DeepSeek may have used Google's Gemini to train its latest model

https://techcrunch.com/2025/06/03/deepseek-may-have-used-googles-gemini-to-train-its-latest-model/

29•samsmithy•8mo ago

Comments

parineum•8mo ago

At this point, they all using each other because so much of the new content they are scraping for data is generated.

These models will converge and plateau because the datasets are only going to get worse as more of their content is incestuous.

wkat4242•8mo ago

Yes indeed some studies were already done on this.

sovietmudkipz•8mo ago

I recall that AI trained on AI output over many cycles eventually becomes something akin to noise texture as the output degrades rapidly.

Won’t most AI produced content put out into the public be human curated, thus heavily mitigating this degradation effect? If we’re going to see a full length AI generated movie it seems like humans will be heavily involved, hand holding the output and throwing out the AI’s nonsense.

AstroBen•8mo ago

Some will be heavily curated, by those who care about quality. This is a lot slower to produce, requires some expertise to do right, so there will be far less of it

The vast majority of content will be (is) the fastest and easiest to create - AI slop

jsheard•8mo ago

The default Llama 4 system prompt even instructs it to avoid using various ChatGPT-isms, presumably because they've already scraped so much GPT-generated material that it noticably skews their models output.

zackangelo•8mo ago

There might be a plateau coming but I’m not sure that will be the reason.

It seems counterintuitive but there is some research suggesting that using synthetic data might actually be productive.

jsheard•8mo ago

I think there's probably a distinction to be made between deliberate, careful use of synthetic data, as opposed to blindly scraping 1PB of LLM generated SEO spam and force-feeding it into a new model. Maybe the former is useful, but the latter... probably not.

philipwhiuk•8mo ago

> Distillation isn’t an uncommon practice, but OpenAI’s terms of service prohibit customers from using the company’s model outputs to build competing AI.

I have the absolute tiniest of violins for this given OpenAI's behaviour vs everyone else's terms of service.

sovietmudkipz•8mo ago

“Copyright must evolve into the 21century (…so that AI can legally steal everything produced by people”

And also “Don’t steal our AI!”

jsheard•8mo ago

The world is not prepared for the mental gymnastics that OpenAI/Google/etc will employ to defend their copyright if their big models ever get leaked.

bitpush•8mo ago

I see no evidence that Google is doing this. Any sources?

Zetaphor•8mo ago

I'm still unclear how they are able to claim this considering their raw thinking traces were never exposed to the end user, only summaries.

dev_hugepages•8mo ago

The raw thinking traces were shown previously in the API.

vb-8448•8mo ago

I wonder if at this point it really matters who used whose data ...

ksymph•8mo ago

Interesting. The tonal change has definitely been noticeable. It also seems a bit more succinct and precise with its word choice, less flowery. That does seem to be in line with Gemini's behavior.

hulitu•8mo ago

> Sam Paech, a Melbourne-based developer who creates “emotional intelligence” evaluations for AI, published what he claims is evidence that DeepSeek’s latest model was trained on outputs from Gemini. DeepSeek’s model, called R1-0528, prefers words and expressions similar to those that Google’s Gemini 2.5 Pro favors, said Paech in an X post.

And if you search for personal information of Android users, including location, sex, political orientation and location data, it is all there. /s

Styling: Search-Text and Other Highlight-Y Pseudo-Elements

Crypto firm accidentally sends $40B in Bitcoin to users

Magnetic fields can change carbon diffusion in steel

Fantasy football that celebrates great games

Show HN: Animalese

StrongDM's AI team build serious software without even looking at the code

John Haugeland on the failure of micro-worlds

Show HN: Velocity - Free/Cheaper Linear Clone but with MCP for agents

Corning Invented a New Fiber-Optic Cable for AI and Landed a $6B Meta Deal [video]

Show HN: XAPIs.dev – Twitter API Alternative at 90% Lower Cost

Near-Instantly Aborting the Worst Pain Imaginable with Psychedelics

Show HN: Nginx-defender – realtime abuse blocking for Nginx

The Super Sharp Blade

Smart Homes Are Terrible

What I haven't figured out

KPMG pressed its auditor to pass on AI cost savings

Open-source Claude skill that optimizes Hinge profiles. Pretty well.

First Proof

I squeezed a BERT sentiment analyzer into 1GB RAM on a $5 VPS

Kagi Translate

Building Interactive C/C++ workflows in Jupyter through Clang-REPL [video]

Tactical tornado is the new default

Full-Circle Test-Driven Firmware Development with OpenClaw

Automating Myself Out of My Job – Part 2

Dependency Resolution Methods

Crypto firm apologises for sending Bitcoin users $40B by mistake

Show HN: iPlotCSV: CSV Data, Visualized Beautifully for Free

There's no such thing as "tech" (Ten years later)

List of unproven and disproven cancer treatments

Me/CFS: The blind spot in proactive medicine (Open Letter)

Styling: Search-Text and Other Highlight-Y Pseudo-Elements

Crypto firm accidentally sends $40B in Bitcoin to users

Magnetic fields can change carbon diffusion in steel

Fantasy football that celebrates great games

Show HN: Animalese

StrongDM's AI team build serious software without even looking at the code

John Haugeland on the failure of micro-worlds

Show HN: Velocity - Free/Cheaper Linear Clone but with MCP for agents

Corning Invented a New Fiber-Optic Cable for AI and Landed a $6B Meta Deal [video]

Show HN: XAPIs.dev – Twitter API Alternative at 90% Lower Cost

Near-Instantly Aborting the Worst Pain Imaginable with Psychedelics

Show HN: Nginx-defender – realtime abuse blocking for Nginx

The Super Sharp Blade

Smart Homes Are Terrible

What I haven't figured out

KPMG pressed its auditor to pass on AI cost savings

Open-source Claude skill that optimizes Hinge profiles. Pretty well.

First Proof

I squeezed a BERT sentiment analyzer into 1GB RAM on a $5 VPS

Kagi Translate

Building Interactive C/C++ workflows in Jupyter through Clang-REPL [video]

Tactical tornado is the new default

Full-Circle Test-Driven Firmware Development with OpenClaw

Automating Myself Out of My Job – Part 2

Dependency Resolution Methods

Crypto firm apologises for sending Bitcoin users $40B by mistake

Show HN: iPlotCSV: CSV Data, Visualized Beautifully for Free

There's no such thing as "tech" (Ten years later)

List of unproven and disproven cancer treatments

Me/CFS: The blind spot in proactive medicine (Open Letter)

DeepSeek may have used Google's Gemini to train its latest model

Comments