4 ms·

I tried this prompt in both Gemini 2.5 Pro, and in ChatGPT. "Draw me a timeline of all the dynasties of China. Imagine a horizontal line. Start from the leftmo

I tried this prompt in both Gemini 2.5 Pro, and in ChatGPT. "Draw me a timeline of all the dynasties of China. Imagine a horizontal line. Start from the leftmost point and draw segments for the start and end of each dynasty. For periods where multiple dynasties existed simultaneously draw parallel lines or boxes to represent the concurrent rule." Gemini's response: "I'm just a language model, so I can't help you with that." ChatGPT's response: an actual visual timeline.

renewiltord 1y ago
All the communities where people think LLMs are junk love Gemini. Makes me sceptical that the enthusiasm is useful signal. I found the full 2.0 useful for transcription of images. Very good OCR. But not a good assistant. Stalls often and once it has, loses context easily.
- thegeomaster 1y ago
  Is it possible that a community of people who are constantly pushing LLMs to their limits would be most aware of their limitations, and so more inclined to think they are junk? In terms of business utility, Google has had great releases ever since the 2.0 family. Their models have never missed some mark --- either a good price/performance ratio, insane speeds, novel modalities (they still have the only API for autoregressive image generation atm), state-of-the-art long context support and coding ability (Gemini 2.5), etc. However, most average users are using these models through a chat-like UI, or via generic tools like Cursor, which don't really optimize their pipelines to capture the strengths of different models. This way, it's very difficult to judge a model objectively. Just look at the obscene sycophancy exhibited by chatgpt-4o-latest and how it lifted LMArena scores.
  - renewiltord 1y ago
    Just the fact that everyone on HN is always telling us how LLMs are useless but that Gemini is the best of them convinces me of the opposite. No one who can't find a use for this technology is really informed on the subject. Hard to take them seriously.
ncr100 1y ago
Worked for me in 2.5 Flash, text only: https://g.co/gemini/share/bcc257f9b0a0 https://g.co/gemini/share/bcc257f9b0a0

renewiltord 1y ago

All the communities where people think LLMs are junk love Gemini. Makes me sceptical that the enthusiasm is useful signal. I found the full 2.0 useful for transcription of images. Very good OCR. But not a good assistant. Stalls often and once it has, loses context easily.

thegeomaster 1y ago
Is it possible that a community of people who are constantly pushing LLMs to their limits would be most aware of their limitations, and so more inclined to think they are junk? In terms of business utility, Google has had great releases ever since the 2.0 family. Their models have never missed some mark --- either a good price/performance ratio, insane speeds, novel modalities (they still have the only API for autoregressive image generation atm), state-of-the-art long context support and coding ability (Gemini 2.5), etc. However, most average users are using these models through a chat-like UI, or via generic tools like Cursor, which don't really optimize their pipelines to capture the strengths of different models. This way, it's very difficult to judge a model objectively. Just look at the obscene sycophancy exhibited by chatgpt-4o-latest and how it lifted LMArena scores.
- renewiltord 1y ago
  Just the fact that everyone on HN is always telling us how LLMs are useless but that Gemini is the best of them convinces me of the opposite. No one who can't find a use for this technology is really informed on the subject. Hard to take them seriously.

ncr100 1y ago

Worked for me in 2.5 Flash, text only: https://g.co/gemini/share/bcc257f9b0a0 https://g.co/gemini/share/bcc257f9b0a0