Google is Using Anthropic's Claude To Improve Its Gemini AI

Contractors working to improve Google's Gemini AI are comparing its answers against outputs produced by Anthropic's competitor model Claude, TechCrunch reported Tuesday, citing internal correspondence. From the report: Google would not say, when reached by TechCrunch for comment, if it had obtained permission for its use of Claude in testing against Gemini.

As tech companies race to build better AI models, the performance of these models are often evaluated against competitors, typically by running their own models through industry benchmarks rather than having contractors painstakingly evaluate their competitors' AI responses. The contractors working on Gemini tasked with rating the accuracy of the model's outputs must score each response that they see according to multiple criteria, like truthfulness and verbosity. The contractors are given up to 30 minutes per prompt to determine whose answer is better, Gemini's or Claude's, according to the correspondence seen by TechCrunch.

duh

[dfghjk]

Yes, of course they are. Would that not be immediately obvious to do? And it doesn't need to be a competitor either.

Re:

[timeOday]

Car companies do actual teardowns of competitors' vehicles, and it's no secret. Barring patent infringement, there isn't a "no peeking!!" law.

Re:

[Seven Spirals]

They need to do something. Both Gemini and Claude absolutely suck donkey balls. ChatGPT and Grok are both leagues ahead of them. The main problem with Gemini is it's inability to answer questions without "context" where it absolutely fucking must lecture you about this or that before it even tries to answer. Claude is even worse in this respect and you can barely get it to answer at all without giving paragraphs of boilerplate bullshit as it "combats" "misinformation". People want the answer to their questi

Need laws

[buck-yar]

There should be laws that force these tech companies to disclose all the sources that were used to train them. But since the rich and powerful's lobbyists write the laws, I'm not holding my breath.

It is all intellectual theft anyways

[gweihir]

I doubt any LLM scammer will be willing to set a precedent here. It is pretty clear LLM output is not copyrightable as something the LLM produced anyways, but making that even clearer could backfire rather catastrophically for this scummy industry.

They all are...

[Inyu]

AI companies in general would definitely be motivated to mine each other's AI answers, at least for comparative analysis, at most for collecting primary data set forn training their AIs.

Because Gemini S*CKS!

[Anonymous Coward]

Google's Gemini is a weak, substandard product. It's not even worth using, and they know it.

My contact said they rushed to get Gemini out, assuming improvement would eventually compete with OpenAI, etc., but this isn't working out. So there is a ton of pressure; desperate measures.

This is what happens when siblings breed.

[RockDoctor]

Write that on a Post-It ,and stick it on your screen where the links to Claudini and Gemaude reside.

I'd get popcorn, if the stuff were edible while waiting for a monster to emerge.