Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Gemini AI

deepmind.google

Post reply on HN
Print view View original

431–440 of 1001 posts

Re: Gemini AI

#431

Post by hulium »

Palmik wrote: post #324
albertzeyer wrote: post #42

So, better than GPT4 according to the benchmarks? Looks very interesting. Technical paper: https://goo.gle/GeminiPaper Some details: - 32k context length - efficient attention mechanisms (for e.g. multi-query attention (Shazeer, 2019)) - audio input via Universal Speech Model (USM) (Zhang et al., 2023) features - no audio output? (Figure 2) - visual encoding of Gemini models is inspired by our own foundational work o…

The table is *highly* misleading. It uses different methodologies all over the place. For MMLU, it highlights the CoT @ 32 result, where Ultra beats GPT4, but it loses to GPT4 with 5-shot, for example. For GSM8K it uses Maj1@32 for Ultra and 5-shot CoT for GPT4, etc. Then also, for some reason, it uses different metrics for Ultra and Pro, making them hard to compare. What a mess of a "paper".

Why is that misleading? It shows Gemini with CoT is the best known combination of prompt and LLM on MMLU.

They simply compare the prompting strategies that work best with each model. Otherwise it would be just a comparison of their response to specific prompt engineering.

Re: Gemini AI

#432

Post by rvnx »

m3kw9 wrote: post #412

You know who’s really f——-ed? Apple, they are now way behind google who is still behind OpenAI even with this.

No they are likely working on offline LLMs and custom chips so they'll be fine.

If you can run a large model locally for most of the cases, you won't want to use the Google Cloud services or OpenAI.

Re: Gemini AI

#433

Post by huqedato »

mg wrote: post #293

To test whether bard.google.com is already updated in your region, this prompt seems to work: Which version of Bard am I using? Here in Europe (Germany), I get: The current version is Bard 2.0.3. It is powered by the Google AI PaLM 2 model Considering that you have to log in to use Bard while Bing offers GPT-4 publicly and that Bard will be powered by Gemini Pro, which is not the version that they say beats GPT-4, it…

from Italy: "You are currently using the latest version of Bard, which is powered by a lightweight and optimized version of LaMDA, a research large language model from Google AI. This version of Bard is specifically designed for conversational tasks and is optimized for speed and efficiency. It is constantly being updated with new features and improvements, so you can be sure that you are always using the best possible version."

Re: Gemini AI

#434

Post by mensetmanusman »

atleastoptimal wrote: post #373
mensetmanusman wrote: post #349

Earlier quoted context omitted.

OpenAI had an almost five-year head-start with relevant data acquisition and sorting, which is the most important part of these models.

Google has the biggest proprietary moat of information of any company in the world I'm sure.

maybe it is too much? If you just train LLM's on the entire Internet, it will be mostly garbage.

Re: Gemini AI

#435

Post by seydor »

aantix wrote: post #401

Hmmm.. Seems like summarizing/extracting information from Youtube videos is a place where Bard/Gemini should shine. I asked it to give me "the best quotes from..." a person appearing in the video (they are explicitly introduced) and Bard says, "Unfortunately, I don't have enough information to process your request."

How about making youtube videos. People already do that

Re: Gemini AI

#436

Post by HlessClaudesman »

arnaudsm wrote: post #225
albertzeyer wrote: post #42

So, better than GPT4 according to the benchmarks? Looks very interesting. Technical paper: https://goo.gle/GeminiPaper Some details: - 32k context length - efficient attention mechanisms (for e.g. multi-query attention (Shazeer, 2019)) - audio input via Universal Speech Model (USM) (Zhang et al., 2023) features - no audio output? (Figure 2) - visual encoding of Gemini models is inspired by our own foundational work o…

I miss when ML scientific papers had actual science in them. Now they all feel like ads.

It's aimed at the AI pontification industrial complex.

Re: Gemini AI

#437

Post by dbrgn »

rvnx wrote: post #410
SeanAnderson wrote: post #340

Gemini Ultra isn't released yet and is months away still. Bard w/ Gemini Pro isn't available in Europe and isn't multi-modal, https://support.google.com/bard/answer/14294096 No public stats on Gemini Pro. (I'm wrong. Pro stats not on website, but tucked in a paper - https://storage.googleapis.com/deepmind-media/gemini/gemini_... ) I feel this is overstated hype. There is no competitor to GPT-4 being released today. I…

Not just Europe: also no Canada, China, Russia, United Kingdom, Switzerland, Bulgaria, Norway, Iceland, etc.

United Kingdom, Switzerland, Bulgaria, Norway and Iceland are all part of Europe.

Re: Gemini AI

#439

Post by bookmark1231 »

EZ-E wrote: post #358
milesward wrote: post #136

This demo is nuts: https://youtu.be/UIZAiXYceBI?si=8ELqSinKHdlGlNpX

Out of curiosity I fed ChatGPT 4 a few of the challenges through a photo (unclear if Gemini takes live video feed as input but GPT does not afaik) and it did pretty well. It was able to tell a duck was being drawn at an earlier stage before Gemini did. Like Gemini it was able to tell where the duck should go - to the left path to the swan. Because and I quote "because ducks and swans are both waterfowl, so the swan d…

The category similarity comment is amusing. My ChatGPT4 seems to have an aversion to technicality, so much that I’ve resorted to adding “treat me like an expert researcher and don’t avoid technical detail” in the prompt

Re: Gemini AI

#440

Post by emaciatedslug »

CamperBob2 wrote: post #301
ACS_Solver wrote: post #267

Earlier quoted context omitted.

To quote Gemini, what the quack! Even with the understanding that these are handpicked interactions that are likely to be among the system's best responses, that is an extremely impressive level of understanding and reasoning.

Calls for a new corollary to Clarke's Third Law. "Any sufficiently-advanced rigged demo is indistinguishable from magic."

And for the development team it's predicated upon Hoffsteader's Law.
Post reply on HN