On: Gemini AI

[aaronharnly]

I did some side-by-side comparisons of simple tasks (e.g. "Write a WCAG-compliant alternative text describing this image") with Bard vs GPT-4V.

Bard's output was significantly worse. I did my testing with some internal images so I can't share, but will try to compile some side-by-side from public images.

[IanCal → aaronharnly]

Bard with pro is apparently text only:

> Important: For now, Bard with our specifically tuned version of Gemini Pro works for text-based prompts, with support for other content types coming soon.

https://support.google.com/bard/answer/14294096

I'm in the UK and it's not available here yet - I really wish they'd be clearer about what I'm using, it's not the first time this has happened.

[sinuhe69 → IanCal]

You can ask Bard directly! Unlike ChatGPT, Bard can answer many things about itself.

[IanCal → sinuhe69]

It lies:

https://imgur.com/a/glPmXp3

I ask it if it's available in the uk and it says no. I say I'm in the uk and it tells me it's not Gemini then.

[aaronharnly → IanCal]

Huh! It has an image upload, and gives somewhat responsive, just not great, responses, so I'm a bit confused by that. So this is the existing Lens implementation?

[staticman2 → aaronharnly]

Bard has been capable of handling images for months.

[IanCal → aaronharnly]

Is palm2 multimodal?

[a_wild_dandan → aaronharnly]

As it should! Hopefully Gemini Ultra will be released in a month or two for comparison to GPT-4V.

[xfalcox → aaronharnly]

I'm researching using LLMs for alt-text suggestion for forum users, can you share your finding so far?

Outside of GPT-4V I had good first results with https://github.com/THUDM/CogVLM

[IanCal → xfalcox]

As a heads up, bard with gemini pro only works with text.