what? Google has some good internal tools, mostly the older ones. They're lucky to be using React instead of Angular at Facebook though.
nostrademons | root | parent | next
Angular was an acquisition and isn't all that commonly used within Google.

Closure was the homegrown framework that's everywhere. It is a different flavor of bad than Angular. Angular was basically "You too can make your Javascript look like HTML", Closure is "You too can make your Javascript look like Java", and nobody bothered to ask Why? For that matter, React was "You too can make your Javascript look like Ocaml." JQuery was the only framework that really let Javascript be Javascript (other than writing in vanilla JS, which post-ES2015 wasn't as insane as it sounds).

frollogaston | root | parent | next
And then there's GWT which was "you can write webpages in Java." Yeah I never really kept up with the web frameworks there, just knew that for non-customer-facing things we were always recommended to use Angular, and that Wiz also exists.
loading story #49195173
mikaraento | root | parent
Though to be more charitable closure was a direct reaction to the difficulty of maintaining a huge JavaScript codebase for Gmail and static typing like Java really helped. Typescript is better but closure was miles better than raw js.
flymasterv | root | parent | next
React? Lucky. I wrote Closure (with an S) at GOOG in anger 5 years ago, and only stopped because I switched teams.
dehrmann | root | parent
It's more like Buck and Tupperware.
mikert89 | root | parent | next
google had the best tooling a decade ago...
fooker | root | parent | next
Okay yeah, fair point.

My comment was from a decade old perspective

fragmede | root | parent
Who has the best tooling now?
loading story #49191891
swrrt | root | parent
Sounds like a big-company issue, large scale lead to large burden and complexity.
solenoid0937 | root | parent | next
I have a friend at Google DeepMind who tells me that Google believes in AGI/superintelligence just as much as HN does, which is probably why no one with conviction wants to work there.
aleph_minus_one | root | parent | next
> I have a friend at Google DeepMind who tells me that Google believes in AGI/superintelligence just as much as HN does, which is probably why no one with conviction wants to work there.

With "believes in AGI/superintelligence just as much as HN does" do you mean that they are superbelievers or rather sceptical of AGI/superintelligence? I have seen both positions on HN.

sheeshkebab | root | parent | next
What’s with fascination with agi/superintelligence? It seems people working on it never had kids and just want to compensate for that. It’s a really horrible thing to try to “solve”.
mdp2021 | root | parent | next
> What’s with fascination with agi/superintelligence

We had relative strength, we needed more strength, we built cranes;

there was something like intelligence, we needed more intelligence, we sought to fill the need;

we were rightly fascinated with intelligence, we studied intelligence itself, we wanted to build it...

And then somebody built things that looked like intelligence, and very rightly some said "Oh, now we have to get to the Real Thing with urgency".

solenoid0937 | root | parent | next
I think death, and most extreme pain/suffering should be optional. Superintelligence, if done safely, lets us solve most of our problems.
sumeno | root | parent | next
The vast majority of pain and suffering in the world is already entirely optional, but as people we allow it to exist. How is another computer going to fix that?

It is far more likely that whoever controls a super intelligence uses it to gain more power and inflict far more suffering

loading story #49201403
a2ff6eeb0 | root | parent | next
I agree, we're doing a wonderful job of safely letting the leaders of various countries and companies, at their option, cause death, extreme pain, and suffering. Look at the innovation of AI companies in the police and military sectors.

I'm extremely bullish on our future ability to make autonomous death an option for anyone.

ethbr1 | root | parent | next
If humans could politically-economically deploy superintelligence safely, then we'd already have less extreme pain/suffering.

Unless there are countervailing forces, it will be deployed, capital will hoard the benefits, and everyone else will be told to fuck off.

I don't have faith in any of the AI labs to make hard financial decisions to deploy hypothetical future AGI in a way that's good for humanity as a whole.

There are too many incentives against, including extreme personal financial incentives for key AI lab stakeholders against.

And if we should take anything from tech history, it's that an exceedingly small number of people look fuck-you money in the face and say "Naw, I'd rather do what I believe in."

loading story #49197168
tomashubelbauer | root | parent | next
What is your position on middle of the road pain and suffering?
loading story #49195941
sheeshkebab | root | parent | next
What? psychologists are for that (or drugs and alcohol).
loading story #49197171
| root | parent | next
{"deleted":true,"id":49194835,"parent":49194210,"time":1786012014,"type":"comment"}
vasco | root | parent
I for one think that's it's been too long that human suffering is alone. Since the machines will take our jobs, they might as well suffer while they do it.

It's like the old adage: "If you make AGI you just want to watch computers cry"

frollogaston | root | parent | next
How much belief is that? I tend to skip the HN posts that are about AI.
| root | parent | next
{"deleted":true,"id":49194917,"parent":49188074,"time":1786012618,"type":"comment"}
solenoid0937 | root | parent
Most of HN does not take the idea of superintelligence seriously, and until this year did not take the idea of AGI seriously.

I think LessWrong is a much better community for rational takes on AI, they've been reasoning about these risks for years under a much more sound logical framework

loading story #49188478
loading story #49189175
loading story #49188245
loading story #49189477
Zigurd | root | parent | next
A few months ago I heard Demis Hassabis say something, albeit vague, I doubt he truly believed regarding AGI. That AGI is relatively near is the party line everywhere. That may contribute to creating toxic environments and lousy investment decisions. So here we go with the FOMO.
oblio | root | parent
I think HN believes in AGI. HN probably doesn't believe LLMs will lead to AGI.

Also lots of tech people, HN included, are waking up to technology not only including penicillin (net positive for humanity) but also dynamite (best case: net neutral).

Nifty3929 | root | parent | next
Dynamite has been incredibly beneficial for humanity.
loading story #49189529
loading story #49189988
Zigurd | root | parent
I believe in AGI to the extent that if what's going on between your ears isn't happening on a network of neurons, it's magic. And I also believe the idea that AGI is near is based on the emergent capabilities of LLMs. There's a chance that AGI will emerge from bigger faster better LLMs. But without a theory of when and how that will happen, I'm not counting on it.
summerlight | root | parent | next
It's not a tooling problem. It's more of a layer of policy problems. If you realize that you cannot run a simple experimental code even in non-production environment for weeks due to 10s of privacy, security, access, process and legal issues where you gotta collect a bunch of approvals, this is critical. And the problem gets worse because the tooling is too good when it enforces. There used to be some holes and circumvention which are all gone these days. This is probably why they said "the infra is good for services but not for research".
frollogaston | root | parent
I don't even think it's good for services. It's not like you go through cumbersome reviews/tools and then things are safe. They have insane homemade config languages and obscure systems that 99% of SWEs don't really understand but won't say it out loud. That's how they dropped cns2, and the postmortem is never going to blame the tools.
vkou | root | parent | next
The tooling is not the problem. If shit takes forever to launch, it's because there are many stakeholders that need to be satisfied (some for security, some for regulatory, some for the kinds of politics you get in a company that employs almost half a million people.)
frollogaston | root | parent
But GDM isn't gated on launches. They were freely releasing things internally for dogfood. Problem is that stuff was just not as good as the competition.
vkou | root | parent
A particular Google product being shit is a data point, but is orthogonal to my opinion about why it is slow to launch products/features.
vintermann | root | parent | next
Internal tooling? Didn't Jeff Dean write their internal tooling?
frollogaston | root | parent
He wrote the good old parts
vintermann | root | parent
And Tensorflow?
loading story #49193759
mhh__ | root | parent | next
If the tools are all in the same monorepo idk if that's actually true
frollogaston | root | parent
What do you mean about them being in the monorepo?
r0b05 | root | parent | next
Maybe they are forced to use Google search.
frollogaston | root | parent
No, they have an actually good internal search. And there are some good things like stubby, but again pretty annoying that Gemini doesn't understand stubby.
VirusNewbie | root | parent | next
> because even Gemini doesn't understand Google-internal tooling.

this is false, it's very good at internal tooling.

frollogaston | root | parent
Only the GFG models know anything internal. Regular Gemini isn't trained on any of that. And GFG is a much older base model, so people use the regular one. If the tools seem to handle google3 code ok, it's only because of skills and not the model itself, and then you run into issues with skill bloat. Sometimes the A/B test would give me the bad model of the day that'd try to grep all of piper.

Start in a blank directory and tell it to spin up a boq Scaffolding stubby server that responds with "hello world." Unless something has changed after I quit a few months ago, it won't know how to do that locally, let alone actually deploy it. Try the same outside Google with like a Flask server on AWS or GCP.

kridsdale1 | root | parent | next
Your information is indeed out of date.
loading story #49191515
rryan | root | parent
No longer the case
flymasterv | root | parent | next
Thank you!

Google’s tooling was, hands down, the worst I have ever encountered. I did 10 years at GOOG, 3 at AMZN, 4 in research, and another 5 at companies you have heard of but wouldn’t be impressed by, and every day GOOG infuriated me.

rixed | root | parent
You don't measure tooling quality with devs' enjoyment, though, but with what the tools make possible.

Technical merit is not correlated to popularity, after all.

frollogaston | root | parent
Made it possible for a cronjob to take 7 days' wall time to set up
Nimitz14 | root | parent
Uh isn’t google known to have the best tooling in the world
frollogaston | root | parent
They earned that reputation in like 2005. Some people have been there so long (without doing side projects) that they don't know what non-Google tooling looks like in this decade or even previous.
HlessClaudesman | parent | next
Maybe Google's decision to waste everyone's time and pollute the truth ecosystem by pushing half baked AI assertions into search was a mistake.
ed_elliott_asc | root | parent
I quite like the ai/search thing - I like that there is a super accessible ai I can ask questions without infecting my ChatGPT history
krige | root | parent | next
The answers it gives are often trash though, even more so the whatever lobotimized model they slap on top of google search.
| root | parent
{"deleted":true,"id":49195472,"parent":49195063,"time":1786017512,"type":"comment"}
loading story #49197590
ShadowOfThePit | root | parent
Why not just use temporary chats for ChatGPT?
overfeed | parent | next
> And all the prominent names Google gained: NULL

As an outsider, Google seems to have a knack for minting prominence for their talent.

geodel | parent | next
Did all these names were huge before joining Google, and Google used their extraordinary hiring skills to get them? Or a whole bunch of them had challenging problems to solve and ample resources at their disposal to become huge?

To me it is good thing in either case. Extraordinary people are leaving to make even faster research and development. A lot of talent in Google hitherto unnamed is going to get chance to shine.

alfalfasprout | parent | next
Sundar Pichai began the downfall of Google by most accounts. It went from an engineer driven culture to one of infighting, politics, and bureaucracy. It's telling when xooglers nowadays complain of maintenance and improvement of products being a career dead end vs. shipping new things. And there being so much empire building.
amelius | parent | next
Some of these ex-Googlers apparently are starting a new company. Any news around that? Curious what they're up to ...
| parent | next
{"deleted":true,"id":49191284,"parent":49185997,"time":1785979632,"type":"comment"}
Kalanos | parent | next
Gemini has been improving by what feels like 1% every week
xnx | root | parent
68% per year is pretty good!
loading story #49200817
nolist_policy | parent | next
> Combined with no gemini frontier GA release in about 14 months. You have to have created an environment pretty hostile to innovation for this to happen

The Trump admin is now regulating frontier models though.

shadowtree | parent | next
All this talent has delivered... Gemini.

Oh well, emperors, clothes, ... you know.

gizajob | root | parent
Gemini is arguably the fulfilment of what AskJeeves promised and never delivered, nor did the rest of Silicon Valley succeed in natural language search and question answering for the 30 years it took to finally arrive. Gemini works great for answering questions and delivering answers for probably 90%+ of the things that people are going to ask Google for, while running on Google’s TPUs paid for with Google’s profits instead of hyper expensive Nvidia racks funded with VC Hopium, and integrated seamlessly free of sign ups to novel portals.
shevy-java | parent | next
Ever since Google became an adCompany, it rarely innovated anymore.
loading story #49196889
loading story #49200303