DeepMind is Holding Back Release of AI Research To Give Google an Edge (arstechnica.com) 31
Google's AI arm DeepMind has been holding back the release of its world-renowned research, as it seeks to retain a competitive edge in the race to dominate the burgeoning AI industry. From a report: The group, led by Nobel Prize-winner Sir Demis Hassabis, has introduced a tougher vetting process and more bureaucracy that made it harder to publish studies about its work on AI, according to seven current and former research scientists at Google DeepMind. Three former researchers said the group was most reluctant to share papers that reveal innovations that could be exploited by competitors, or cast Google's own Gemini AI model in a negative light compared with others.
The changes represent a significant shift for DeepMind, which has long prided itself on its reputation for releasing groundbreaking papers and as a home for the best scientists building AI. Meanwhile, huge breakthroughs by Google researchers -- such as its 2017 "transformers" paper that provided the architecture behind large language models -- played a central role in creating today's boom in generative AI. Since then, DeepMind has become a central part of its parent company's drive to cash in on the cutting-edge technology, as investors expressed concern that the Big Tech group had ceded its early lead to the likes of ChatGPT maker OpenAI.
"I cannot imagine us putting out the transformer papers for general use now," said one current researcher. Among the changes in the company's publication policies is a six-month embargo before "strategic" papers related to generative AI are released. Researchers also often need to convince several staff members of the merits of publication, said two people with knowledge of the matter.
The changes represent a significant shift for DeepMind, which has long prided itself on its reputation for releasing groundbreaking papers and as a home for the best scientists building AI. Meanwhile, huge breakthroughs by Google researchers -- such as its 2017 "transformers" paper that provided the architecture behind large language models -- played a central role in creating today's boom in generative AI. Since then, DeepMind has become a central part of its parent company's drive to cash in on the cutting-edge technology, as investors expressed concern that the Big Tech group had ceded its early lead to the likes of ChatGPT maker OpenAI.
"I cannot imagine us putting out the transformer papers for general use now," said one current researcher. Among the changes in the company's publication policies is a six-month embargo before "strategic" papers related to generative AI are released. Researchers also often need to convince several staff members of the merits of publication, said two people with knowledge of the matter.
General Research and Capitalism don't mesh (Score:4, Insightful)
this is news? (Score:2)
Wow what a story. Nobody is talking about google AI so google announces that the reason for this is "because their AI is secret". Oooookay, i'm sure that's going to move the stock price. /rolleyes
Re:this is news? (Score:4, Insightful)
To be fair, lies likt this one work on many people. If I have learned one thjing fromt his AI hype, it is how many ways are there to pretend you have something incredibly valuable, when in fact, you do not.
Re: (Score:3)
So ya, I bet they're a bit sore about it, since they're not even the biggest player in the game.
If $200 billion a year isn't valuable, then I don't know what the fuck is.
Your take on this topic is, as usual, trash.
Re: (Score:2)
Currently, this is maybe an industry that _burns_ $200B per year, but not one that earns this sum. It is all unfounded hopes, greed and stupidity.
Re: (Score:2)
Party A pays Party B X dollars, that is revenue for Party B.
You can say that Party A burned it, but that doesn't fucking matter, because that's not how economics works.
Re: (Score:2)
Hit a nerve there, have I? I suggest you stop being stupid and let reality in.
Re: (Score:2)
There are many who talk about Google AI, but I agree with you that the amount of those is far less than what you would expect.
Google Deepmind is clear leader in AI Development. Other companies are so far behind that it is not even a race. Problem is that majority of humans don't know it. Other companies are doing cute little tricks with their AI, but Google is already making scientific discoveries with their AI, some small and some worth a nobel:
https://arstechnica.com/inform... [arstechnica.com]
https://aimagazine.com/articl [aimagazine.com]
It is all smoke and mirrors anyways (Score:1)
This posibly just serves to create a false impression of them having something more they do not publish. Of course, they do not. LLMs are niche specialist tools except for the two uses we already see and which only have gotten cosmetic improvements for over a year now: Better search and better crap. Nothing else of general use will come from them.
Re: (Score:2)
This posibly just serves to create a false impression of them having something more they do not publish. Of course, they do not. LLMs are niche specialist tools except for the two uses we already see and which only have gotten cosmetic improvements for over a year now: Better search and better crap. Nothing else of general use will come from them.
The type of people who think this way also convince themselves that no practical use cases and no profits have arisen from AI over the past decade. I imagine ten years from now they'll think the same about the delusional people who tricked the stock market into minting millionaires and billionaires based on nothing useful. They know better and are astounded that all these supposedly smart people are being delusional or are being gaslighted. They point to the many dead-end AI use cases to support their th
Re: (Score:2)
Re: (Score:2)
And where do you see "worth" in LLMs or exponential growth? Because all they do is _burn_ money.
Re: (Score:2)
And where do you see "worth" in LLMs or exponential growth?
In their revenue. I don't understand your question.
AI companies are raking in $200B a year in revenue, right now.
Because all they do is _burn_ money.
Oh- they do a lot of that too, loll, particularly the companies making the frontier models.
But beyond that, they also make money. Fucking boat loads of it.
Re: (Score:2)
In their revenue. I don't understand your question.
AI companies are raking in $200B a year in revenue, right now.
Hahhahah, no, they do not. What they rake in is stupid money from "investors".
Re: (Score:2)
Hahhahah, no, they do not.
Yes, they do. Of course that counts the DC and GPU hardware manufacturing side of the business as well.
What they rake in is stupid money from "investors".
No, venture capital is not revenue.
OpenAI makes almost $13B a year right now in revenue.
Their most recent round of VC was $40B
It's that big, precisely because they're pulling in the money they are.
I don't understand why your brain gets so religiously shut down whenever this topic comes up. Where you diddled by an AI when you were a child?
Re: (Score:2)
You are without insight. I am well aware of the applications of non-LLM AI. Go play somewhere else, idiot.
Re: (Score:2)
If only there were some other use for AI than chatbots. Maybe even something that might even get you a Nobel prize.
Re: (Score:2)
If only there were some other use for AI than chatbots. Maybe even something that might even get you a Nobel prize.
As that one probably went "woosh" over most people heads here, let me highlight here that Google Deepmind's AlphaFold2 developers won the Nobel prize for Chemistry last year: https://www.nobelprize.org/prizes/chemistry/2024/press-release/ [slashdot.org]
Hypocritical (Score:2)
Re: (Score:2)
I haven't seen them once demand that everyone else needs to share all their research.
I know they're arguing that copyright should allow for AI training as fair use, but that's not remotely the same thing.
Re: (Score:2)
Re: (Score:2)
They have not demanded that people hand their research over to them.
To make this hypocritical, someone would have to leak it, and then some AI company would have to use it for training and then they'd have to cry that their copyright has been violated. Then I'd be right with you laughing at them.
Re: Hypocritical (Score:2)
They have not demanded that people hand their research over to them.
I must have missed them saying they would pay for it.
Re: (Score:2)
You're trying to make a fatally flawed analogy here. Let it go.
As I said, you had no intention of getting anything straight. Your intention was to make a really really bad argument and hope that people were dumb enough to swallow it.
Begin the siloing! (Score:4, Insightful)
This seems like the opening move in a gambit where everyone starts holding back their research on AI in order to give their own company an edge. Honestly, I don't mind because with everyone siloing their research it means that development stagnation is right around the corner. This in turn will burst the already faltering AI bubble. I look forward to when AI is no longer under development for purely financial reasons.
Re: (Score:3)
When we ask, "are these guys going to actually be able to monetize this?" the answer is: "Yes". Because they already have.
Whether or not they're overvalued is certainly an open question, but the players are pulling in major money from actual people right now- they're not just floating on VC.
Gemini is poop (Score:1)
Don't worry, they are not sitting on any secrets that give Gemini a competitive edge, because it suuuuuuuucks.
Re: Gemini is poop (Score:2)
Re: (Score:2)
Actually way more impressed with Gemma 3, which is from the same people and downloadable. They have versions that will run on your home PC. But honestly, all of them spew bullshit.
There go the scientists (Score:2)
.
Not any more.