Do we know what hardware they used for training? Google's own or Nvidia?
Technical report: https://storage.googleapis.com/deepmind-media/gemini/gemini_...
"We trained Gemini models using TPUv5e and TPUv4"
Gemini AI
131–140 of 1001 posts
Re: Gemini AI
#132One observation: Sundar's comments in the main video seem like he's trying to communicate "we've been doing this ai stuff since you (other AI companies) were little babies" - to me this comes off kind of badly, like it's trying too hard to emphasize how long they've been doing AI (which is a weird look when the currently publicly available SOTA model is made by OpenAI, not Google). A better look would simply be to sh…
In fairness, the performance/size ratio for models like BERT still gives GPT-3/4 and even Llama a run for it's money. Their tech isn't as product-ized as OpenAI's, but Tensorflow and it's ilk have been an essential part of driving actual AI adoption. The people I know in the robotics and manufacturing industries are forever grateful for the out-front work Google did to get the ball rolling.
You seem to be saying the same thing- Googles best work is in the past, their current offerings are underwhelming, even if foundational to the progress of others.
Re: Gemini AI
#133So it's basically just GPT-4, according to the benchmarks, with a slight edge for multimodal tasks (ie audio, video). Google does seem to be quite far behind, GPT-4 launched almost a year ago.
GPT-4 launched a little less than 9 months ago.
GPT-4 had finished training 8 months prior to release.
Re: Gemini AI
#134Earlier quoted context omitted.
That's for Ultra right? Which is an amazing accomplishment, but it sounds like I won't be able to access it for months. If I'm lucky.
Yep, the announcement is quite cheeky. Ultra is out sometime next year, with GPT-4 level capability. Pro is out now (?) with ??? level capability.
Pro benchmarks are here: https://storage.googleapis.com/deepmind-media/gemini/gemini_...
Sadly it's 3.5 quality, :(
Re: Gemini AI
#135I asked Bard, "Are you running Gemini Pro now?" And it told me, "Unfortunately, your question is ambiguous. "Gemini Pro" could refer to..." and listed a bunch of irrelevant stuff.
Is Bard not using Gemini Pro at time of writing? The blog post says, "Starting today, Bard will use a fine-tuned version of Gemini Pro for more advanced reasoning, planning, understanding and more."
(EDIT: it is... gave me a correct answer after some pushing. I just had to clarify that I meant "Bard". Not off to a great start)
It's weird, right? If I ask it:
Q: "Who are you?" A: "I am bard." Q: "Is Bard running Gemini Pro?" A: "Yes." Q: "Are you running Gemini Pro?" A: "Ambiguous Question."
Re: Gemini AI
#136This demo is nuts: https://youtu.be/UIZAiXYceBI?si=8ELqSinKHdlGlNpX
Re: Gemini AI
#137So it's basically just GPT-4, according to the benchmarks, with a slight edge for multimodal tasks (ie audio, video). Google does seem to be quite far behind, GPT-4 launched almost a year ago.
GPT-4 launched a little less than 9 months ago.
Yeah and GPT-4V is even younger (at least it was available later)
Re: Gemini AI
#138Bard still not available in Canada so i can't use it ¯\_(ツ)_/¯. Wonder why Google is the only one that can't release their model here.
Anthropic's Claude is still not available in Canada either. Anyone have insight into why its difficult to bring these AI models to Canada when on the surface its political and legal landscape isn't all that different from the US?
Google's embargo seemed to relate to their battle with the Canadian government over news. Given that they settled on that I'd expect its availability very soon.
Anthropic is a bit weird and it almost seems more like lazy gating. It's available in the US and UK, but no EU, no Canada, no Australia.
Re: Gemini AI
#139> For Gemini Ultra, we’re currently completing extensive trust and safety checks, including red-teaming by trusted external parties, and further refining the model using fine-tuning and reinforcement learning from human feedback (RLHF) before making it broadly available. > As part of this process, we’ll make Gemini Ultra available to select customers, developers, partners and safety and responsibility experts for ear…
Well not exactly. Not coming out until later when presumably GPT4 will have grown as much as well. So far each time, Google has failed to catch up to OpenAI. Hopefully they do however eventually.
Re: Gemini AI
#140> For Gemini Ultra, we’re currently completing extensive trust and safety checks, including red-teaming by trusted external parties, and further refining the model using fine-tuning and reinforcement learning from human feedback (RLHF) before making it broadly available. > As part of this process, we’ll make Gemini Ultra available to select customers, developers, partners and safety and responsibility experts for ear…
won't be available for regular devs until probably Q2 next year, OpenAI will have probably released GPT5 or whatever new model by then. And GPT4 was done training in 2022, the fact Google is thumping their chest about being 2 years behind a much smaller company is kind of pathetic